應用相關回饋之語詞資訊於概念建立之方法

The application of the term information residing in relevance feedback for concept construction

張明竣(Ming-Chun Chang)

指導教授 ： 周世傑

國立中央大學/管理學院/資訊管理學系/碩士(2015年)

摘要

過去的概念萃取研究中，對於一篇文件應該萃取多少概念去表達，沒有一個依據。因此，本研究旨在探討概念萃取數量與文件集離散程度間的關聯，並且利用公開資料集TREC-6做實驗，驗證其是否對文件分類效能上有所提升。如果將文件集分群後，文件平均分佈在各個群中，代表文件集的離散程度很高，本研究假設應該萃取更多概念才能夠足以表達文件集中大部份文章。反之如果文件集中在某幾個群中，代表文件集的離散程度很低，表示文件的分佈是很集中的，萃取少量概念就足以表達大部份文章。在本研究中提出動態概念萃取策略，利用文件分群得知其離散程度，並利用此數據動態控制萃取的概念數量，經實驗驗證可以初步證實本研究所提出的動態概念萃取策略，對於文件分類上的效能有進一步提升。

關鍵字

概念萃取 ； 文件概念化 ； 相關回饋 ； 向量空間模型 ； 文件集離散程度

並列摘要

In the past, we did not have a method to determine how many concpts to represent the document. The aim of this study is to discuss the relation between the number of concepts extraction and the dispersion of document dataset. This study uses public document dataset TREC-6 to validate the effectiveness of text classification. This study proposes that a document dataset has high dispersion if the documents distribute evenly in a cluster. In this case, this study assumes that more concepts are needed to represent the document. On the contrary, if the documents has a centralized distribution in a cluster, the document dataset has low dispersion. In this case, this study assumes that less concepts are needed to represent the document. This study proposes a dynamic concept extraction method which applies the degree of dispersion as the basis to dynamically determine the number of concepts. Empirical results show that the proposes method can improve the effectiveness in text classification.

並列關鍵字

concept extraction ； bag-of-concepts ； relevance feedback ； vector space model ； dispersion of document dataset

參考文獻

[1] G. Salton, A. Wong and C. Yang, 'A vector space model for automatic indexing', Commun. ACM, vol. 18, no. 11, pp. 613-620, 1975.

[2] G. Salton and C. Buckley, 'Term-weighting approaches in automatic text retrieval', Information Processing & Management, vol. 24, no. 5, pp. 513-523, 1988.

[3] J. Wu, 'Applying relevance feedback to construct a vector space model with concepts as the dimension value', National Central University, Taiwan, 2013.

[4] M. Sahlgren and R. Cöster, 'Using Bag-of-Concepts to Improve the Performance of Support Vector Machines in Text Categorization', in Proceedings of the 20th international conference on Computational Linguistics, Geneva, Switzerland, August 23-27, 2004, p. 487.

[5] A. Alahmadi, A. Joorabchi and A. E. Mahdi, 'A New Text Representation Scheme Combining Bag-of-Words and Bag-of-Concepts Approaches for Automatic Text Classification', in GCC Conference and Exhibition (GCC), 2013 7th IEEE, Doha, November 17-20, 2013, pp. 108-113.