2002 Volume 16 Issue 4
Published: 15 August 2002
-
Select all|
-
2002, 16(4): 2-12.Lexical knowledge representation and knowledge extraction are bottlenecks of natural language processing. In this paper ,we proposed a preliminary framework for representing lexical knowledge and a learning model for extracting common sense knowledge. The key technologies to achieve this proposed learning model are (1) unknown word identification and its syntactic and semantic categorization , (2) lexical semantic disambiguation , (3) parsing technologies by using syntactic ,semantic and common sense knowledge.
-
2002, 16(4): 13-21,45.Chinese Concept Dictionary (CCD) is a WordNet-like semantic lexicon ,developed by the Institute of Computational Linguistics ,Peking University. This article focuses on the structure of CCD ,which presents a concept defined by a set of synonyms (SynSet) and a network of concepts based on the hypernymy hierarchy ,the basic relationship ,with other supplementary relationships. The deductive rules on this semantic network are mathematically formalized ,which could be well applied to Chinese semantic analysis.
-
2002, 16(4): 22-32.Lexical Semantic Relations in English and other European languages are well-studied. For instance , EuroWordNet (Vossen 1998) is a lexical-semantic database in which the meaning of each word is characterized by its semantic relations with other word meanings. That is to say , the understanding of the meaning of a word is by its semantic relations with others. In order to ensure the quality and consistence of the database during the construction of it ,the EuroWordNet project provided linguistic tests for each semantic relation in each language it handled. Empirically ,using these linguistic semantic tests , people can more easily and consistently check if two word meanings have a certain kind of semantic relations. And anyone who accesses the database can verify the correctness as to each relation linking. In other words ,these linguistic tests for lexical semantic relations provide a building block for a verifiable and language-independent theory of lexical semantics. In this paper ,we explore the possibility of a similar set of semantic relation tests for Chinese. In particular ,we attempted to provide test formulae and rules for significant semantic relations in order to evaluate the possibility. In addition to the construction of the theoretical infrastructure of studies on Chinese lexical semantics , our study will also provide strong support for the universality of the relation-based approach to lexical representation and linguistic ontology pioneered by Miller’s WordNet (Fellbaum 1998) .
-
2002, 16(4): 33-39.Machine-readable dictionaries have been regarded as a rich knowledge source from which various relations in lexical semantics can be effectively extracted. These semantic relations have been found useful for supporting a wide range of natural language processing tasks ,from information retrieval to interpretation of noun sequences ,and to machine translation. In this paper ,we address issues related to problems in building a linkage between two existing lexical resources ,WordNet and LDOCE-E/C (English-Chinese version of Longman dictionary of Contemporary English) by lexical overlap and genus agreement . In the LDOCE-E/C ,each headword definition is given a Chinese translation. When we successfully linked a WordNet synset to a definition of bilingual dictionary ,the Chinese definition of bilingual dictionary will be connected to that synset .
-
2002, 16(4): 40-45.Context environment can be used to describe the meaning similarity between words. Corpus based similarity word extraction can be used in various kind of fields such as lexicon compiling and intelligent search engine. Based on 6.4 million chars of Chinese ancient poetry ,a statistic model was defined to extract contextual similarity words from the corpus. A concept based intelligent search engine for Chinese ancient poetry was developed on top of the word similarity relations. The result is encouraging.
-
2002, 16(4): 46-52.This paper proposes a Chinese text analysis method on specific domain. According to the texts character , this method combines shallow parsing technology with deep parsing and semantic analysis technology. Drawing on finite-state cascades method ,its shallow parsing module recognizes named entities in the texts. So that it greatly eases the burden of the deep analysis module. Principally depending on word collocation information ,its deep analysis module combines syntactic analysis and semantic analysis to determine sentence structure. It gains good effect at resolving the ambiguity of phrase structure in specific domain.
-
2002, 16(4): 53-60.A hierarchical structure model of the law literature based on structure identifier is presented in this paper. The model reflects the hierarchical characteristics and the continuity and parallelity of the chapters ,articles ,and sub-articles. A system for indexing and aligning Chinese-English law literature at sub-article level based on the model is implemented which supports fault-correction and fault-tolerance. The test result shows that the average indexing time is 3.31ms ,and the alignment precision is 98.6%. Combining this method and the lexical criteria results in an alignment precision of 99.3%.
-
2002, 16(4): 61-66.In this paper ,the similitude problems of Tibetan Characters are discussed. The same conclusion are showed by recognition system. On the base of Tibetn Character characteristic of graph construction ,we explain what the peculiar processing in normalization ,features selection and so on.
