基于语境关联的Web信息过滤算法

设计一个文本过滤实验 ,首先从语料库的词频统计结果中挖掘出词频的二元关联度 ,然后用一个Hop field网络将词频的二元关联关系转化为语境关联关系 ,训练语言单位在整个上下文环境下的权重 ,并建立用户模板 .该算法改善了词频特征提取算法与文本上下文环境的匹配状况 ,实验结果表明 ,对专业性Web文档的过滤可达到更高的精确度

Web information filtering arithmetic based on context relevancy

Web Information Filtering, which is to filter the most expected documents from the target Web Set. As known from our reading experience, the context information contained in user-interested documents is an indispensability fact in estimating text subjects. To design a text-filter experiment, first, the dualistic relevancy of word coexistence frequency should be mined by the word frequency Stat. from corpus, then a Hopfield NN can be used to convert the dualistic relevancy to context relevancy and to train the weights of language units in the whole context, at last the user-template is built. This arithmetic can improve the matching condition between word frequency characters and the context information. The result of experiment shows that more precision can be achieved for professional Web text filtering.