output
stringlengths
1
483
hypotheses
listlengths
5
5
input
stringlengths
20
2.42k
instruction
stringclasses
1 value
because you are making counts for word forms right
[ "Because you make encounters for word forms, right?", "Because you make encounts for word forms, right?", "Because you're making counts for word forms, right?", "Because you make accounts for word forms, right?", "Because you making counts for word forms, right?" ]
['because you make encounters for word forms right', 'because you make encounts for word forms right', 'because you are making counts for word forms right', 'because you make accounts for word forms right', 'because you making counts for word forms right']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
alright okay
[ "All right, okay.", "All right okay.", "All right. Okay.", "All right, OK.", "All right, Okay." ]
['all right okay', 'all right okay', 'all right okay', 'all right ok', 'all right okay']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so how much do they display and starting position at where the or maybe the mid position of it
[ "So how much the display and starting position at where the or maybe the mid position of it.", "So how much the display and starting position at where there or maybe the mid position of it.", "So how much of the display and starting position at where the or maybe the mid position of it.", "So how much the display and starting position at where there, or maybe the mid position of it.", "So how much of the display and starting position at where there or maybe the mid position of it." ]
['so how much the display and starting position at where the or maybe the mid position of it', 'so how much the display and starting position at where there or maybe the mid position of it', 'so how much of the display and starting position at where the or maybe the mid position of it', 'so how much the display and starting position at where there or maybe the mid position of it', 'so how much of the display and starting position at where there or maybe the mid position of it']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
cause you would probably be quite likely if they are talking about a conference or a person
[ "Because you'd probably be quite likely if they're talking about a conference or a person.", "Cause you'd probably be quite likely if they're talking about a conference or a person.", "'Cause you'd probably be quite likely if they're talking about a conference or a person.", "Because you'd probably be quite likely, if they're talking about a conference or a person.", "Cause you'd probably be quite likely, if they're talking about a conference or a person." ]
['because you would probably be quite likely if they are talking about a conference or a person', 'cause you would probably be quite likely if they are talking about a conference or a person', 'cause you would probably be quite likely if they are talking about a conference or a person', 'because you would probably be quite likely if they are talking about a conference or a person', 'cause you would probably be quite likely if they are talking about a conference or a person']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
that i do not know
[ "that I do not know.", "that I don't know.", "that I do not know,", "that I do not know", "that I do not know?" ]
['that i do not know', 'that i do not know', 'that i do not know', 'that i do not know', 'that i do not know']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
just thinking what is the simp so
[ "Just thinking what's this simp so.", "Just thinking what's this simp So.", "Just thinking, what's this simp? So.", "Just thinking what's this simp. So.", "Just thinking, what's this simp so." ]
['just thinking what is this simp so', 'just thinking what is this simp so', 'just thinking what is this simp so', 'just thinking what is this simp so', 'just thinking what is this simp so']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
for the information density and we just tie that to the existing utterances and tie them to the existing speaker changes
[ "For the information density, we just tie that to the existing utterances and tie them to the existing speaker changes.", "For the information density we just tie that to the existing utterances and tie them to the existing speaker changes.", "For the information density, we just tie that to the existing utterances, and tie them to the existing speaker changes.", "For the information density. We just tie that to the existing utterances and tie them to the existing speaker changes.", "For the information density and we just tie that to the existing utterances and tie them to the existing speaker changes." ]
['for the information density we just tie that to the existing utterances and tie them to the existing speaker changes', 'for the information density we just tie that to the existing utterances and tie them to the existing speaker changes', 'for the information density we just tie that to the existing utterances and tie them to the existing speaker changes', 'for the information density we just tie that to the existing utterances and tie them to the existing speaker changes', 'for the information density and we just tie that to the existing utterances and tie them to the existing speaker changes']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i mean otherwise i can give you the code for loading a dictionary
[ "I mean otherwise I can give you the code for loading a dictionary.", "I mean, otherwise I can give you the code for loading a dictionary.", "I mean, otherwise, I can give you the code for loading a dictionary.", "I mean otherwise, I can give you the code for loading a dictionary.", "Otherwise, I can give you the code for loading a dictionary." ]
['i mean otherwise i can give you the code for loading a dictionary', 'i mean otherwise i can give you the code for loading a dictionary', 'i mean otherwise i can give you the code for loading a dictionary', 'i mean otherwise i can give you the code for loading a dictionary', 'otherwise i can give you the code for loading a dictionary']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
because i i in my outline i talked about using the discourse acts 1st
[ "Because I in my outline I talked about using the discourse acts first.", "Because in my outline, I talked about using the discourse acts first.", "Because I in my outline, I talked about using the discourse acts first.", "Because in my outline I talked about using the discourse acts first.", "Because I, in my outline, I talked about using the discourse acts first." ]
['because i in my outline i talked about using the discourse acts 1st', 'because in my outline i talked about using the discourse acts 1st', 'because i in my outline i talked about using the discourse acts 1st', 'because in my outline i talked about using the discourse acts 1st', 'because i in my outline i talked about using the discourse acts 1st']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
they would probably they would like you more if you s s h into another computer compress it there and then sort of copy it into the into the gateway machine
[ "They'd probably they'd like you more if you SSH into another computer, compress it there and then sort of copy it into the gateway machine.", "They probably they'd like you more if you SSH into another computer, compress it there and then sort of copy it into the gateway machine.", "They'd probably they'd like you more if you SSH into another computer, compress it there, and then sort of copy it into the gateway machine.", "They'd probably they'd like you more if you SSH into another computer, compress it there and then sort of copy it into their into the gateway machine.", "They'd probably they'd like you more if you SSH into another computer, compress it there, and then sort of copy it into their into the gateway machine." ]
