My research activities leads with writing recognition and document analysis and interpretation.
 

• For writing recognition, the methods investigated are oriented towards lexical context modeling by using stochastic techniques. Hidden Markov models with first and second order, trajectory models and Random fields (NSHP-HMM) are experimented on unconstrained handwritten words for Latin and Arabic scripts. The pilot application concerns now the recognition of postal cheque amounts. In addition to this, we study the problem of multi-oriented and multi-scale character recognition using different neural network topologies including rotation transformation and dynamic neuron link changing.

• In document analysis, problems studied correspond to structure representation (physical and logical), to content comprehension and production. For structure modeling, we have used an unified representation of their objects based on international standards such as XML. The originality of our models is in the object abstraction (representation by abstract types) and in rules given to establish the correspondence physical-logical between conceptual objects and their representation. Our comprehension strategies are based on agent cooperation by hypotheses prediction and verification. We have developed several document analysis systems, such as Graphein which is a platform for block-document analysis such as scientific papers, Lib-more for structure recognition of library notices, and Bascet which is an emergent architecture used for the semantic validation of library references structures. Interfaces also constitute a big part of our research. We studied them in a system to help the user to introduce the generic model for documents and in an another system for Library interrogation. Some protocols are defined in these two applications. Different domains are investigated : Digital Libraries, administration, archives, etc.

• For document indexing and retrieval, our research is moving towards document reformulation and requalification. We focus our research on the user modeling and query interpretation and integration in the reformulation procedure.  A final part concerns the document reformating for electronic reedition on personal electronic supports. The topic concerns ancient documents.

• My research group is taking part at different european and national projects on document analysis and retrieval.