PGR: A Cooperative Legal IR System on the Web

In this paper we present a web based legal information retrieval system which is able to support cooperative interactions between users and a legal knowledge base. The legal knowledge base is composed by four levels of knowledge: Text knowledge; Domain knowledge; Information-retrieval knowledge; and Interaction Knowledge. The Text Level has knowledge about the words and sequence of words that are in each text of the knowledge base. The Domain level includes knowledge about the text domain such as juridical knowledge (for instance, under which conditions a pension for relevant services may be given to someone). It has rules that encoded the domain knowledge, such as: normally those pensions are attributed to militaries or to civilians (firemen, doctors, nurses, etc.) and the conditions to be fulfilled by them are different. The Information Retrieval Level includes knowledge about what we should expect to find in texts about a subject, for instance that in texts about pensions for relevant services, the pension may be attributed or refused. The Interaction level is responsible for the multimodal interaction management. This includes the ability of the system to infer user intentions and attitudes and to build the representation of the user interaction. Introduction In this paper we present a web based legal information retrieval system which is able to support cooperative interactions between users and a legal knowledge base. The legal knowledge base is composed by four levels of knowledge: 1. Text knowledge; 2. Domain knowledge; 3. Information-retrieval knowledge; 4. Interaction Knowledge. The text level is composed by the set of documents produced by the Portuguese Attorney General since 1940. The domain level includes knowledge about specific laws. For instance, we have rules to represent under which conditions a pension for relevant services may be given to someone. We also have a taxonomy of legal concepts constructed by our partner the Portuguese Attorney General Office, which has more than 6000 concepts. As another source of domain knowledge we have built a list of relevant expressions from the documents using statistical methodologies. The information retrieval level includes knowledge about what we should expect to find in texts about a specific subject. The idea is to find classes (or clusters) of concepts. For instance, in texts about pensions for relevant services, one important feature is the fact the pension may be attributed or refused. The interaction level is responsible for the interaction management. It is composed by rules that allow the system to infer the interaction structure and the user attitudes (beliefs and intentions). As we have different levels of knowledge we need different representation techniques and inference tools: 1. Text level – Texts are represented by HTML documents and the SINO search engine (from AustLII) is used to perform inferences. SINO was changed in order to deal with the Portuguese Language. Namely, the new system uses the Portuguese lexicon (more than 900,000 words) to handle morphological errors and to obtain the base queried word. Moreover, we are using a Portuguese tagger to disambiguate the morphological categories of the word. 2. Other levels – Knowledge is represented by logic programs and inference is done through the use of dynamic logic programming semantics (Alferes et al.). Dynamic logic programming defines how a logic program can be updated by other logic programs. In our approach, each event is represented by logic facts, which are used to update the previous program and to obtain a new one. The interaction manager is basically a logic programming module, which communicates with the user and the different knowledge bases. Communication with the user is done through a web interface using CGI and the Java language. Cooperation is achieved through the inference of the user attitudes behind the user acts. These inferred attitudes allow the system to interact with the legal knowledge base in order to refine the user query and to be more helpful in the answer. For instance, if the user act is to search for documents with expression: User A1: pensions for relevant services The system may suggest a way for the user to specify the query: System A2: Pensions that were granted, refused or both? The user may select any of the suggestions User A3: Both. The system will consider that the user does not want to refine its query any more: System – A4: There are 129 documents, listed in the right frame. Another cooperative characteristic is the possibility to manage the context of the interaction and to try to join one action with the previous context: