{ Thomas Heitz }

Presentation

♦ Publications ♦ Softwares

Who I am ?

• I work at Qwam CI as a data scientist. Before I worked as a research assistant for the University of Sheffield NLP Team and as a consultant for Ontotext. I use notably GATE, one of the main open source frameworks in NLP.
• You can consult my [HTML] résumé.
• And if really curious my [SITE] personal page.

What I do ?

• Objective: to make computers able to understand texts.
• Keywords: Natural Language Processing, Text mining, Text Analytics, Computational Linguistics.
• Applications: Search Engine, Email Filtering, Intelligence, etc.

How to contact me ?

• thomashtz123@gmail.com

Publications

♦ Presentation ♦ Softwares

Reviewed papers in international conferences

• 2006. Modélisation du prétraitement des textes.
  T. Heitz. In Proceedings of JADT'06 (International Conference on Statistical Analysis of Textual Data), volume 1, pages 499-506, 2006. [PDF] Présentation  [PDF] Poster  [PS] Article  [BIB] BibTeX
• 2004. EXIT: Un système itératif pour l'extraction de la terminologie du domaine à partir de corpus spécialisés.
  M. Roche, T. Heitz, O. Matte-Tailliez, Y. Kodratoff. In Proceedings of JADT'04 (International Conference on Statistical Analysis of Textual Data), volume 2, pages 946-956, 2004. [PDF] Présentation  [PS] Article  [BIB] BibTeX

Reviewed papers in international workshop

• 2008. Large-scale, Parallel Automatic Patent Annotation.
  Agatonovic, M., Aswani, N., Bontcheva, K., Cunningham, H., Heitz, T., Li, Y., et al. In Proceedings of 1st International CIKM Workshop on Patent Information Retrieval – PaIR'08. Napa Valley, California, USA, pages 1-8. Admission rate: 7/16. [PDF] Presentation  [PDF] Paper  [BIB] BibTeX

Papers in international challenges

• 2004. From the texts to the concepts they contain: a chain of linguistic treatments.
  Amrani A., Azé J., Heitz T., Kodratoff Y., Roche M. In proceedings of TREC'04 (Text REtrieval Conference), National Institute of Standards and Technology, Gaithersburg Maryland USA, p712-722, 2004. [PDF] Article  [BIB] BibTeX

Reviewed papers in national conferences

• 2005. Extraction de termes centrée autour de l'expert.
  Heitz T., Roche M., Kodratoff Y. Dans la revue RNTI, numéro spécial EGC'2005 (session démonstrations), 2:685-690. 2005. [PDF] Présentation  [BIB] BibTeX
• 2004. EXIT : EXtraction itérative de la terminologie.
  M. Roche, T. Heitz, O. Matte-Tailliez, Y. Kodratoff. Revue RNTI, numéro spécial EGC'2004 (résumé), 2:478, 2004. [PDF] Poster  [PDF] Présentation  [BIB] BibTeX

Miscellaneous

• 2008. Thèse intitulée Une méthode récursive pour le prétraitement des textes. [PDF] Présentation [PDF] Thèse 

Softwares

♦ Presentation ♦ Publications

GATE, General Architecture for Text Engineering

• I contribute to GATE since 2007, notably the user interface.
• [JAVA1.6] Download page.

EXIT 1.0, EXtraction Iterative of the Terminology

• Screenshots: Expressions Panel, Extraction Panel
• Document: Manuel, First start guide
• Presentation: EGC'2005
• [JAVA1.4] Download archive

Syntactic relations analyzer for French and English

• Transformation d'un texte étiqueté grammaticalement par Brill en relations syntaxiques de type Shallow Parser de Xerox.
• Documentation
• [PERL] Download archive

Data and MetaData annotation GATE plugin

• Its purpose is to annotate each line of a corpus of texts into data and metadata. You will have to tag a sample (Parameter: corpusSampleDocument) of the full corpus in step 2. At the end of the process, step 4, you will have all files in the corpus (Parameter: corpus) rewritten in new files with XML tags. All files will be put in the results directory (Parameter: resultsDirectoryURL).
  1. Choose if you start with or without initial annotations
  2. Prepare corpus to annotate
  3. Learn & Show learning errors
  4. Apply learned model on full corpus
• [JAVA1.4] Download plugin for Gate.

Sentences segmentation GATE plugin

• Its purpose is to segment a corpus of texts into sentences but not on abbreviations. You will have to tag a sample (Parameter: corpusSampleDocument) of the full corpus (Parameter: corpus), steps 2, 3, 4, 5 and 6. At the end of the process, step 7, you will have all files in the corpus (Parameter: corpus) rewritten in new files with one sentence per line and two lists of abbreviations and capitalized words. All files will be put in the results directory (Parameter: resultsDirectoryURL).
  1. Compute statistics on the full corpus
  2. Use lists and statistics to do a first tagging of abbreviations
  3. