An Intelligent Email Agent
- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
An Intelligent Email Agent
Adina Magda Florea Alexandru Moldovanu
adina@apolo.cs.pub.ro sandum@sky.dsp.pub.ro
"Politehnica" University of Bucharest
Abstract
We propose an email agent that assists the user in the management of
email messages. The agent is autonomous and runs in an UNIX
environment. It has a learning capability that permits the agent to learn
the user's interests and preferences when dealing with messages, and to
automatically perform the required actions. The learning algorithm is a
variant of Quinlan's ID3 but the acquired knowledge is stored in a rule
knowledge base. The agent can learn from several sources of
knowledge, including the possibility to consult other similar email
agents, presumably more knowledgeable than itself.
1. Interface Agents
The term agent is used frequently these days. This is true both for software applications in general and for artificial intelligence in particular. Most often, when people use the term agent they refer to an entity that functions continuously and autonomously in an environment in which other processes take place and other agents can exist [6]. Wooldridge and Jennings [7] defined two increasingly strong notions of agency: weak agency, primarily to describe the software agents community, and strong agency, to describe artificial intelligence agents, namely agents being capable of rationality, mentalistic notions, adaptability and learning. The agent proposed in this paper belongs to this second notion of agency, being user adaptable and capable of learning. It is also an interface agent, in the sense that it provides assistance to a user dealing with a particular application. Interface agents modify the currently dominant interaction metaphor of direct manipulation [5], which requires the user to initiate all tasks explicitly and to monitor all events. The newly metaphor used in this case is that of a personal assistant who is collaborating with the user in the same work environment to initiate communication, monitor events and perform tasks [2]. The assistant becomes gradually more effective as it learns the user's interests, habits and preferences.
2 Functional Structure of the Agent
The email agent develops and uses a set of decision rules to classify incoming email messages on behalf of the user. The goal of the agent is to relieve the user from the burden of dealing with a large number of email messages by performing the actions the user would take on these messages. Because different users may take different actions on the same message, it would be rather difficult to handcode the user's preferences. Instead, the agent has a learning component that enables it to accustom to its user. The agent will gather training examples and will apply a learning algorithm to