DocumentCode :
1862791
Title :
Preliminary Study of a New Approach to NMF Based Text Summarization Fused with Anaphora Resolution
Author :
Batcha, Nowshath Kadhar ; Zaki, Ahmed M.
Author_Institution :
Dept. of Comput. Sci., Int. Islamic Univ., Kuala Lumpur, Malaysia
fYear :
2010
fDate :
9-10 Jan. 2010
Firstpage :
367
Lastpage :
370
Abstract :
Recently Non-negative Matrix Factorization (NMF) has captured a quantifiable attention in the field of information retrieval. Especially in the dimension of text summarization, it has leaped a large step in minimizing the gap between the summary produced by the man and machine. This paper is aimed at a preliminary study of a novel approach to automatic text summarization using extraction method by fusing the output of an anaphoric resolver with NMF. This initial study is aimed at studying the impact of changes in the weights of H matrix of NMF matrices by treating anaphoric pronoun replacements in the term by sentence matrix of the original document. It is done to emphasize how anaphoric resolutions can improve the performance of the output when the original term by document matrix is properly treated with anaphoric replacements using NMF method. Results of the preliminary study demonstrate that when the output of anaphoric resolver is fused with NMF better results can be achieved in automatic text summarization.
Keywords :
image fusion; image resolution; matrix decomposition; text analysis; NMF; anaphora resolution; anaphoric pronoun replacements; automatic text summarization; document matrix; information retrieval; nonnegative matrix factorization; Automated highways; Computer science; Data mining; Information retrieval; Linear algebra; Matrix decomposition; Singular value decomposition; Sparse matrices; Vectors; Web sites; anaphora resolution; non-negative matrix factorization; text summarization;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Knowledge Discovery and Data Mining, 2010. WKDD '10. Third International Conference on
Conference_Location :
Phuket
Print_ISBN :
978-1-4244-5397-9
Electronic_ISBN :
978-1-4244-5398-6
Type :
conf
DOI :
10.1109/WKDD.2010.100
Filename :
5432584
Link To Document :
بازگشت