Krzysztof Simiński
Abstract:
The sentence boundary verification is the crucial task in the natural language processing. The text analysis is very often based on sentences. The origin of the problem originates from the haplography of the full stop. Many approaches have been proposed (as direct, character, feature and phrase analyses). This paper presents the attempt at applying the neuro-fuzzy approach to the feature analysis of the lexems in the neighbourhood of the full stop in question. The results of the experiments reveal that the neuro-fuzzy systems achieve similar or better results with two or three rules whereas the crisp system generated rule bases with more than 20 rules.
Reference:
Krzysztof Simiński, Neuro-fuzzy Approach to Sentence Boundary Verification in Polish Text, [in] OWD2008, X International PhD Workshop, 2008, pp. 267-270.
Bibtex Entry:
@INPROCEEDINGS{id:Siminski2008Neuroapproach,
author = {Krzysztof Simi\'{n}ski},
title = {Neuro-fuzzy Approach to Sentence Boundary Verification in Polish Text},
booktitle = {OWD2008, X International PhD Workshop},
year = {2008},
pages = {267--270},
address = {Wis{\l}a, Poland},
abstract = {The sentence boundary verification is the crucial task in the natural
language processing. The text analysis is very often based on sentences.
The origin of the problem originates from the haplography of the
full stop. Many approaches have been proposed (as direct, character,
feature and phrase analyses). This paper presents the attempt at
applying the neuro-fuzzy approach to the feature analysis of the
lexems in the neighbourhood of the full stop in question. The results
of the experiments reveal that the neuro-fuzzy systems achieve similar
or better results with two or three rules whereas the crisp system
generated rule bases with more than 20 rules.},
timestamp = {2008.10.13},
url = {http://sun.aei.polsl.pl/~ksim/data/publikacje/2008. Siminski. Neuro-fuzzy Approach to Sentence Boundary Verification in Polish Text.pdf}
}