Building a Parser That can Afford to Interact with Semantics
نویسنده
چکیده
Natural language understanding programs get bogged down by the multiplicity of possible syntactic structures while processing real world texts that human understanders do not have much difficulty with. In this work, I analyze the relationships between parsing strategies, the degree of local ambiguity encountered by them, and semantic feedback to syntax, and propose a parsing algorithm called Head-Signaled Left Corner Parsing (HSLC) that minimizes local ambiguities while supporting interactive syntactic and semantic analysis. Such a parser has been implemented in a sentence understanding program called COMPERE (Eiselt, Ma-hesh, & Holbrook 1993). A parser could quickly eliminate many possible syntactic structures for a sentence by using (a) the grammar to generate syntactic expectations, (b) structural preferences such as Minimal Attachment or Right Association, (c) feedback from semantic analysis, (d) statistical preferences based on a corpus , or (e) case-based preferences arising from prior texts about stereotypical situations. None of the above strategies suffices by itself for handling real text. In this work, I assume that (a) we must strive to design parsing strategies capable of analyzing general , real life text, (b) it is beneficial to produce immediate , incremental interpretations ('meanings') of incoming texts, and (c) semantic (and pragmatic) analysis can provide useful feedback to syntax without requiring unbounded resources. Given these, my objective is to design a parsing strategy that makes the best use of linguistic preferences– both grammatical and structural, and also semantic and conceptual preferences, while minimizing local ambiguities. Strong cognitive motivations for devising such a solution were presented earlier in (Eiselt, Mahesh, & Holbrook 1993). The question this leads to is: When should the parser interact with the semantic analyzer? It should interact only when such interaction is beneficial to one or both, that is, when one can provide some information to the other to help reduce the number of choices being considered. Parsing strategies can be distinguished along a dimension of " eagerness " depending on when they make commitments to a syntactic unit and are ready for interaction with semantics. At one end of the spectrum lies pure bottom-up parsing, being too circumspect and precluding the use of syntactic expectations. Pure top-down parsing, at the other end, is too eager and leads to unwarranted back-tracking. Such nondeterminism is a problem for incremental interaction with semantics. A combination strategy called Left Corner (LC) Parsing has been shown to be a good middle ground for using …
منابع مشابه
Studying impressive parameters on the performance of Persian probabilistic context free grammar parser
In linguistics, a tree bank is a parsed text corpus that annotates syntactic or semantic sentence structure. The exploitation of tree bank data has been important ever since the first large-scale tree bank, The Penn Treebank, was published. However, although originating in computational linguistics, the value of tree bank is becoming more widely appreciated in linguistics research as a whole. F...
متن کاملBuilding a computational lexicon and ontology with FrameNet
This paper explores FrameNet as a resource for building a lexicon for deep syntactic and semantic parsing with a practical multipledomain parser. The TRIPS parser is a wide-coverage parser which uses a domain-independent ontology to produce semantic interpretations in 5 different application domains. We show how semantic information from FrameNet can be useful for developing a domainindependent...
متن کاملEvaluation of E-Trust Building Structures Interact With Transportation
Transportation industry is the most dynamic components of any society. In the twenty-first century, with the growth of technology and the widespread use of the Internet and the emergence of e-commerce and e-business interaction and active transportation industry deserves to have a wide range of electronic services to the transportation community to take advantage of the investors of the new and...
متن کاملBuilding lexical resources for PrincPar, a large coverage parser that generates principled semantic representations
Parsing, one of the more successful areas of Natural Language Processing, has mostly been concerned with syntactic structure. Though uncovering the syntactic structure of sentences is very important, in many applications a meaning representation for the input must be derived as well. We report on PrincPar, a parser that builds full meaning representations. It integrates LCFLEX, a robust parser,...
متن کاملFeature Engineering in Persian Dependency Parser
Dependency parser is one of the most important fundamental tools in the natural language processing, which extracts structure of sentences and determines the relations between words based on the dependency grammar. The dependency parser is proper for free order languages, such as Persian. In this paper, data-driven dependency parser has been developed with the help of phrase-structure parser fo...
متن کامل