Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Semantic bootstrapping: a theoretical perspective
Wu W., Li H., Wang H., Zhu K. IEEE Transactions on Knowledge and Data Engineering29 (2):446-457,2017.Type:Article
Date Reviewed: Jun 27 2017

Language semantics can be seen as the natural language expressions of knowledge during learning, that is, during the knowledge acquisition process. Because learning is a bootstrapping process, it is natural for language semantics evaluation to be seen as a bootstrapping process. The bootstrapping process during semantic determination is mainly controlled by the syntactic structure of the phrase rather than by the knowledge discovered within a bootstrap step. Due to the ambiguous nature of the syntactic structure of language phrases, the one-to-one mapping between syntactic patterns and the underlying knowledge cannot always be determined.

The main points when using bootstrapping in a computational process are answers to the questions: What is the first level of this bootstrapping process, and what are the operators that allow one to build on previous computation layers while determining the next bootstrap layers? The answer to these questions for semantic evaluation is affected by the meaning of the language concept. Natural language is a communication tool used by language speakers to communicate knowledge to each other; hence, natural language semantics is natural language. Therefore, communication between human language speakers does not need semantic evaluation. However, during language processing by machine, semantic evaluation is a necessary process. The semantics of a text processed by a machine is determined by a computation process from the syntax of the processed text that is artificially generated, usually using a grammar specification of text expressions.

The paper provides an algorithm that uses the bootstrapping process for the evaluation of the isA relationship over a language corpus. For that, the authors look at the natural language as a collection of patterns of the form “subject, predicate, object.” The first level of the bootstrapping process is the database containing the language corpus. The bootstrapping process is controlled by the syntactic structure of the pattern searched over the corpus or by the knowledge discovered in the corpus. The paper uses the Hearst patterns as syntactic structures of the text searched and disambiguates the result by leveraging the knowledge identified in the previous step to help extract more knowledge in the current step, until no new knowledge is discovered. Knowledge formalization is provided by considering the pairs (x,y) where y isA x, and the phrase s searched has the syntactic structure (X, ⟨P⟩, Y) where X = (x1,...,xm) is the set of all super-concepts, ⟨P⟩ is the pattern keywords, and Y = (y1,..., yn) is the set of all candidate subconcepts. Consequently, this is not really a semantic search. For a true semantic search within this scheme, one should consider the corpus as a collection of knowledge domains. Then, semantic search over the corpus can be performed searching for tuples (x,y,z) in the relationship isA ˆ inDom, that is, y isA x inDom z.

Reviewer:  T. Rus Review #: CR145382 (1709-0633)
Bookmark and Share
  Reviewer Selected
 
 
Natural Language Processing (I.2.7 )
 
 
Data Mining (H.2.8 ... )
 
 
Knowledge Acquisition (I.2.6 ... )
 
 
Semantics Of Programming Languages (F.3.2 )
 
Would you recommend this review?
yes
no
Other reviews under "Natural Language Processing": Date
Current research in natural language generation
Dale R. (ed), Mellish C. (ed), Zock M., Academic Press Prof., Inc., San Diego, CA, 1990. Type: Book (9780122007354)
Nov 1 1992
Incremental interpretation
Pereira F., Pollack M. Artificial Intelligence 50(1): 37-82, 1991. Type: Article
Aug 1 1992
Natural language and computational linguistics
Beardon C., Lumsden D., Holmes G., Ellis Horwood, Upper Saddle River, NJ, 1991. Type: Book (9780136128137)
Jul 1 1992
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy