Please use this identifier to cite or link to this item:

Software agents for issues identification : an experiment on financial documents

Authors Yen, Jerome
Ma, Pai Chun
Sivakumar, V.
Chen, H.
Issue Date 1996
Source Proceedings of the first Asia-Pacific Decision Sciences Institute Conference, Hong Kong, Hong Kong University of Science and Technology, Hong Kong, 21-22 June 1996, p. 805-814
Summary This paper presents a knowledge-based approach to building an intelligent agent to help overcome the information overflow problems when dealing with large textual databases. The proposed agent is able to identify key issues from unstructured textual documents and reorganize the contents according to the issues identified. After such process, the processed documents contain tree-like structure. Transfer unstructured textual documents into semi-structured documents is extremely important in achieving higher efficiency in information retrieval and retention. The proposed approach is based on a number of techniques: automatic indexing, concept space generation, and neural network classification. In this paper, we describe how these techniques are used to extract subject descriptors, their semantic relationships, and the texts (documents or paragraphs) related to each descriptor, from the textual databases. The proposed agent has been implemented and tested with the annual reports from thirteen of the largest international banks. We compared the system's performance in terms of the users' satisfaction and speed in identifying the key issues and supporting texts. The results indicate that with the support from the agent, users are easier and faster to identify the key issues and supporting texts. The conclusion we have is that software agents are able to assist human decision makers to more efficiently handle large textual databases. By identifying the structure or semantics and allow the users to read or browse the documents based on such structure or semantics can improve the efficiency and utilization of the available information.
Language English
Format Conference paper
Files in this item:
File Description Size Format
software.PDF 1269177 B Adobe PDF