Optimization Method Of Adaptive Filtering Utility Target

TREC(Text Retrieval Conference) is the most important conference in IR field. Adaptive Filtering track is one of the most important tasks in TREC. Evaluation measure optimization with threshold is one of the most important research areas in Adaptive Filtering. This paper takes TREC evaluation measure as our target function, and then compares maximum likelihood estimation in the threshold tuning (ML) with local target optimization, then proposes Local optimization Method of Adaptive Filtering Utility Target Based on Maximum Likelihood Estimation (MMLOR). ML is the best method for threshold in adaptive filtering in past TREC. ML is an unbias method which jointly estimates the parameters of the density distributions for relevant and non-relevant documents and the ratio of the relevant document in the corpus.MMLOR takes the advantage of ML and combines a local optimization method. MMLOR represents not only the character of holistic corpus but also the character of local corpus. MMLOR can tune the threshold according to the character of local corpus. MMLOR is based on ML and tunes the threshold according to the need of local corpus. MMLOR is not put two methods together a complex iterative process simply. MMLOR overcomes the shortcomings of single method. It proves that this method is effective in improving the quality of filtering.In order to prove that MMLOR is better than ML and local target optimization, this paper put three methods into the same filtering condition. In addition, this paper ameliorates ICTCLAS. ICTCLAS is a system of segmentation. And this paper modifies the rocchio means in order to make it fit the corpus better.