S-CASE Blog | Natural Language Processing

We recently got another research paper on our work in S-CASE accepted  at a conference on natural language processing. The accepted paper describes our efforts on improving a parsing model that can  automatically map software requirements written in natural language to formal representations based on semantic roles.

State-of-the-art semantic role labelling systems require large  annotated corpora to achieve full performance. Unfortunately, such  corpora are expensive to produce and often do not generalise well  across domains. Even in domain, errors are often made where syntactic  information does not provide sufficient cues. In this paper, we mitigate both of these problems by employing distributional word representations gathered from unlabelled data. The rationale for this  approach lies in the so-called distributional hypothesis by Zellig Harris, which states that words that occur in the same contexts tend  to have similar meanings.

While straight-forward word representations  of predicates and arguments have already been shown to be useful for  semantic analysis tasks, we show that further gains can be achieved by composing representations that model the interaction between predicate  and argument, and capture full argument spans.

Leave a Reply

Your email address will not be published.