CMPUT 650/497 - Computational Semantics

Overview

CMPUT 650 is a project-based graduate seminar on Computational Semantics and Artificial Intelligence (AI), with the focus on solving semantic tasks using Large Language Models (LLMs). There are no formal prerequisites, but students are expected to be familiar with Natural Language Processing (NLP), Machine Learning (ML), and vector semantics (embeddings). The students are expected to read research papers, think critically, and participate in group discussions. In addition to individual projects, students will participate in a joint team project, which is organized as a series of assignments, and provides opportunities for hands-on application of state-of-the-art NLP tools. Both the individual projects and the team project are aimed at subsequent peer-reviewed publication, with students having an option to participate as co-authors.

Objectives

The goal of the course is for the students to become familiar with the concepts and algorithms in Computational Semantics, as well as their relation to LLMs. The students will improve their research communication skills, including paper writing, reviewing, and presenting, and potentially even co-author a peer-reviewed publication.

Course Topics

  • lexical and sentence-level semantics
  • word senses and wordnets
  • semantic relations
  • lexical and textual similarity
  • translational equivalence
  • universality and language independence
  • large language models (LLMs)
  • AI-completeness

Course Work

  • Course project (50%)
    • Proposal (5%)
    • Interim report (5%)
    • Final report (25%)
    • Presentation (10%)
    • Peer reviews (5%)
  • Assignments (30%)
  • Paper presentations (10%)
  • Seminar discussions (10%)

References

Student projects from the last offering of CMPUT 650 (Wi24) have been included in the following papers:

UAlberta at SemEval-2024 Task 1: A Potpourri of Methods for Quantifying Multilingual Semantic Textual Relatedness and Similarity
Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024)

Semi-Automated Construction of Sense-Annotated Datasets for Practically Any Language
Proceedings of the 31st International Conference on Computational Linguistics (COLING 2025)

Course Details
Loading ...