Finding lexically related Maltese word pairs in corpora

Abstract

This project introduces a weakly-supervised system which is able to extract patterns from the context by using a predefined set of word pairs in a particular lexical relation, and use the previously extracted patterns to be able to gather new word pairs in that specific relation. This is done by using methods such as regular expressions to extract patterns and the Espresso algorithm (Pantel and Pennacchiotti, 2006) to give weighting to each pattern. In its current state, the system returns rather promising results, showing potential as a preliminary system of its kind, which will allow researchers to build on top of what has already been done to improve its accuracy further. It is quite likely that data which this system is able to gather might be used in the near future to be able to build a Maltese WordNet.