| The intelligent agent is an important application of distributive artificial intelligence (AI). It aims at lessening the burden of operating complex software or fussy job for people in order to enhance the amity of software and improve the efficiency as far as possible. General speaking, agent is a computing entity that can run in a certain environment and responses to its changes so that it can take action actively and flexibly to satisfy the purpose of design.With the high development of information society, email has replaced the traditional post and become one of rapid communications among people. Email bring more convenience to people, it has also made more chance for those commercial organizations. Those companies can send some advertisement email to any address at any time. These dust emails take up the memory of mailbox and can hinder people's life. At the same time, it is difficult to classify and settle more and more emails. The email classification based agent system is just designed to classify emails intelligently so as to liberate users from these troubles.Though there is completed software for email intelligent processing on abroad (e.g. Lydia Email Agent), there is no similar software in China. There are many email service agents, which can filter emails. But they only can process email by its address. Therefore, these email systems say nothing of real intelligence.The Email Classification Agent System discussed in this paper aims at classifying and settling emails. The first job of this paper is to compare the functions of current email systems. And then the filter and classification function in present email systems has been analyzed. Finally, the conclusion, which most email systems are not really intelligent, can be achieved. A new algorithm on email classification is given in this paper based on above analysis. That is to divide emails into four different degrees according to the idea of different users. The four degrees are primary email, secondary email, strange email and deleting email. Learning can do the classification of emails from the address, the subject, the content of received emails and the habit of user's operation. In addition, a corresponding system has been completed to validate the feasibility of the algorithm discussed in the paper. The new algorithm in the paper can classify most emails correctly and is feasible.The method is fairly important whether a problem can be solved successfully or not. The search strategy based on heuristic depth priority is adapted in the paper. The process of email by algorithm is completed, according to planning flow chart, from up to down and also from left to right. Once the terminal state is achieved (primary email, secondary email, deleting email and strange email), and then the algorithm ends. There is a key phrase in the algorithm that is discussed in chapter four in detail. Splitting words need to be done during classifying email according to its subject and content. Splitting words is a key to success of email classification. Since every sentence is made up of various words, it is necessary to understand the meaning of every word in the sentence in order to catch on the meaning of the sentence more perfectly. There is separator among the word of document in English, but the situation is completely different in Chinese documents. All kinds of words are continuously Chinese string. There are many ways of splitting word, such as way of mechanical matching, way of characteristic library, way of restrict matrix, way of grammar analyzing, way of understanding syncopation etc. This paper explained the way of splitting word in detail in chapter three. The way of mechanical matching is used in the paper for the characteristic of email.The search way based on enlightening and depth priority is adopted in the new algorithm. The algorithm will end as soon as the ultimate state appears (it refers to primary email, secondary email, deleting email and strange email). The step of analyzing words is the key to a successful class...
|
PDF Full Text Request