Frantz Vichot

Using Machine Learning to Maintain Rule-based Named-Entity Recognition and Classification Systems

This paper presents a method that assists in maintaining a rule-based named-entity recognition and classification system. The underlying idea is to use a separate system, constructed with the use of machine learning, to monitor the performance of the rule-based system. The training data for the second system is generated with the use of the rule-based system, thus avoiding the need for manual tagging. The disagreement of the two systems acts as a signal for updating the rule-based system. The generality of the approach is illustrated by applying it to large corpora in two different languages: Greek and French. The results are very encouraging, showing that this alternative use of machine learning can assist significantly in the maintenance of rule-based systems.

Automatic processing of proper names in texts

This paper shows first the problems raised by proper names in natural language processing. Second, it introduces the knowledge representation structure we use based on conceptual graphs. Then it explains the techniques which are used to process known and unknown proper names. At last, it gives the performance of the system and the further works we intend to deal with.

Feeding a Financial Decision Support System with Textual Information

Many expert systems need a lot of data. For a long time this point has appeared to be a bottleneck for the growth of Artificial Intelligence applications. A major way to provide an expert system with knowledge is to enter it by hand. With the maturity of Natural Language Processing (NLP), a new way has been opened with automatic Information Extraction (IE) from text. This paper briefly presents a financial decision support system, named SAPE, connected with an IE system. This application is used by Caisse des dépôts et consignations (CDC) in order to anticipate takeover bids on the European stock markets. It provides ways to manage the highly complex and moving network of European shareholdings. SAPE is available on CDC groups intranet and is used by fund managers as part of their everyday work. This paper also describes how our NLP system, Exosème, uses the economic newswire from the Agence France-Presse (French press agency) in order to extract information on shareholdings and how this information is managed by the user to provide SAPE database with the large amount of information needed for its computations. After months of use, IE has appeared to be a powerful concrete solution. Moreover, if the economic value of takeover bids has lead us to pay particular attention to shareholdings, this approach can be extended to other events. In fact, with IE, new possibilities for portfolio decision support systems are coming. This paper presents the improvements we plan, and discusses those, though tempting, that are still out of reach due to the lack of adaptive tools.