['they would probably they would like you more if you ssh into another computer compress it there and then sort of copy it into the gateway machine', 'they probably they would like you more if you ssh into another computer compress it there and then sort of copy it into the gateway machine', 'they would probably they would like you more if you ssh into another computer compress it there and then sort of copy it into the gateway machine', 'they would probably they would like you more if you ssh into another computer compress it there and then sort of copy it into their into the gateway machine', 'they would probably they would like you more if you ssh into another computer compress it there and then sort of copy it into their into the gateway machine']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
and for all together like the display itself
[ "And for altogether, like the display itself.", "And for altogether like the display itself.", "And for all together, like the display itself.", "And for all together like the display itself.", "And for altogether, like, the display itself." ]
['and for altogether like the display itself', 'and for altogether like the display itself', 'and for all together like the display itself', 'and for all together like the display itself', 'and for altogether like the display itself']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i the
[ "I wondered that.", "I wonder that.", "I wanted that.", "I wonder that?", "I wondered." ]
['i wondered that', 'i wonder that', 'i wanted that', 'i wonder that', 'i wondered']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so that we we take the the ready made parts and just see how we get them work together in the interface the way we want and and then we have a working prototype
[ "So that we take the ready made parts and just see how we get them worked together in the interface the way we wanted. And then we have a working prototype.", "So that we take the ready made parts and just see how we get them work together in the interface the way we want it. And then we have a working prototype.", "So that we take the ready made parts and just see how we get them worked together in the interface the way we want it. And then we have a working prototype.", "So that we take the ready made parts and just see how we get them to work together in the interface the way we want it. And then we have a working prototype.", "So that we take the ready made parts and just see how we get them to work together in the interface the way we wanted. And then we have a working prototype." ]
['so that we take the ready made parts and just see how we get them worked together in the interface the way we wanted and then we have a working prototype', 'so that we take the ready made parts and just see how we get them work together in the interface the way we want it and then we have a working prototype', 'so that we take the ready made parts and just see how we get them worked together in the interface the way we want it and then we have a working prototype', 'so that we take the ready made parts and just see how we get them to work together in the interface the way we want it and then we have a working prototype', 'so that we take the ready made parts and just see how we get them to work together in the interface the way we wanted and then we have a working prototype']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i was
[ "I was.", "Shows.", "She was.", "Showers.", "Show us." ]
['i was', 'shows', 'she was', 'showers', 'show us']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
and they are all numbered
[ "A little number.", "A little numbered.", "A narrow number.", "a little number.", "And they're all numbered." ]
['a little number', 'a little numbered', 'a narrow number', 'a little number', 'and they are all numbered']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
well yeah that is not it is not an issue
[ "Well yeah, that is not, it's not an issue.", "Well yeah, that is not it's not an issue.", "But yeah, that is not it's not an issue.", "But yeah, that is not, it's not an issue.", "Well yeah, that is not, is not an issue." ]
['well yeah that is not it is not an issue', 'well yeah that is not it is not an issue', 'but yeah that is not it is not an issue', 'but yeah that is not it is not an issue', 'well yeah that is not is not an issue']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
and you
[ "and you.", "in you.", "and EU.", "and you?", "in Europe." ]
['and you', 'in you', 'and eu', 'and you', 'in europe']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
but it will need to be calculated at word level though because otherwise there will not be enough occurrences of the terms to make any meaningful sense
[ "They'll need to be calculated at word level though because otherwise there won't be enough occurrences of the terms to make any meaningful sense.", "They'll need to be calculated at word level though, because otherwise there won't be enough occurrences of the terms to make any meaningful sense.", "They'll need to be calculated at word level, though, because otherwise there won't be enough occurrences of the terms to make any meaningful sense.", "They'll need to be calculated at word level, though, because otherwise, there won't be enough occurrences of the terms to make any meaningful sense.", "They'll need to be calculated at word level though, because otherwise, there won't be enough occurrences of the terms to make any meaningful sense." ]
['they will need to be calculated at word level though because otherwise there will not be enough occurrences of the terms to make any meaningful sense', 'they will need to be calculated at word level though because otherwise there will not be enough occurrences of the terms to make any meaningful sense', 'they will need to be calculated at word level though because otherwise there will not be enough occurrences of the terms to make any meaningful sense', 'they will need to be calculated at word level though because otherwise there will not be enough occurrences of the terms to make any meaningful sense', 'they will need to be calculated at word level though because otherwise there will not be enough occurrences of the terms to make any meaningful sense']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
okay
[ "Okay.", "Ok.", "OK.", "Okay,", "okay." ]
['okay', 'ok', 'ok', 'okay', 'okay']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i think we are easier if we if it is sitting on the x m l than if it is sitting on the s q l stuff because if it is sitting on the x m l
[ "I think we are easier if we if it's sitting on the XML than if it's sitting on the SQL stuff because if it's sitting on the XML.", "I think we are easier if we if it's sitting on the Xml than if it's sitting on the SQL stuff because if it's sitting on the Xml.", "I think we are easier if we if it's sitting on the XML than if it's sitting on the SQL stuff, because if it's sitting on the XML.", "I think we are easier if we, if it's sitting on the XML than if it's sitting on the SQL stuff because if it's sitting on the XML.", "I think we are easier if we, if it's sitting on the Xml than if it's sitting on the SQL stuff because if it's sitting on the Xml." ]
['i think we are easier if we if it is sitting on the xml than if it is sitting on the sql stuff because if it is sitting on the xml', 'i think we are easier if we if it is sitting on the xml than if it is sitting on the sql stuff because if it is sitting on the xml', 'i think we are easier if we if it is sitting on the xml than if it is sitting on the sql stuff because if it is sitting on the xml', 'i think we are easier if we if it is sitting on the xml than if it is sitting on the sql stuff because if it is sitting on the xml', 'i think we are easier if we if it is sitting on the xml than if it is sitting on the sql stuff because if it is sitting on the xml']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
and i think it even has its own annotation like digits or something
[ "And I think it even has a its own annotation like digits or something.", "And I think it even has a its own annotation, like digits or something.", "And I think it even has, uh, its own annotation like digits or something.", "And I think it even has its own annotation like digits or something.", "And I think it even has, uh, its own annotation, like digits or something." ]
['and i think it even has a its own annotation like digits or something', 'and i think it even has a its own annotation like digits or something', 'and i think it even has its own annotation like digits or something', 'and i think it even has its own annotation like digits or something', 'and i think it even has its own annotation like digits or something']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Mm.", "Mhm.", "Hm.", "Mmm.", "Yeah." ]
['', '', 'hm', '', 'yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i think we will have to buffer the audio
[ "I think we have to buffer the audio.", "I think we'll have to buffer the audio.", "I think we have to buffer the audience.", "I think we'll have to buffer the audience.", "I think we have to buffer the audio" ]
['i think we have to buffer the audio', 'i think we will have to buffer the audio', 'i think we have to buffer the audience', 'i think we will have to buffer the audience', 'i think we have to buffer the audio']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
it is just we are half way through the project time table
[ "It's just we are halfway through the project timetable.", "It's just, we are halfway through the project timetable.", "And it's just we are halfway through the project timetable.", "And just we are halfway through the project timetable.", "That's just we are halfway through the project timetable." ]
['it is just we are halfway through the project timetable', 'it is just we are halfway through the project timetable', 'and it is just we are halfway through the project timetable', 'and just we are halfway through the project timetable', 'that is just we are halfway through the project timetable']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
okay
[ "Yeah.", "You can't see it.", "You have to see.", "You can't see that.", "Yeah, I'm just here." ]
['yeah', 'you can not see it', 'you have to see', 'you can not see that', 'yeah i am just here']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so only words per meeting series
[ "So only words per meeting series.", "So, only words per meeting series.", "So only words. Per meeting series.", "So, uh, only words, um, per meeting series.", "So, uh, only words. Um, per meeting series." ]
['so only words per meeting series', 'so only words per meeting series', 'so only words per meeting series', 'so only words per meeting series', 'so only words per meeting series']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
that like the highest level of merging it is not the whole icsi corpus but individual series
[ "That like at the highest level of merging, it's not the whole exicorpus, but individual series.", "That like at the highest level of merging, it's not the whole exicorpus but individual series.", "That like at the highest level of merging, it's not the whole Exi Corpus, but individual series.", "That like at the highest level of merging, it's not the whole Exi Corpus but individual series.", "That like, at the highest level of merging, it's not the whole Exi Corpus, but individual series." ]
['that like at the highest level of merging it is not the whole exicorpus but individual series', 'that like at the highest level of merging it is not the whole exicorpus but individual series', 'that like at the highest level of merging it is not the whole exi corpus but individual series', 'that like at the highest level of merging it is not the whole exi corpus but individual series', 'that like at the highest level of merging it is not the whole exi corpus but individual series']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
and it will just string them all together with maybe i do not know 10th of a 2nd silence in between each one or something like that
[ "It'll just string them all together with like maybe, I don't know, tenth of a second silence in between each one or something like that.", "It'll just string them all together with like maybe I don't know, tenth of a second silence in between each one or something like that.", "It'll just string them all together with, like, maybe, I don't know, tenth of a second silence in between each one or something like that.", "So it'll just string them all together with like maybe I don't know, tenth of a second silence in between each one or something like that.", "So it'll just string them all together with like maybe, I don't know, tenth of a second silence in between each one or something like that." ]
['it will just string them all together with like maybe i do not know 10th of a 2nd silence in between each one or something like that', 'it will just string them all together with like maybe i do not know 10th of a 2nd silence in between each one or something like that', 'it will just string them all together with like maybe i do not know 10th of a 2nd silence in between each one or something like that', 'so it will just string them all together with like maybe i do not know 10th of a 2nd silence in between each one or something like that', 'so it will just string them all together with like maybe i do not know 10th of a 2nd silence in between each one or something like that']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
alright
[ "", ".", "Thank you.", "Mr President,", "Mm." ]
['', '', 'thank you', 'mister president', '']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
or are we doing it on word level like the information density calculation
[ "are we doing it on world level, like the information density calculation?", "are we doing it on world level like the information density calculation?", "are we doing it at world level, like the information density calculation?", "are we doing it on world level, like the information density calculation.", "are we doing it on a world level, like the information density calculation?" ]
['are we doing it on world level like the information density calculation', 'are we doing it on world level like the information density calculation', 'are we doing it at world level like the information density calculation', 'are we doing it on world level like the information density calculation', 'are we doing it on a world level like the information density calculation']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah that is just an i d or something
[ "Yeah, That's just an I. D. or something.", "Yeah, That's just an ID or something.", "Yeah that's just an I. D. or something.", "Yeah, That's just an I D or something.", "Yeah, That's just an I. D. Or something." ]
['yeah that is just an i d or something', 'yeah that is just an id or something', 'yeah that is just an i d or something', 'yeah that is just an i d or something', 'yeah that is just an i d or something']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yes.", "Yeah,", "yeah.", "Yeah, yeah." ]
['yeah', 'yes', 'yeah', 'yeah', 'yeah yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah i am pretty sure it is already there
[ "Yeah, I'm pretty sure it's already there.", "Yeah, Pretty sure it's already there.", "And I'm pretty sure it's already there.", "Yeah. Pretty sure it's already there.", "Yeah, Place Sure it's already there." ]
['yeah i am pretty sure it is already there', 'yeah pretty sure it is already there', 'and i am pretty sure it is already there', 'yeah pretty sure it is already there', 'yeah place sure it is already there']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
file with whatever hierarchical structure we want
[ "File with whatever hierarchical structure we want.", "Fire with whatever hierarchical structure we want.", "Far with whatever hierarchical structure we want.", "file with whatever hierarchical structure we want.", "far with whatever hierarchical structure we want." ]
['file with whatever hierarchical structure we want', 'fire with whatever hierarchical structure we want', 'far with whatever hierarchical structure we want', 'file with whatever hierarchical structure we want', 'far with whatever hierarchical structure we want']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i am not biased
[ "I'm not biased.", "I am not biased.", "I'm not bias.", "I am not bias.", "am not biased." ]
['i am not biased', 'i am not biased', 'i am not bias', 'i am not bias', 'am not biased']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
the other hand i mean it should not be like should be like 50 mega byte in ram or something it should not be massive should it
[ "The other hand, I mean, it shouldn't like should be like fifty megabytes in rum or something. It shouldn't be massive, should it?", "The other hand, I mean, it shouldn't like should be like fifty megabyte in rum or something. It shouldn't be massive, should it?", "The other hand, I mean, it shouldn't be like, should be like fifty megabytes in rum or something. It shouldn't be massive, should it?", "The other hand, I mean, it shouldn't like, should be like fifty megabytes in rum or something. It shouldn't be massive, should it?", "The other hand, I mean, it shouldn't, like, should be like fifty megabytes in rum or something. It shouldn't be massive, should it?" ]
['the other hand i mean it should not like should be like 50 megabytes in rum or something it should not be massive should it', 'the other hand i mean it should not like should be like 50 megabyte in rum or something it should not be massive should it', 'the other hand i mean it should not be like should be like 50 megabytes in rum or something it should not be massive should it', 'the other hand i mean it should not like should be like 50 megabytes in rum or something it should not be massive should it', 'the other hand i mean it should not like should be like 50 megabytes in rum or something it should not be massive should it']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
just let us be careful about that because
[ "Just let's be careful about that, because.", "Just let's be careful about that because.", "Just let's be careful about that. Because.", "Just let us be careful about that, because.", "Just, let's be careful about that, because." ]
['just let us be careful about that because', 'just let us be careful about that because', 'just let us be careful about that because', 'just let us be careful about that because', 'just let us be careful about that because']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
and speaker
[ "And speaker.", "and speaker.", "Speaker.", "And Speaker.", "and Speaker." ]
['and speaker', 'and speaker', 'speaker', 'and speaker', 'and speaker']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yeah?", "Yes.", "Yeah,", "yeah." ]
['yeah', 'yeah', 'yes', 'yeah', 'yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
well we alternatively we can probably just make another directory on the beefy scratch space
[ "Alternatively, we can probably just make another directory on the Beefy Scratch space.", "Will be alternatively we can probably just make another directory on the Beefy Scratch space.", "Alternatively, we can probably just make another directory on the Beefy Scratch Space.", "Will be alternatively, we can probably just make another directory on the Beefy Scratch space.", "Alternatively we can probably just make another directory on the Beefy Scratch space." ]
['alternatively we can probably just make another directory on the beefy scratch space', 'will be alternatively we can probably just make another directory on the beefy scratch space', 'alternatively we can probably just make another directory on the beefy scratch space', 'will be alternatively we can probably just make another directory on the beefy scratch space', 'alternatively we can probably just make another directory on the beefy scratch space']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
then it is b like a more coherent framework
[ "Then it's like a more coherent framework.", "then it's like a more coherent framework.", "Then it's, like, a more coherent framework.", "Then it's like, a more coherent framework.", "Then, it's like a more coherent framework." ]
['then it is like a more coherent framework', 'then it is like a more coherent framework', 'then it is like a more coherent framework', 'then it is like a more coherent framework', 'then it is like a more coherent framework']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so you ha yeah
[ "So, yeah.", "So yeah.", "So you yeah.", "So, Yes.", "So, yeah, yeah." ]
['so yeah', 'so yeah', 'so you yeah', 'so yes', 'so yeah yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
that everyone is
[ "that everyone's, um,", "the everyone's, um,", "the everyone's. um,", "the everyone's um,", "that everyone's um," ]
['that everyone is', 'the everyone is', 'the everyone is', 'the everyone is', 'that everyone is']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
there there are time stamps for well
[ "There there are time sums um for less.", "There there are time sums for less.", "There there are time sums um for less", "There there are time sums, um, for less.", "There, there are time sums, um, for less." ]
['there there are time sums for less', 'there there are time sums for less', 'there there are time sums for less', 'there there are time sums for less', 'there there are time sums for less']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah i think i think those segments for each utterance are split up
[ "Yeah, I think I think those segments for each utterance were split up.", "Yeah, I think I think those segments for each utterance are split up.", "Yeah, I think, I think those segments for each utterance were split up.", "Yeah, I think, I think those segments for each utterance are split up.", "Yeah I think I think those segments for each utterance are split up." ]
['yeah i think i think those segments for each utterance were split up', 'yeah i think i think those segments for each utterance are split up', 'yeah i think i think those segments for each utterance were split up', 'yeah i think i think those segments for each utterance are split up', 'yeah i think i think those segments for each utterance are split up']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
but if they give us access to it here 0 sitting on a dice machine
[ "But if they give us access to it here, sitting on a dice machine,", "But if they give us access to it here, sitting on a dice machine.", "But if they give us access to it here, sitting on a dice machine", "but if they give us access to it here, sitting on a dice machine,", "but if they give us access to it here, sitting on a dice machine." ]
['but if they give us access to it here sitting on a dice machine', 'but if they give us access to it here sitting on a dice machine', 'but if they give us access to it here sitting on a dice machine', 'but if they give us access to it here sitting on a dice machine', 'but if they give us access to it here sitting on a dice machine']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
should be
[ "Should it be?", "Should that be?", "That'd be.", "Should I be?", "Is that beat?" ]
['should it be', 'should that be', 'that would be', 'should i be', 'is that beat']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
how long would it take to make the frequency counts with a java hash table
[ "How long would it take to make the frequency counts with the Java hash table?", "How long would it take to make the frequency counts with the Java Hash table?", "How long would it take to make the frequency counts with the Java Hatch table?", "How long would it take to make the frequency counts with the java hash table?", "How long would it take to make the frequency counts with the Java Hashtabel?" ]
['how long would it take to make the frequency counts with the java hash table', 'how long would it take to make the frequency counts with the java hash table', 'how long would it take to make the frequency counts with the java hatch table', 'how long would it take to make the frequency counts with the java hash table', 'how long would it take to make the frequency counts with the java hashtabel']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "yeah.", "Yeah,", "Yes.", "Yea." ]
['yeah', 'yeah', 'yeah', 'yes', 'yea']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yeah,", "Yes.", "yeah.", "Yeah?" ]
['yeah', 'yeah', 'yes', 'yeah', 'yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
wait wait wait
[ "Wait, wait, wait.", "Wait wait wait.", "wait wait wait.", "Wait, wait.", "Wait, wait, wait, wait." ]
['wait wait wait', 'wait wait wait', 'wait wait wait', 'wait wait', 'wait wait wait wait']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah maybe
[ "Yeah, Maybe.", "Yeah, maybe.", "Yeah. Maybe.", "Yeah maybe.", "Yeah, maybe?" ]
['yeah maybe', 'yeah maybe', 'yeah maybe', 'yeah maybe', 'yeah maybe']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah and if you make that structurally very similar to the the le like one level down
