File size: 4,516 Bytes
fbb1cfd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2e2fcca
 
 
 
 
e9aeca3
2e2fcca
fbb1cfd
 
 
 
 
 
 
 
 
 
 
 
e9aeca3
 
 
 
 
 
 
 
 
 
fbb1cfd
e9aeca3
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
import numpy as np
import spacy
import nltk
import tensorflow as tf
import streamlit as st
from streamlit_extras.add_vertical_space import add_vertical_space
from bs4 import BeautifulSoup
from gensim.models import Word2Vec
from nltk.tokenize import word_tokenize
from warnings import filterwarnings
filterwarnings('ignore')



def streamlit_config():

    # page configuration
    st.set_page_config(page_title='Document Classification', layout='centered')

    # page header transparent color
    page_background_color = """

    <style>



    [data-testid="stHeader"] 

    {

    background: rgba(0,0,0,0);

    }



    </style>

    """
    st.markdown(page_background_color, unsafe_allow_html=True)

    # title and position
    st.markdown(f'<h1 style="text-align: center;">Financial Document Classification</h1>',
                unsafe_allow_html=True)
    add_vertical_space(4)


def text_extract_from_html(html_file):

    # Read the uploaded HTML file
    html_content = html_file.read().decode('utf-8')

    # Parse the HTML Content
    soup = BeautifulSoup(html_content, 'html.parser')

    # Extract the Text
    text = soup.get_text()

    # Split the Text and Remove Unwanted Space
    result = [i.strip() for i in text.split()]

    return result


def text_processing(text):

    # spaCy Engine
    nlp = spacy.load('en_core_web_lg')

    # Process the Text with spaCy
    doc = nlp(' '.join(text))

    # Tokenization, Lemmatization, and Remove Stopwords, punctuation, digits
    token_list = [
                  token.lemma_.lower().strip()
                  for token in doc
                  if token.text.lower() not in nlp.Defaults.stop_words and token.text.isalpha()
                 ]

    if len(token_list) > 0:
        return ' '.join(token_list)
    else:
        return 'empty'
    

def sentence_embeddings(sentence):

    # split the sentence into separate words
    words = word_tokenize(sentence)                         

    # load the trained model
    model = Word2Vec.load('word2vec_model.bin')

    # get the vectors of each words
    vectors = [model.wv[word] for word in words if word in model.wv]
    
    if vectors:
        # return the average of vectors
        return np.mean(vectors, axis=0)

    else:
        # we set the model parameter in training ---> vector_size = 300
        return np.zeros(model.vector_size)


def prediction(input_file):

    # Extract the Text from HTML Document
    html_content = text_extract_from_html(input_file)

    # Preprocess the Text
    preprocessed_text = text_processing(html_content)

    # Text Convert into Embeddings
    features = sentence_embeddings(preprocessed_text)

    # Reshape the features into match the expected input shape of Model
    features = np.expand_dims(features, axis=0)
    features = np.expand_dims(features, axis=2)

    # Convert into Tensors
    features_tensors = tf.convert_to_tensor(features, dtype=tf.float32)

    # Load the Model and Prediction
    model = tf.keras.models.load_model('model.h5', custom_objects = {'Orthogonal': tf.keras.initializers.Orthogonal})
    prediction = model.predict(features_tensors)

    # Find the Maximum Probability Value
    target_label = np.argmax(prediction)

    # Find the Target_Label Name
    target = {0:'Balance Sheets', 1:'Cash Flow', 2:'Income Statement', 3:'Notes', 4:'Others'}
    predicted_class = target[target_label]

    # Find the Confidence
    confidence = round(np.max(prediction)*100, 2)

    add_vertical_space(2)
    st.markdown(f'<h4 style="text-align: center; color: orange;">{confidence}% Match Found</h4>', 
                    unsafe_allow_html=True)
    
    add_vertical_space(1)
    st.markdown(f'<h3 style="text-align: center; color: green;">Prdicted Class = {predicted_class}</h3>', 
                    unsafe_allow_html=True)



# Streamlit Configuration Setup
streamlit_config()
    

# File uploader to upload the HTML file
input_file = st.file_uploader('Upload an HTML file', type='html')

if input_file is not None:

    try:
        # Predict the Input_HTML_File_Class
        prediction(input_file)

    except:
        # Check 'punkt' Already Downloaded or Not
        try:
            nltk.data.find('tokenizers/punkt')
        except LookupError:
            nltk.download('punkt')

        # Predict the Input_HTML_File_Class
        prediction(input_file)