Extending the Lexicon by Exploiting Subregularities

Robert Wilensky

EECS Department
University of California, Berkeley
Technical Report No. UCB/CSD-91-618
1991

http://www2.eecs.berkeley.edu/Pubs/TechRpts/1991/CSD-91-618.pdf

Many linguistic phenomena are not predictable from one another, but are nevertheless subject to regularities. One phenomenon sometimes ``motivates'' another, in that it provides a post hoc explanation for some usage, but falls short of predicting it. Since motivations do not seem to play an obvious role in language use, and do not eliminate the need for any information, they have not played a major role in most linguistic theorizing, nor have they been exploited by language processing systems. We believe that motivations may play an important role in language acquisition. In particular, motivations are frequently found among word senses, and can be utilized to aid in the large problem of learning the lexicon. Previously, we have studied a subclass of this problem, namely, acquiring metaphoric sense extensions. However, there appear to be a large number of other kinds of word sense extension that can also be so exploited. These relations constitute a potentially unlimited number of ``subregularities'' within and about the lexicon. We propose that acquisition is performed by a form of analogical reasoning that can be applied to hypothesizing new polysemous word senses. We speculate that this proposal has some cognitive validity, and suggest that it can usefully supplement other computational lexical acquisition methods based on corpora and on-line dictionaries.


BibTeX citation:

@techreport{Wilensky:CSD-91-618,
    Author = {Wilensky, Robert},
    Title = {Extending the Lexicon by Exploiting Subregularities},
    Institution = {EECS Department, University of California, Berkeley},
    Year = {1991},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/1991/6376.html},
    Number = {UCB/CSD-91-618},
    Abstract = {Many linguistic phenomena are not predictable from one another, but are nevertheless subject to regularities.  One phenomenon sometimes ``motivates'' another, in that it provides a post hoc explanation for some usage, but falls short of predicting it.  Since motivations do not seem to play an obvious role in language use, and do not eliminate the need for any information, they have not played a major role in most linguistic theorizing, nor have they been exploited by language processing systems. We believe that motivations may play an important role in language acquisition.  In particular, motivations are frequently found among word senses, and can be utilized to aid in the large problem of learning the lexicon.  Previously, we have studied a subclass of this problem, namely, acquiring metaphoric sense extensions.  However, there appear to be a large number of other kinds of word sense extension that can also be so exploited.  These relations constitute a potentially unlimited number of ``subregularities'' within and about the lexicon. We propose that acquisition is performed by a form of analogical reasoning that can be applied to hypothesizing new polysemous word senses.  We speculate that this proposal has some cognitive validity, and suggest that it can usefully supplement other computational lexical acquisition methods based on corpora and on-line dictionaries.}
}

EndNote citation:

%0 Report
%A Wilensky, Robert
%T Extending the Lexicon by Exploiting Subregularities
%I EECS Department, University of California, Berkeley
%D 1991
%@ UCB/CSD-91-618
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/1991/6376.html
%F Wilensky:CSD-91-618