[ "And if you make that structurally very similar to the like one level down.", "And if you make that structurally very similar to the one level down.", "And if you make that structurally very similar to the, like one level down.", "Yeah, and if you make that structurally very similar to the like one level down.", "Yeah, and if you make that structurally very similar to the, like one level down." ]
['and if you make that structurally very similar to the like one level down', 'and if you make that structurally very similar to the one level down', 'and if you make that structurally very similar to the like one level down', 'yeah and if you make that structurally very similar to the like one level down', 'yeah and if you make that structurally very similar to the like one level down']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah and x m l files
[ "Yeah, The XML files.", "Yeah, and the XML files.", "Yeah, And the XML files.", "Yeah, the XML files.", "Yeah, and XML files." ]
['yeah the xml files', 'yeah and the xml files', 'yeah and the xml files', 'yeah the xml files', 'yeah and xml files']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
that sounds quite reasonable
[ "It sounds quite reasonable.", "sounds quite reasonable.", "That sounds quite reasonable.", "It sounds quite reasonable,", "Sounds quite reasonable." ]
['it sounds quite reasonable', 'sounds quite reasonable', 'that sounds quite reasonable', 'it sounds quite reasonable', 'sounds quite reasonable']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
no do not bother
[ "I don't bother.", "I don't know.", "I don't believe that.", "I don't know about that.", "No, I don't bother." ]
['i do not bother', 'i do not know', 'i do not believe that', 'i do not know about that', 'no i do not bother']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
you have got no way of getting in and extracting just that word
[ "We've got no way of getting in and extracting just that word.", "You've got no way of getting in and extracting just that word.", "You got no way of getting in and extracting just that word.", "We've got no way of getting in, and extracting just that word.", "we've got no way of getting in and extracting just that word." ]
['we have got no way of getting in and extracting just that word', 'you have got no way of getting in and extracting just that word', 'you got no way of getting in and extracting just that word', 'we have got no way of getting in and extracting just that word', 'we have got no way of getting in and extracting just that word']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so i would not need it
[ "So I wouldn't need it.", "So, I wouldn't need it.", "so I wouldn't need it.", "So I wouldn't need it", "So i wouldn't need it." ]
['so i would not need it', 'so i would not need it', 'so i would not need it', 'so i would not need it', 'so i would not need it']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
wait are we are we using this for the for the for the do for the weighting in the end now this this measure you are calculating
[ "Are we are we using this um for the for the for the for the waiting in the end now this this measure you're calculating?", "Are we are we using this for the for the for the for the waiting in the end now this this measure you're calculating?", "Wait, are we are we using this for the for the for the for the waiting in the end now this this measure you're calculating?", "Wait, are we are we using this um for the for the for the for the waiting in the end now this this measure you're calculating?", "Are we are we using this um for the for the for the for the waiting in the end now this this measure you're calculating." ]
['are we are we using this for the for the for the for the waiting in the end now this this measure you are calculating', 'are we are we using this for the for the for the for the waiting in the end now this this measure you are calculating', 'wait are we are we using this for the for the for the for the waiting in the end now this this measure you are calculating', 'wait are we are we using this for the for the for the for the waiting in the end now this this measure you are calculating', 'are we are we using this for the for the for the for the waiting in the end now this this measure you are calculating']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yes.", "yeah.", "Yeah,", "Yeah, yeah." ]
['yeah', 'yes', 'yeah', 'yeah', 'yeah yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so i guess that would be solvable if not
[ "So I guess that would be solvable if not.", "So I guess that would be solvable, if not.", "So, I guess that would be solvable if not.", "So I guess that would be solvable if not?", "So i guess that would be solvable if not." ]
['so i guess that would be solvable if not', 'so i guess that would be solvable if not', 'so i guess that would be solvable if not', 'so i guess that would be solvable if not', 'so i guess that would be solvable if not']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
only words
[ "Only works.", "only works.", "Only words.", "Only works?", "only words." ]
['only works', 'only works', 'only words', 'only works', 'only words']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Mm.", "Yes.", "Mhm.", "Mm," ]
['yeah', '', 'yes', '', '']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
that means you have to through every single utterance in a series of 70 hours of meetings
[ "That means you have to go through every single utterance in a series of seventy hours of meetings.", "That means, you have to go through every single utterance in a series of seventy hours of meetings.", "that means you have to go through every single utterance in a series of seventy hours of meetings.", "That means you have to go through every single utterance, in a series of seventy hours of meetings.", "That means you have to go through every single utterance in a series of seventy hours of meeting." ]
['that means you have to go through every single utterance in a series of 70 hours of meetings', 'that means you have to go through every single utterance in a series of 70 hours of meetings', 'that means you have to go through every single utterance in a series of 70 hours of meetings', 'that means you have to go through every single utterance in a series of 70 hours of meetings', 'that means you have to go through every single utterance in a series of 70 hours of meeting']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
there is still some merit on within utterances cutting out stuff which clearly is not relevant at all and that maybe also for the audio we would have to do
[ "There's still some merit on within utterances cutting out stuff which clearly isn't relevant at all and that maybe also for the audio we'd have to do.", "There's still some merit on within utterances cutting out stuff which clearly isn't relevant at all. And that maybe also for the audio we'd have to do.", "There's still some merit on within utterances, cutting out stuff which clearly isn't relevant at all. And that maybe also for the audio we'd have to do.", "There's still some merit on within utterances, cutting out stuff which clearly isn't relevant at all, and that maybe also for the audio we'd have to do.", "There's still some merit on within utterances, cutting out stuff, which clearly isn't relevant at all. And that maybe also for the audio we'd have to do." ]
['there is still some merit on within utterances cutting out stuff which clearly is not relevant at all and that maybe also for the audio we would have to do', 'there is still some merit on within utterances cutting out stuff which clearly is not relevant at all and that maybe also for the audio we would have to do', 'there is still some merit on within utterances cutting out stuff which clearly is not relevant at all and that maybe also for the audio we would have to do', 'there is still some merit on within utterances cutting out stuff which clearly is not relevant at all and that maybe also for the audio we would have to do', 'there is still some merit on within utterances cutting out stuff which clearly is not relevant at all and that maybe also for the audio we would have to do']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
well you need the raw frequency as well
