My research interests for over a decade have been within empirical and learning methods for language processing, often operating over rich linguistic representations, with particular emphasis on unsupervised semantic learning. In the last few years we have introduced textual entailment as a generic framework for applied semantic inference over texts.
With our colleagues, we organized the eight rounds of the Recognizing Textual Entailment (RTE) Challenges (2004-2013) under the PASCAL Network of Excellence and since 2008 as part the NIST Text Analysis Conference (TAC), which attracted dozens of research groups and became the primary forum for empirical evaluation of semantic inference systems.
At the Bar Ilan NLP group we develop computational models of textual entailment, including automatic knowledge acquisition, semantic inference, and information extraction and retrieval applications.
My core area of research is within natural language processing:
algorithms that act on human generated text. I am primarily interested in algorithms where the input is a sentence or a paragraph, and the output is a complex object such as a syntactic structure (parse tree), a graph denoting relations between words (co-reference, similarity), or a different sentence (translation, summarization). I tend to focus on core language components, which can serve as building blocks to other higher-level tasks such as textual entailment. My main tools are machine learning algorithms for structured prediction, and I prefer the resulting models to be practical, in the sense that they should be efficient and produce accurate results. If they can be simple it is even better, though “simple” in this context is a bit relative and somewhat of an acquired taste. Currently, I am focusing on better understanding of greedy learning algorithms, domain adaptation, and inference beyond the sentence boundary. Another emerging research direction is trying to model human sentence processing. I have a particular interest in algorithms that work specifically for Hebrew (or similar languages), and am also interested in applying the kinds of algorithms I use for language processing to structured tasks in computational biology.