School of Computer Science
Technical Workshop Series: Introduction to Language Models
Presenter: Zahra Taherikhonakdar
Date: Wednesday, December 6th, 2023
Time: 1:00 PM - 2:00 PM
Location: 4th Floor (Workshop space) at 300 Ouellette Avenue (School of Computer Science Advanced Computing Hub)
LATECOMERS WILL NOT BE ADMITTED once the presentation has begun.
Abstract:
A language model is a probabilistic representation of a natural language, capable of estimating the likelihood of word sequences based on the textual data it was trained on, encompassing one or more languages. The initial significant statistical language model emerged in 1980, followed by IBM's 'Shannon-style' experiments in the same decade. These experiments involved observing and analyzing human subjects' performance in predicting or correcting text, leading to the identification of potential avenues for enhancing language modeling.
These models find application in various tasks, such as speech recognition (aiding in avoiding predictions of improbable or nonsensical sequences), machine translation, natural language generation (producing text that resembles human language more closely), optical character recognition, handwriting recognition, grammar induction, and information retrieval.
Workshop Outline:
In this workshop, I am going to introduce the techniques that are used in LM:
- Introduction to N-grams
- Estimating N gram Probabilities
- Evaluation and Perplexity
- Generalization
Prerequisites:
Computer Science knowledge
Biography:
Zahra is a PhD student at the University of Windsor. Her research is in the area of Information Retrieval. Mainly her research is about improving query refinement to make search engines retrieve the most related documents based on users' initial query.