Skip to Main content Skip to Navigation
Theses

Application du raisonnement à partir de cas à l'analyse de documents administratifs

Hatem Hamza 1
1 READ - READ
LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
Abstract : This thesis deals with administrative document analysis and recognition. The continuous arrival of documents lead us to choose a methodology taking into account the previous processing experiences.We chose case-based reasoning for this reason. After extracting the document's structures like adresses, amount zones and tables, a document model is built as a graph, representing the problem to be solved. This problem is then compared to a document case base using graph probing. If a similar case exists, it is then adapted to analyze and interpret the current case. Otherwise, a structure by structure analysis is done using a document structure case base. The continuous arrival of data requires an incremental learning scheme that could be done as processing goes on. For this purpose, we proposed an improvement of an already existing neural network called Incremental Growing Neural Gas. This improvement consisted in taking into account only the local neighborhood of the nearest neuron while creating a new neuron. The proposed neural network was successfully tested on real documents (invoices, forms) and other synthetic data. This thesis was done thanks to a collaboration with the company ITESOFT. All the steps of the proposed approach were tested on real cases.
Document type :
Theses
Complete list of metadatas

Cited literature [98 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00586317
Contributor : Hatem Hamza <>
Submitted on : Friday, April 15, 2011 - 3:07:46 PM
Last modification on : Tuesday, April 24, 2018 - 1:54:43 PM
Long-term archiving on: : Saturday, July 16, 2011 - 2:54:13 AM

Identifiers

  • HAL Id : tel-00586317, version 1

Collections

Citation

Hatem Hamza. Application du raisonnement à partir de cas à l'analyse de documents administratifs. Génie logiciel [cs.SE]. Université Nancy II, 2008. Français. ⟨tel-00586317⟩

Share

Metrics

Record views

535

Files downloads

2786