I am a PhD student at the Centre for Language Technology at Macquarie University, Sydney. My work is supervised by Dr. Mark Dras.

My PhD research project is in the area of Textual Entailment. The field has become popular quite recently, namely in the beginning of 2004 when the first Recognizing Textual Entailment Challenge has been announced. At this time I have just started my PhD. Being attracted by the task I have submitted my first entailment recognition engine to the competion (link).

My interests in NLP methods changed quite a bit since that and the focus of my PhD as well. Rather than concentrating on participating in the RTE Challenge I have decided to contribute to the task by approching entailment pairs of different types, see, for example, ESESR entailment type.

I use statistical methods, Machine learning, coding mostly in Java in Eclipse environment using GATE and WEKA libraries. Often use WordNet and its API for different tasks.

Recognizing Textual Entailment. What's that?

Recognizing Textual Entailment (RTE) is a task where, given two text snippets, the goal is to determine whether the meaning of one text snippet can be inferred from the meaning of the other. The first of the text snippets in such a pair is referred to as the text and the other one as the hypothesis. The pair of text and hypothesis is called a text-hypothesis pair or entailment pair, with the two names considered to be synonymous.
The text is usually much longer than the hypothesis. It can be represented by one or more coherent sentences,while the hypothesis is usually one short sentence. It is the meaning of the hypothesis that might or might not be entailed from the text. Thus, given a text-hypothesis pair, we recognize the relation between the meanings of the text and the hypothesis in the pair as a true entailment if the meaning of the hypothesis is entailed from the meaning of the text. Otherwise, we recognize the relation between the meanings of the texts as a false entailment.

Text: Japan s Kyodo news agency said the US could be ready to set up a liaison office the lowest level of diplomatic representation in Pyongyang if it abandons its nuclear program.
Hypothesis: Kyodo news agency is based in Japan.

The broader context of our work is to investigate different ways of subclassifying entailment pairs. In this framework, a generic system would have additional special components that take care of the special subclasses of entailment pairs. Such a component is involved when a pair of its subclass is recognized. Note that we do not envisage classifying all the entailment pairs to give a partitioning of the space, a probably infeasible task. We suggest dividing into classes the entailment pairs that are problematic for all the state-of-the-art generic systems and develop separate RTE algorithms for these particular classes. The broad question that we aim to answer is whether this will improve the overall performance of the RTE engine.

Contact:

E6A-341, Department of Computing,
Macquarie University, North Ryde, NSW-2109
e-mail: elena@ics.mq.edu.au
web-page: http://www.comp.mq.edu.au/~elena/