Email Processing System Based On Personal Information Management

World Wide Web is becoming one of the main media for people and it is alsobecoming the huge data warehouse and the source of latent knowledge for humanbeing. With knowledge blast and the need for huge data processing, people are tendedto use computer for transaction, so as the information management computer becamemore and more important. Meanwhile email was the most popular informationmedium for people with high speed and performance. So how to make it easy forpeople to process and utilize the online data turns to be an inevitable challenge forpeople.This paper works on email process techniques from two aspects: email clusteringand summarization based on personal information management and some relevantalgorithm was brought forward.Our email clustering was clustering for search result. So the clustering result waspost clustering and depended on query that is similar to the vivisimo of searchengineer. Traditional clustering techniques are inadequate since they don't generateclusters with highly readable names. In this paper, we reformat the clustering problemas a salient phrase ranking problem so the unsupervised clustering problem wasconverted to a supervised learning problem. Given a query and the ranked list ofemails returned by search part, our method first extracts and ranks salient phrases ascandidate cluster names, based on a regression model learned from human labeledtraining data. The documents are assigned to relevant salient phrases to formcandidate clusters, and the final clusters are generated by merging these candidateclusters.To email summarization, it is summarization about clustering result and dependson query. Traditional text summarizing methods were lack of support for clusteringand structured text. So we have improved some algorithms on similarity in the sameclustering and different structure's affection on summarization.Finally, a demo was introduced to show the algorithms mentioned in this paper.