Update README.md

README.md

@@ -4,10 +4,13 @@ tags:
 license: gpl-3.0
 ---
 ## A conversational agent with many personalities (PersonaGPT)
-PersonaGPT is an open-domain conversational agent capable of decoding _personalized_ responses based on input . 
-It builds on the [DialoGPT-medium](https://huggingface.co/microsoft/DialoGPT-medium) pretrained model based on the [GPT-2](https://github.com/openai/gpt-2) architecture. 
-This model is trained on the [Persona-Chat](https://arxiv.org/pdf/1801.07243) dataset, with added special tokens to better distinguish between conversational history and personality traits for dyadic conversations. Furthermore, some active learning was used to train the model to do _controlled_ decoding based on certain "action codes" (e.g., "talk about work", "ask about favorite music"). 
+PersonaGPT is an open-domain conversational agent designed to do 2 tasks:
+
+1. decoding _personalized_ responses based on input personality facts (the "persona" profile of the bot). 
+2. incorporating _turn-level goals_ into its responses through "action codes" (e.g., "talk about work", "ask about favorite music").
 
+It builds on the [DialoGPT-medium](https://huggingface.co/microsoft/DialoGPT-medium) pretrained model based on the [GPT-2](https://github.com/openai/gpt-2) architecture. 
+This model is trained on the [Persona-Chat](https://arxiv.org/pdf/1801.07243) dataset, with added special tokens to better distinguish between conversational history and personality traits for dyadic conversations. Furthermore, some active learning was used to train the model to do _controlled_ decoding using turn-level goals.
 
 ## Full Repo