Dialogue System Implementation

This repository houses the implementation of a dialogue system I developed as part of my Master's degree dissertation in Natural Language Processing (NLP).

Motivation

The primary motivation behind this project was to investigate the significant influence of a speaker's intention on the progression and dynamics of a conversation.

Framework and Model

• Framework: PyTorch
• Models:
  • Bidirectional Gated Recurrent Unit (GRU)
  • Linear Chain Conditional Random Field (CRF)

Preprocessing Pipeline

1. Tokenization: Breaking down text into individual words or tokens.
2. Replacement: Substituting specific words or characters with predefined ones.
3. Padding: Standardizing sentences or sequences to a uniform length.
4. Numerization:
   • Word2Vec: Employed for converting words into numerical vectors.
   • Char2Vec: Utilized for managing unknown words.
5. Compressing: For lengthy sentences, we compressed them in a consistent manner.

Key Insights

From our experiments, we derived:

• The intention and emotion of the speaker play a pivotal role in determining the flow and direction of the conversation.
• The impact of the speaker's intention on the conversation's progression becomes increasingly noticeable as the dialogue extends.

Acknowledgements

probability theory, machine learning

https://library.korea.ac.kr/detail/?cid=CAT000045999362&ctype=t