Information about ongoing research at the lab.

High Dimensional Models of Lexical Semantics

Over the last 10 years there has been a growing interest in the field of psycholinguistics in mathematical models of the mental lexicon. One such model is the HAL model proposed by Burgess (1996). We are are interested in exploring these statistical models, and finding more ways to use them to explain psycholinguistic phenomena.

Here in the lab we have been working on exploring the parameter space of these types of models to better understand how they perform. In the process we have made our own implementation of HAL and a family of models that resemble HAL. We call this software HiDEx.

HiDEx: The High Dimensional Explorer

HiDEx is a software package developed at the Westbury Lab to conduct research into HAL and other related models of lexical semantic memory. It has the following features:
  • It can process very large corpora (multi-billion word corpora).
  • It can use different window sizes,  window weighting vectors, normalization algorithms and similarity metrics algorithms.
  • It calculate inter-word distances as well as the N closest neighbors.

As of Dec 8th, 2008, HiDEx was released under the GPL v3. This means that HiDEx is free software. Please read the licence before using HiDEx to learn about the implications of the GPL v3 licence.

Download HiDEx!

If you have any questions about HiDEx, please contact Cyrus Shaoul.
For more information about HiDEx and other software projects, please read our publications.


©2008  WestburyLab   chrisw at ualberta dot ca