xsyn logo  
  Home | People | Research | Publications | Calendar | Contact

Computational Semantics for Natural Language using Synchronous Formal Grammars

Funding: NSERC Discovery Grant
Principal Investigator: Chung-hye Han
Duration: 2008-2013

The long-term goal of this research program is to develop a wide-coverage synchronous formal grammar for syntax-semantics mapping in natural language, using Synchronous Tree Adjoining Grammars (STAG). An STAG is created by a pairing of Tree Adjoining Grammars (TAG). Under the STAG approach to compositional semantics, each syntactic elementary tree is paired with one or more semantic trees that represent its meaning. In our approach, the semantic trees have unreduced lambda calculus terms. A synchronous derivation proceeds by mapping a derivation from the source syntax side to an isomorphic derivation in the target semantics side. This isomorphism requirement not only guarantees that the derivation in syntax determines the meaning components needed for semantic composition and the way in which these meaning components are combined, but also ensures computational tractability. Using the STAG formalism, we have been developing analyses for syntax-semantics mapping of relative clauses, it-clefts, and coordination, and binding of reflexives. Through this research, we are developing a theory of syntax-semantics mapping in computational linguistics. On a practical side, the resulting wide-coverage STAG will be used to develop a semantic parser, which can then facilitate many Natural Language Processing applications that make use of computational semantics such as natural language interfaces to database queries, dialog systems, question/answering and summarization.

The nature of grammar competition in the clause structure of head final languages

Funding: SSHRC Standard Research Grant
Principal Investigator: Chung-hye Han
Collaborator: Jeffrey Lidz (University of Maryland)
Duration: 2007-2010

The main goal of this project is to better understand the split in the population for the acquisition of verb placement in the single speech community of a head final language like Korean and Japanese, and further refine the two-grammar hypothesis that the observed split is the consequence of the paucity of evidence from the input data regarding the placement of the verb in the clause structure of head final languages. The project employs experimental methodology to obtain data to address the following specific research questions: (i) How stable is the split in the speech community? (ii) Is the split in the speech community predictable or random? (iii) Are there any other syntactic/semantic phenomena that correlate with the split? Our findings could potentially lead to an identification of a new pattern of language acquisition available to the speech community, which in turn will have interesting implications for the theory of linguistic competence.

Syntax of head final languages

Funding: SSHRC Standard Research Grant
Principal Investigator: Chung-hye Han
Duration: 2003-2006

The main goal of the research program is to advance our understanding of the syntax of head-final languages in comparison to head-initial languages. The fundamental question driving the research will be how children acquire the placement of verbs in head-final languages. In head-initial languages, such as French and English, direct evidence is abundant from string word order through adverb or negation placement. But in a head-final language, verb-raising is hard to detect since there is no evidence from the word order to support a raising analysis. The project investigates various issues arising from this paucity of evidence, with much of the data obtained from both adults and children through carefully designed experiments: (i) the possibility of a split in the population for the acquisition of verb raising; (ii) the implication for the clause structure in general and syntax of negative imperatives in particular; (iii) the implication for the syntax of light verb constructions; (iv) cross-linguistic variation among head-final languages, in particular between Korean and Japanese. By addressing these issues, we aim to shed new light on cross-linguistic variation between head-initial and head-final languages that is directly related to the architecture of clause structure.