An attentive self-organizing neural model for text mining

This paper utilizes an attention concept approach in text mining to address the deficiencies of existing keyword search engines. We show how an attention concept in conjunction with a traditional search approach can be used to develop an adaptive text mining model with user-oriented, time-based and attentive knowledge. Without changing a user’s search behavior, this paper considers some specific post-search operations as attentive targets for building the personalized interest base. This interest base is further shown on an interest map via the self-organizing map algorithm (SOM). By comparing the personalized interest map, the original search results from a keyword search engine are re-ranked. Experimental results demonstrate that the attentive search mechanism is able to improve user satisfaction.