ReSim: System for the semantic representation of interrogative sentences

The objective of this project was the development of an automatic model of semantic representation for interrogative sentences based on Jackendoff’s (1990) representation language, Lexical Conceptual Structures.

The implemented system builds the semantic representation of interrogatives sentences by starting from the verbal predicate as the central axis of this representation. The verbal meaning is broken down into semantic constituents which are labeled with a metalinguistic language in order to describe the meaning of the sentence. Semantic constituents are associated to phrases and, if necessary, adjuncts can be added to the final representation.

The result is a formal representation of a high level of abstraction of the sentence meaning which, differing from other systems, allows us to account for syntactic and lexical variation. The use of this kind of representation languages allows us a more flexible and natural representation in contrast ot other more formal systems of semantic representation that are excesively dependent on the linguistic chain being analyzed. In this system, equivalent sentences, from a semantic point of view, are expressed by means of the same semantic representation.

For the construction of this prototype, we have been working with a corpus of interrogative sentences extracted from a restricted domain. This corpus has been mainly elaborated from the questions written by students taken the course “Computational Linguistics” offered at the UDL.