A Probe into Ambiguities of Determinative-Measure Compounds

759

Authors

nan

Abstract

This paper aims to further probe into the problems of ambiguities in automatic identification of determinative-measure compounds (DMs) in Chinese. It is known that Chinese DMs are identifiable by regular expression rules. However, rule matching only partially solve structural and lexical ambiguities. In this paper, a deep analyses based on corpus data was studied. With the subtle analyses of error identification and disambiguation of DM compounds, we classified three types of ambiguities, i.e. structural, sense, and functional ambiguities. We also proposed resolution principles to eliminate the problems and thus to improve word segmentation and POS (Part-Of-Speech) tagging. 1

Keywords

nan

Full Content

A Probe into Ambiguities of Determinative-Measure Compounds Shih-Min Li, Su-Chu Lin, Keh-Jiann Chen CKIP, Institute of Information Science, Academia Sinica, Taipei {shihmin, jess}@hp.iis.sinica.edu.tw; kchen@iis.sinica.edu.tw Abstract This paper aims to further probe into the problems of ambiguities in automatic identification of determinative-measure compounds (DMs) in Chinese. It is known that Chinese DMs are identifiable by regular expression rules. However, rule matching only partially solve structural and lexical ambiguities. In this paper, a deep analyses based on corpus data was studied. With the subtle analyses of error identification and disambiguation of DM compounds, we classified three types of ambiguities, i.e. structural, sense, and functional ambiguities. We also proposed resolution principles to eliminate the problems and thus to improve word segmentation and POS (Part-Of-Speech) tagging. 1 Introduction To a speaker of English, one of the most striking features of the Mandarin noun phrase is the classifier. A classifier is a word that must occur with a number and/or a demonstrative, or certain quantifiers before the noun (Li and Thompson 1981: 104). Furthermore, Li and Thompson (1981) assert any measure word can be a classifier, so the combination of demonstrative and/or number or quantifier plus a classifier or a measure is defined as a classifier phrase or a measure phrase. For example, san ge in san ge ren (三個人), zhe zhan in zhe zhan deng (這盞燈), ji jian in ji jian yifu (幾 件衣服), liu li in liu li lu (六里路), na jin in na jin yangrou (那斤羊肉) and ji gang in ji gang cu (幾缸 醋) are classifier phrases/measure phrases, which are called as D-M compounds in Chao 1968. A determinative (D) and a measure normally make a compound with unlimited versatility and form a transient word of no lexical import (Chao 1968: 389). Although the demonstratives, numerals and measures may be listed exhaustedly, their combination is inexhaustible. Certain constructions of DMs are ambiguous, for example: (1) 廣告裡全篇多具聳人聽聞的口號 guanggao li quan pian duo ju hairentingwen de kouhao A whole advertisement mostly has shocking slogans. (2) 取此名 qu ci ming choose this one (person) / name this name (3) 二十五年的審核、排隊、等待 ershiwu nian de shenhe paidui dengdai examining, lining up and waiting in the year of twenty five / for twenty five years The morpheme ju in Academia Sinica Balanced Corpus (Sinica Corpus) has two parts of speech, VJ and Nf.1 Thus the phrase duo ju in sentence (1) can be either a verb phrase with the meaning of ‘mostly have’ or a DM. When ju functions as a measure, it always modifies corpses, not slogans. Since ju never co-occurs with slogans, the phrase duo ju here is certainly a verb phrase and then the lexical ambiguity of ju is reduced. In example (2), ming can function as a measure as well as a noun so that this verb phrase has two meanings. In example (3), ershiwu nian can be a time point specifying the event-time of the verb, or denotes the period of time delimitating the time length of the event. The former temporal …

Content truncated to 500 words

Labeled Text

This paper aims to further probe into the problems of ambiguities in automatic identification of determinative-measure compounds (DMs) in Chinese. It is known that Chinese DMs are identifiable by regular expression rules. However, rule matching only partially solve structural and lexical ambiguities. In this paper, a deep analyses based on corpus data was studied. With the subtle analyses of error identification and disambiguation of DM compounds, we classified three types of ambiguities, i.e. structural, sense, and functional ambiguities. We also proposed resolution principles to eliminate the problems and thus to improve word segmentation and POS (Part-Of-Speech) tagging. 1

Publication Details

Paper ID:
759
Year:
2005
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan