Linguistic Analysis Meets Artificial Intelligence: Developing Discourse Parsing Solutions for Dialogue Understanding

In recent years, there has been a noticeable surge in the prevalence of dialogues across various domains. The widespread adoption of dialogues, also facilitated by technologies such as ChatGPT, Bard, and Claude, underscores the considerable significance of comprehending their intricate structures. The analysis of dialogue structures offers invaluable benefits, particularly in sectors like healthcare, where analyzing patient-doctor dialogues can enhance diagnoses and treatments. Similarly, in legal proceedings, parsing courtroom dialogues can boost the justice system’s efficiency. However, the complex organizational structure underlying dialogic interactions poses a challenge for computational modeling. Notably, previous automated tools, primarily designed for monologues, have struggled to adapt to the nuances of dialogues. Handling dialogues requires parsing not only the individual contributions but also the dynamic interplay between speakers, involving shifts in perspectives and topic transitions. This project aims to develop solutions capable of extracting pragmatic and semantic relationships between discourse units within unformatted written or oral dialogues. Natural language processing and machine learning techniques will be employed to analyze lexical, syntactic, and semantic information, thereby identifying discourse segments and classifying the nature of connections between them. The expected outcomes include analytical tools for deeper linguistic examination and improved machine comprehension of idea progression.

Faculty Supervisor:

Giuseppe Carenini

Student:

Partner:

University of Salerno

Discipline:

Computer science

Sector:

Artificial Intelligence; Information and Communications Technology

University:

The University of British Columbia

Program: