Method and system for text classification of short messages

Abstract

The invention discloses a short message text classifying method and system. The short message text classifying method comprises the following steps: dividing a short message classifying process into different tasks in advance, and allocating different tasks onto different ports of a working node of a stream-oriented computation system; extracting the keywords of a short message text to be classified, determining a class library needing to be updated, updating the class library needing to be updated, and computing the characteristic vector of the short message text to be classified; acquiring the similarity between the short message text to be classified and the characteristic vectors of different class library members according to the computed characteristic vector, and determining the class of the short message text to be classified according to the acquired similarity. According to a short message text classifying scheme, the updating of the class library and the computation of the characteristic vector are performed in parallel through different task ports of the working node of the stream-oriented computation system respectively after characteristic preprocessing of short message text, thereby greatly increasing the response speed of short message text processing, and increasing the speed and accuracy of junk short message filtering.

Description

technical field [0001] The invention relates to short message text processing technology, in particular to a short message text classification method and system. Background technique [0002] In the era of Internet big data, real-time processing and analysis of user behavior is more important. Taking SMS text processing as an example, the flood of spam messages, such as fraudulent SMS, advertising promotion, and reactionary SMS, has brought great harm to users. Therefore, operators need to filter spam messages by identifying the content of SMS messages. The timeliness of the short message determines that the processing and distribution of the short message must be completed within a short period of time, so the short message processing system is undoubtedly required to have high real-time performance. [0003] At present, there are mainly two text classification methods for short messages, one is a classification method based on "keywords" + matching rules, and the other is...

AI Technical Summary

Problems solved by technology

[0008] These two characteristics determine the recognition and classification method based on the content of text messages, which cannot meet the classification requirements of short messages with high real-time requirements.

Images