A Study of Using Text Mining Techniques to Construct the Legal Maps of Financial Markets

Following the trends of financial liberalization and internationalization, the financial industry in Taiwan has gradually moved toward the deregulation environment from the highly regulated environment. The total number of financial commodities in financial markets thus increases a lot. It makes the heavy competition in the financial industry and therefore accompanies some problems, including the crisis caused from poor qualities of credit cards and cash cards, non-performing loans, sales persons』 mal-marketing new financial commodities, etc. Hence, the legal regulations of the government have to attain the dual goals of deregulation and prevention of the preceding financial problems so that the legal system of Taiwan』s financial markets has become very complicated. Besides, after the promulgation of financial holding company law in 2001, the businesses of financial company cover a lot of sub-industries, such as banking, insurance, securities, futures, securities investment trust and consulting, etc. It increases the burden of personnel of financial companies and consumers of financial markets in legal information processing regarding their privileges and obligations. Therefore, a good legal knowledge representation system, capable of effectively providing the personnel of financial companies and consumers of financial markets with comprehensive legal knowledge, may be one of the resolutions for resolving the above problem. It can assist them in preventing erratic behavior in advance and making their ways in financial business in compliance with laws and regulations. This research will apply the novel text mining techniques on constructing a legal knowledge map system of Taiwan』s financial market. We hope that the construction of the system not only can represent the legal knowledge maps of financial markets effectively but also can be automated to efficiently handle the fast-growing and changeable legal information. The text mining techniques applied in this research cover automatic Chinese term extraction techniques, document vector models, ontology-based methods, growing hierarchical self-organizing maps, support vector machines, backpropogation neural network, case-based reasoning, etc. Finally, we will evaluate the effectiveness of different methods by recruiting system users to test the system and then report the characteristics and pros and cons of various map representation methods. We hope the legal maps can help users study and retrieve legal knowledge regarding financial markets.

Project ID:PF9611-0123
External Project ID:NSC96-2416-H182-011