<span style="font-size: 10pt; font-family: &quot;Times New Roman&quot;; mso-fareast-font-family: 宋体; mso-ansi-language: EN-US; mso-fareast-language: EN-US; mso-bidi-language: AR-SA; mso-bidi-font-size: 9.0pt;" lang="EN-US">We proposed a novel index-based online text classification method, investigated two index models, and compared the performances of various index granularities for English and Chinese SMS message. Based on the proposed method, six individual classifiers were implemented according to various text features of Chinese message, which were further combined to form an ensemble classifier. The experimental results from English corpus show that the relevant feature among words can increase the classification confidence and the trigram co-occurrence feature of words is an appropriate relevant feature. The experimental results from real Chinese corpus show that the performance of classifier applying word-level index model is better than the one applying document-level index model. The trigram segment outperforms the exact segment in indexing, so it is not necessary to segment Chinese text exactly when indexing by our proposed method. Applying parallel multi-thread ensemble learning, our proposed method has constant time complexity, which is critical to large scale data and online filtering.</span>

No abstract

Through the investigation of email document structure, this paper proposes a multi-field learning (MFL) framework, which breaks the multi-field document Text Classification (TC) problem into several sub-document TC problems, and makes the final category prediction by weighted linear combination of several subdocument TC results. Many previous statistical TC algorithms can be easily rebuilt within the MFL framework via turning binary result to spamminess score, which is a real number and reflects the likelihood that the classified email is spam. The experimental results in the TREC spam track show that the performances of many TC algorithms can be improved within the MFL framework.

No abstract

No abstract