Refine
Year of publication
- 2013 (2) (remove)
Document Type
- Doctoral Thesis (2) (remove)
Keywords
Institute
- Institut für Informatik (2) (remove)
This dissertation investigates the usage of theorem provers in automated question answering (QA). QA systems attempt to compute correct answers for questions phrased in a natural language. Commonly they utilize a multitude of methods from computational linguistics and knowledge representation to process the questions and to obtain the answers from extensive knowledge bases. These methods are often syntax-based, and they cannot derive implicit knowledge. Automated theorem provers (ATP) on the other hand can compute logical derivations with millions of inference steps. By integrating a prover into a QA system this reasoning strength could be harnessed to deduce new knowledge from the facts in the knowledge base and thereby improve the QA capabilities. This involves challenges in that the contrary approaches of QA and automated reasoning must be combined: QA methods normally aim for speed and robustness to obtain useful results even from incomplete of faulty data, whereas ATP systems employ logical calculi to derive unambiguous and rigorous proofs. The latter approach is difficult to reconcile with the quantity and the quality of the knowledge bases in QA. The dissertation describes modifications to ATP systems in order to overcome these obstacles. The central example is the theorem prover E-KRHyper which was developed by the author at the Universität Koblenz-Landau. As part of the research work for this dissertation E-KRHyper was embedded into a framework of components for natural language processing, information retrieval and knowledge representation, together forming the QA system LogAnswer.
Also presented are additional extensions to the prover implementation and the underlying calculi which go beyond enhancing the reasoning strength of QA systems by giving access to external knowledge sources like web services. These allow the prover to fill gaps in the knowledge during the derivation, or to use external ontologies in other ways, for example for abductive reasoning. While the modifications and extensions detailed in the dissertation are a direct result of adapting an ATP system to QA, some of them can be useful for automated reasoning in general. Evaluation results from experiments and competition participations demonstrate the effectiveness of the methods under discussion.
The amount of information on the Web is constantly increasing and also there is a wide variety of information available such as news, encyclopedia articles, statistics, survey data, stock information, events, bibliographies etc. The information is characterized by heterogeneity in aspects such as information type, modality, structure, granularity, quality and by its distributed nature. The two primary techniques by which users on the Web are looking for information are (1) using Web search engines and (2) browsing the links between information. The dominant mode of information presentation is mainly static in the form of text, images and graphics. Interactive visualizations offer a number of advantages for the presentation and exploration of heterogeneous information on the Web: (1) They provide different representations for different, very large and complex types of information and (2) large amounts of data can be explored interactively using their attributes and thus can support and expand the cognition process of the user. So far, interactive visualizations are still not an integral part in the search process of the Web. The technical standards and interaction paradigms to make interactive visualization usable by the mass are introduced only slowly through standardatization organizations. This work examines how interactive visualizations can be used for the linking and search process of heterogeneous information on the Web. Based on principles in the areas of information retrieval (IR), information visualization and information processing, a model is created, which extends the existing structural models of information visualization with two new processes: (1) linking of information in visualizations and (2) searching, browsing and filtering based on glyphs. The Vizgr toolkit implements the developed model in a web application. In four different application scenarios, aspects of the model will be instantiated and are evaluated in user tests or examined by examples.