[ "Well, you need the raw frequency as well.", "Well you need the raw frequency as well.", "Well, n you need the raw frequency as well.", "Well n you need the raw frequency as well.", "Well, you need the roar frequency as well." ]
['well you need the raw frequency as well', 'well you need the raw frequency as well', 'well n you need the raw frequency as well', 'well n you need the raw frequency as well', 'well you need the roar frequency as well']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
i think there are quite a lot of numbers in the beginning where n there is no time stamp for the numbers
[ "Um, I think there are quite a lot of numbers in the beginning where there's no timestamp for the numbers.", "Um, I think there are quite a lot of numbers in the beginning where there's no time stamp for the numbers.", "Um, I think there are quite a lot of numbers in the beginning where there is no timestamp for the numbers.", "Um, I think there are quite a lot of numbers in the beginning where there is no time stamp for the numbers.", "Um, I think there are quite a lot of numbers in the beginning when there's no time stamp for the numbers." ]
['i think there are quite a lot of numbers in the beginning where there is no timestamp for the numbers', 'i think there are quite a lot of numbers in the beginning where there is no time stamp for the numbers', 'i think there are quite a lot of numbers in the beginning where there is no timestamp for the numbers', 'i think there are quite a lot of numbers in the beginning where there is no time stamp for the numbers', 'i think there are quite a lot of numbers in the beginning when there is no time stamp for the numbers']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so you could just
[ "so you could just", "So you could just", "so that you could just", "so you could just,", "so you can just" ]
['so you could just', 'so you could just', 'so that you could just', 'so you could just', 'so you can just']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah if like i d i do not know anything about audio and i have never seen the player
[ "Yeah, if like, I don't know anything about audio and I have never seen the player.", "Yeah, if like I don't know anything about audio and I have never seen the player.", "Yeah, if, like, I don't know anything about audio and I have never seen the player.", "Yeah, if like, I don't know anything about Audion. I have never seen the player.", "Yeah, if, like, I don't know anything about Audion. I have never seen the player." ]
['yeah if like i do not know anything about audio and i have never seen the player', 'yeah if like i do not know anything about audio and i have never seen the player', 'yeah if like i do not know anything about audio and i have never seen the player', 'yeah if like i do not know anything about audion i have never seen the player', 'yeah if like i do not know anything about audion i have never seen the player']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
but it would be nice to have some basic system which just displays some stuff
[ "But it would be nice to have some basic system which just displays some stuff.", "But it would be nice to have some basic system, which just displays some stuff.", "But it would be nice to have some basic system which just display some stuff.", "but it would be nice to have some basic system which just displays some stuff.", "But, It would be nice to have some basic system which just displays some stuff." ]
['but it would be nice to have some basic system which just displays some stuff', 'but it would be nice to have some basic system which just displays some stuff', 'but it would be nice to have some basic system which just display some stuff', 'but it would be nice to have some basic system which just displays some stuff', 'but it would be nice to have some basic system which just displays some stuff']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
okay
[ "Like,", "Like.", "Okay.", "like,", "I can't." ]
['like', 'like', 'okay', 'like', 'i can not']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Hm.", "Mm.", "Yes.", "Hmm." ]
['yeah', 'hm', '', 'yes', '']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah in the beginning as well sometimes i think
[ "Yeah, in the beginning as well, sometimes I think.", "Yeah, in the beginning as well sometimes I think.", "Yeah, in the beginning as well sometimes, I think.", "Yeah, in the beginning as well, sometimes, I think.", "Yeah in the beginning as well sometimes I think." ]
['yeah in the beginning as well sometimes i think', 'yeah in the beginning as well sometimes i think', 'yeah in the beginning as well sometimes i think', 'yeah in the beginning as well sometimes i think', 'yeah in the beginning as well sometimes i think']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
does anyone want to like just sit with me and like play for 3 hours with nite x m l at some point
[ "Does anyone want to like, just sit with me and like, play for three hours with Nitex Mel at some point?", "Does anyone want to, like, just sit with me and, like, play for three hours with Nitex Mel at some point?", "Does anyone want to, like, just sit with me and, like, play for three hours with Night Ex Mail at some point?", "Does anyone want to like, just sit with me and like, play for three hours with Night Ex Mail at some point?", "Does anyone want to, like, just sit with me and, like, play for three hours with Night XML at some point?" ]
['does anyone want to like just sit with me and like play for 3 hours with nitex mel at some point', 'does anyone want to like just sit with me and like play for 3 hours with nitex mel at some point', 'does anyone want to like just sit with me and like play for 3 hours with night ex mail at some point', 'does anyone want to like just sit with me and like play for 3 hours with night ex mail at some point', 'does anyone want to like just sit with me and like play for 3 hours with night xml at some point']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yes i am going to build a dictionary then from that
[ "Yes, I'm gonna build a dictionary then from that.", "Yes, I'm going to build a dictionary then from that.", "Yes I'm gonna build a dictionary then from that.", "Yes I'm going to build a dictionary then from that.", "Yes, I'm gonna build a dictionary, then, from that." ]
['yes i am going to build a dictionary then from that', 'yes i am going to build a dictionary then from that', 'yes i am going to build a dictionary then from that', 'yes i am going to build a dictionary then from that', 'yes i am going to build a dictionary then from that']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so it it seems that the data structure is not a big problem and that basically we do not have to have all these massive discussions of how we exactly interact with the data structure because most of our work is not done with that data structure in memory in the browser
[ "So it seems that the data structure isn't a big problem and that basically we don't have to have all these massive discussions of how we exactly interact with the data structure because most of our work isn't done with that data structure in memory in the browser.", "It seems that the data structure isn't a big problem and that basically we don't have to have all these massive discussions of how we exactly interact with the data structure because most of our work isn't done with that data structure in memory in the browser.", "So it seems that the data structure isn't a big problem, and that basically we don't have to have all these massive discussions of how we exactly interact with the data structure, because most of our work isn't done with that data structure in memory in the browser.", "So it seems that the data structure isn't a big problem and that basically we don't have to have all these massive discussions of how we exactly interact with the data structure, because most of our work isn't done with that data structure in memory in the browser.", "So it seems that the data structure isn't a big problem, and that basically we don't have to have all these massive discussions of how we exactly interact with the data structure because most of our work isn't done with that data structure in memory in the browser." ]
