Student research opportunities

Latent Probabilistic Analysis/Inference on Text

Project Code: CECS_816

This project is available at the following levels:
Masters, PhD
Please note that this project is only for higher degree (postgraduate) applicants.

Keywords:

text entailment, sentiment analysis, topic models, machine learning

Supervisor:

Dr Wray Buntine

Outline:

Tasks such as inferring sentiment or mood from text, or inferring discourse or conversational structure from text are performing inference about the text using auxiliary resources such as word lists and/or the syntactic/semantic structures. Text Entailment is a recent research area concerned more with extracting factual information rather than sentiment. How can we set these tasks up as probabilistic inference on latent variables, inferring over the grammatical structures and linguistic resources? Part of the problem here is the availability of tagged data. Little tagged data is available, and invariably it is only partially tagged (e.g., tagged as a whole, not individual sentences or phrases). So what is needed to deal with partially tagged data? How do we set this up as a semi-supervised learning problem?

NOTE 1: not really an Honours project, but some variation could work.

NOTE 2: feel free to be creative and suggest an alternative!