استفاده از نقشه خود سازمان ده سلسله مراتبی رشد برای ساخت نقشه های قانونی برای بازارهای بورس اوراق بهادار و معاملات آتی تایوان

A good legal knowledge representation system, capable of effectively providing investors with comprehensive legal knowledge, is needed for investors to prevent erratic behavior before investment decisions. This is especially important in Taiwan’s securities and futures markets because the majority of market participants are individual investors who have limited access to legal knowledge about markets. Besides, the construction of the knowledge representation has to be automatic in order to efficiently handle the fast-growing and changeable legal information. Thus, we use the GHSOM algorithm to present a content-based and easy-to-use map hierarchy for Chinese legal documents in the securities and futures markets in the Chinese language. Meanwhile, an enhanced topic selection module and a web-based user interface are also proposed. The maps can be browsed on the web site (http://synteny.iis.sinica.edu.tw/legalmap/). To evaluate the legal maps, we apply two approaches, namely a validity test and a task experiment.

In legal informatics, successful searching of huge legal digital libraries depends a great deal on the user’s ability to master the legal terminology (Schweighofer, Rauber, & Dittenbach, 2001). Due to cost considerations, traditional approaches to legal knowledge representation based on thesauri and classification by professional persons have gradually developed into semi-automatic or automatic approaches. Thus, how to automatically categorize, index, organize, present, and summarize the enormous amounts of legal documents to enable quick and efficient retrieval of accurate legal information is a key issue (Merkl and Schweighofer, 1997, Schweighofer et al., 2001, Schweighofer et al., 1995 and Thompson, 2001). Most previous research in legal informatics has focused on Western languages – seldom on oriental languages, such as Chinese, Japanese, or Korean. In this paper, we study the effects of applying legal informatics to legal documents written in one of the above oriental languages by using an unsupervised learning algorithm called the growing hierarchical self-organizing map (GHSOM). As this algorithm has a proven performance record in Western language legal knowledge systems, we apply it to the construction of Chinese legal maps. According to Schweighofer et al. (2001), it is better to segment the topics in legal documents when presenting legal knowledge. We, therefore, selected a very important topic, namely, legal knowledge of Taiwan’s securities and futures markets, as our test environment, because – according to statistics published by the ROC government1 – the majority of participants in these markets are individual investors. In contrast, institutional investors are the key players in the American and European markets. To protect individual investors, who are not as sophisticated as their institutional counterparts in gathering market information, Taiwan’s securities and futures markets are highly regulated by several government agencies and the markets’ self-regulating bodies, including the Securities and Futures Bureau (SFB), the Taiwan Securities Exchange (TSE), the ROC. Over-the-counter Securities Exchange (OTC), and Taiwan Futures Exchange (Taifex). However, as most investors are not familiar with the large number of laws, rules, and regulations promulgated by the competent authorities, it is hard for them to understand the relationships among the different types of legal information. As a result, they sometimes violate laws, and inadvertently commit crimes. Currently, legal information retrieval systems used in Taiwan’s markets emphasize key word search functions, but the search results are not presented in a meaningful way. As most market participants are not experienced in retrieving legal information, the retrieval procedure should be uncomplicated and user friendly so that legal knowledge about Taiwan’s securities and futures markets can be accessed easily. Hence, the goal of this paper is to provide helpful and cost-effective legal guidance for market participants, including institutional and individual investors, employees of public companies, and securities and futures service providers. The remainder of this paper is organized as follows. In Section 2, we introduce related research about legal knowledge retrieval. In Section 3, we review the self-organizing map (SOM) related literature and point out some drawbacks of applying growing hierarchical SOM (GHSOM) to legal maps. We then describe how to construct legal maps for Taiwan’s securities and futures markets in Section 4, and present the legal maps in Section 5. In Section 6, we evaluate GHSOM by a validity test and a task experiment. Finally, in Section 7, we present our conclusions.

Legal knowledge representation of the securities and futures markets can help market participants understand the relationships between the laws, regulations, rules, criteria, and contracts in both markets. Access to legal document archives would be enhanced by providing investors with effective tools to search an environment of dynamically classified legal documents. In this paper, we integrate a dictionary-based term extraction technology, a vector space model for processing documents, and the GHSOM and LabelSOM algorithms to construct legal maps. Our method enables users to access legal documents in a convenient, efficient, and cost-effective way. We also enhance the visualization and readability of the GHSOM model by adding a tree display to show the topics generated by our topic selection module. In addition, information regarding the volume of documents clustered in each topic and a color display are shown on the maps to improve visualization. The maps can be browsed on the web site (http://synteny. iis.sinica.edu.tw/legalmap/). In this study, clustering of legal documents using the GHSOM model is shown to be effective, because most similar documents can be clustered together and the geographic distance between clusters seems to also represent the semantic distance. The results show that the model performs well in clustering Chinese language legal documents. In our experience, term extraction is the only critical difference between different language legal informatics. It is hard to identify terms in Chinese articles, because there are no spaces between terms. Some methodologies have been developed to resolve the problem, such as dictionary, linguistic, and statistical approaches (Ong et al., 1999). With regard to the clustering method, GHSOM, we find that there is no difference between Chinese and Western language formats, because the input vectors are composed of tf · idf weights generated by calculating the term and document frequencies. Also, the terms used have already been extracted by a term extraction technology. To evaluate the applicability of GHSOM to legal maps, we invited 15 experts4 to use the system and give some suggestions to us. After reviewing the respondents opinions, some users remarked that the distinction between nearby clusters is not clear enough for them to browse legal maps and needs to be improved, therefore, we suggest that future research should pay more attention to clarifying some terms that appear in most legal documents (e.g., ‘‘party’’ and ‘‘securities’’) in order to make the distinction between nearby clusters on the legal maps more obvious to users. This would reduce the difficulties that many users experience when browsing maps. Besides, they also suggested that the addition of keyword search and notification about amendments to laws, as well as more information about banking and insurance related laws and regulations, would make the system easier to retrieve legal information about the securities and futures markets. In legal informatics of oriental languages, we suggest that future research could apply other major methodologies from legal informatics of Western languages by developing an ontology defined in terms of specific users. For example, an ontology could be developed for people working in financial holding companies, and used to constructing a knowledge base for their specific purposes, including consumer financing, corporate financing, wealth management, etc.