['so it seems that the data structure is not a big problem and that basically we do not have to have all these massive discussions of how we exactly interact with the data structure because most of our work is not done with that data structure in memory in the browser', 'it seems that the data structure is not a big problem and that basically we do not have to have all these massive discussions of how we exactly interact with the data structure because most of our work is not done with that data structure in memory in the browser', 'so it seems that the data structure is not a big problem and that basically we do not have to have all these massive discussions of how we exactly interact with the data structure because most of our work is not done with that data structure in memory in the browser', 'so it seems that the data structure is not a big problem and that basically we do not have to have all these massive discussions of how we exactly interact with the data structure because most of our work is not done with that data structure in memory in the browser', 'so it seems that the data structure is not a big problem and that basically we do not have to have all these massive discussions of how we exactly interact with the data structure because most of our work is not done with that data structure in memory in the browser']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
we have the the nite x m l framework with all its functionality for synchronizing through all the different levels whenever there is a change whenever something is moving forward and stuff
[ "We have the Knight XML framework with all its functionality for synchronizing through all the different levels whenever there's a change, whenever something's moving forward and stuff.", "We have the Knight XML framework with all its functionality for synchronizing through all the different levels, whenever there's a change, whenever something's moving forward and stuff.", "We have the Knight XML framework with all its functionality for synchronizing through all the different levels whenever there's a change whenever something's moving forward and stuff.", "We have the night XML framework with all its functionality for synchronizing through all the different levels whenever there's a change, whenever something's moving forward and stuff.", "We have the knight XML framework with all its functionality for synchronizing through all the different levels whenever there's a change, whenever something's moving forward and stuff." ]
['we have the knight xml framework with all its functionality for synchronizing through all the different levels whenever there is a change whenever something is moving forward and stuff', 'we have the knight xml framework with all its functionality for synchronizing through all the different levels whenever there is a change whenever something is moving forward and stuff', 'we have the knight xml framework with all its functionality for synchronizing through all the different levels whenever there is a change whenever something is moving forward and stuff', 'we have the night xml framework with all its functionality for synchronizing through all the different levels whenever there is a change whenever something is moving forward and stuff', 'we have the knight xml framework with all its functionality for synchronizing through all the different levels whenever there is a change whenever something is moving forward and stuff']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
like do they look like they are utterances numbers
[ "Like do they look like their utterances numbers?", "Like do they look like they're utterances numbers?", "Like do they look like they're utterances numbers.", "Like do they look like their utterances, numbers.", "Like do they look like they're utterances, numbers." ]
['like do they look like their utterances numbers', 'like do they look like they are utterances numbers', 'like do they look like they are utterances numbers', 'like do they look like their utterances numbers', 'like do they look like they are utterances numbers']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
when you click on a segment it is going to have like words or whatever that are important
[ "When you click on a segment, it's gonna have like words or whatever that are important.", "When you click on a segment it's gonna have like words or whatever that are important.", "When you click on a segment, it's going to have like words or whatever that are important.", "When you click on a segment it's going to have like words or whatever that are important.", "When you click on a segment, it's gonna have like, words or whatever that are important." ]
['when you click on a segment it is going to have like words or whatever that are important', 'when you click on a segment it is going to have like words or whatever that are important', 'when you click on a segment it is going to have like words or whatever that are important', 'when you click on a segment it is going to have like words or whatever that are important', 'when you click on a segment it is going to have like words or whatever that are important']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah and and a topic is basically they are just on the i d probably with a start time or something
[ "And the topics basically they're just an ID probably with a start time or something.", "And the topics basically they're just an ID, probably with a start time or something.", "And the topics basically, they're just an ID, probably with a start time or something.", "And the topics, basically, they're just an ID, probably with a start time or something.", "Yeah, and the topics basically they're just an ID, probably with a start time or something." ]
['and the topics basically they are just an id probably with a start time or something', 'and the topics basically they are just an id probably with a start time or something', 'and the topics basically they are just an id probably with a start time or something', 'and the topics basically they are just an id probably with a start time or something', 'yeah and the topics basically they are just an id probably with a start time or something']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah i reckon you can just mean it over the sentence
[ "Yeah, I reckon you can just mean it over the sentence.", "Yeah. I reckon you can just mean it over the sentence.", "Yeah I reckon you can just mean it over the sentence.", "Yeah, I reckon you could just mean it over the sentence.", "Yeah. I reckon you could just mean it over the sentence." ]
['yeah i reckon you can just mean it over the sentence', 'yeah i reckon you can just mean it over the sentence', 'yeah i reckon you can just mean it over the sentence', 'yeah i reckon you could just mean it over the sentence', 'yeah i reckon you could just mean it over the sentence']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "yeah.", "Yes.", "Yeah,", "yeah," ]
['yeah', 'yeah', 'yes', 'yeah', 'yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah i know
[ "Yeah, and I.", "Yeah and I.", "Yeah, And I.", "Yeah, and I,", "Yeah, and I" ]
['yeah and i', 'yeah and i', 'yeah and i', 'yeah and i', 'yeah and i']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yeah,", "yeah.", "Yes.", "No." ]
['yeah', 'yeah', 'yeah', 'yes', 'no']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
the right hand corner far right
[ "the right hand corner, far right.", "the right hand corner far right.", "right hand corner, far right.", "the right hand corner, far right?", "the right hand corner far right?" ]
['the right hand corner far right', 'the right hand corner far right', 'right hand corner far right', 'the right hand corner far right', 'the right hand corner far right']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah i think it is quite easy after the
[ "Yeah, I think it's quite easy. I think it is.", "Yeah, I think it's quite easy, perhaps it is.", "Yeah, I think it's quite easy. perhaps it is.", "Yeah. I think it's quite easy. I think it is.", "Yeah, I think it's quite easy. Yep, It is." ]
['yeah i think it is quite easy i think it is', 'yeah i think it is quite easy perhaps it is', 'yeah i think it is quite easy perhaps it is', 'yeah i think it is quite easy i think it is', 'yeah i think it is quite easy yep it is']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yes
[ "Yes.", "Yes,", "yes.", "Yes?", "Yeah." ]
['yes', 'yes', 'yes', 'yes', 'yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
wa
[ "What?", "What", "But.", "what?", "Wait." ]
['what', 'what', 'but', 'what', 'wait']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
we do not have to
[ "We don't have to.", "we don't have to.", "We do not have to.", "we do not have to.", "You don't have to." ]
['we do not have to', 'we do not have to', 'we do not have to', 'we do not have to', 'you do not have to']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so far i extracted the dura durations
[ "So far, I extracted the durations.", "So far I extracted the durations.", "So far, I extracted the duration.", "So far I extracted the duration.", "So far, I extracted the duration space." ]
['so far i extracted the durations', 'so far i extracted the durations', 'so far i extracted the duration', 'so far i extracted the duration', 'so far i extracted the duration space']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so so the utterances are tied to topic segments
[ "So the utterances are tied to topic segments.", "So, the utterances are tied to topic segments.", "So, so the utterances are tied to topic segments.", "So so the utterances are tied to topic segments.", "So the utterances are tied to topics segments." ]
['so the utterances are tied to topic segments', 'so the utterances are tied to topic segments', 'so so the utterances are tied to topic segments', 'so so the utterances are tied to topic segments', 'so the utterances are tied to topics segments']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah but w it do not you have to like go sort of like for in a document versus the whole thing
[ "Yeah, but don't you have to like go sort of like for in a document versus the whole thing.", "Yeah, but don't you have to like go sort of like for in a document versus the whole thing?", "Yeah, but don't you have to like go sort of like four in a document versus the whole thing.", "Yeah, but don't you have to like go sort of like four in a document versus the whole thing?", "Yeah but don't you have to like go sort of like for in a document versus the whole thing." ]
['yeah but do not you have to like go sort of like for in a document versus the whole thing', 'yeah but do not you have to like go sort of like for in a document versus the whole thing', 'yeah but do not you have to like go sort of like 4 in a document versus the whole thing', 'yeah but do not you have to like go sort of like 4 in a document versus the whole thing', 'yeah but do not you have to like go sort of like for in a document versus the whole thing']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
well yeah
[ "Well, yeah.", "Well yeah.", "Well, Yeah.", "Well. Yeah.", "Well Yeah." ]
['well yeah', 'well yeah', 'well yeah', 'well yeah', 'well yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yes.", "Yeah,", "yeah.", "Yea." ]
['yeah', 'yes', 'yeah', 'yeah', 'yea']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so they still get all the data and just they internally say 0 no this is less than 3 and i am not going to display it or something
[ "So they still get all the data and just they internally say, oh no, this is less than three and I'm not going to display it or something.", "So they still get all the data and just they internally say oh no this is less than three and I'm not going to display it or something.", "So they still get all the data and just they internally say, oh no this is less than three and I'm not going to display it or something.", "So they still get all the data and just they internally say oh no, this is less than three and I'm not going to display it or something.", "So they still get all the data and just they internally say, oh no, this is less than three and I'm not gonna display it or something." ]
['so they still get all the data and just they internally say 0 no this is less than 3 and i am not going to display it or something', 'so they still get all the data and just they internally say 0 no this is less than 3 and i am not going to display it or something', 'so they still get all the data and just they internally say 0 no this is less than 3 and i am not going to display it or something', 'so they still get all the data and just they internally say 0 no this is less than 3 and i am not going to display it or something', 'so they still get all the data and just they internally say 0 no this is less than 3 and i am not going to display it or something']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah but i think actually like at the moment the integration comes 1st i mean it is sort of at the moment the building the browser comes 1st
[ "Yeah, but I think actually like at the moment the integration comes first. I mean, it's sort of at the moment the building, the browser comes first.", "Yeah, but I think actually like at the moment the integration comes first. I mean, at sort of at the moment the building, the browser comes first.", "Yeah, but I think actually like at the moment, the integration comes first. I mean, it's sort of at the moment, the building, the browser comes first.", "Yeah, but I think actually like at the moment, the integration comes first. I mean, at sort of at the moment, the building, the browser comes first.", "Yeah, but I think actually, like, at the moment the integration comes first. I mean, it's sort of at the moment the building, the browser comes first." ]
['yeah but i think actually like at the moment the integration comes 1st i mean it is sort of at the moment the building the browser comes 1st', 'yeah but i think actually like at the moment the integration comes 1st i mean at sort of at the moment the building the browser comes 1st', 'yeah but i think actually like at the moment the integration comes 1st i mean it is sort of at the moment the building the browser comes 1st', 'yeah but i think actually like at the moment the integration comes 1st i mean at sort of at the moment the building the browser comes 1st', 'yeah but i think actually like at the moment the integration comes 1st i mean it is sort of at the moment the building the browser comes 1st']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "yeah.", "They are.", "There.", "Yeah, yeah." ]
['yeah', 'yeah', 'they are', 'there', 'yeah yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
so if it is from s well actually i think if it is mounted you can directly burn from there
[ "So if it's from well actually I think if it's mounted you can directly burn from there.", "So if it's from well actually, I think if it's mounted, you can directly burn from there.", "So if it's from well actually I think if it's mounted, you can directly burn from there.", "So if it's from well, actually, I think if it's mounted, you can directly burn from there.", "So if it's from well, actually I think if it's mounted, you can directly burn from there." ]
['so if it is from well actually i think if it is mounted you can directly burn from there', 'so if it is from well actually i think if it is mounted you can directly burn from there', 'so if it is from well actually i think if it is mounted you can directly burn from there', 'so if it is from well actually i think if it is mounted you can directly burn from there', 'so if it is from well actually i think if it is mounted you can directly burn from there']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
like you think directly feeding
[ "like you think directly feeding", "like, you think directly feeding", "Like, you think directly feeding", "Like you think directly feeding", "Like you think, directly feeding" ]
['like you think directly feeding', 'like you think directly feeding', 'like you think directly feeding', 'like you think directly feeding', 'like you think directly feeding']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.
yeah
[ "Yeah.", "Yeah,", "yeah.", "Yes.", "Yeah?" ]
['yeah', 'yeah', 'yeah', 'yes', 'yeah']
The following text contains 5-best hypotheses from an Automatic Speech Recognition system. As part of a speech recognition task, please perform error correction on the hypotheses to generate the most accurate transcription of the spoken text.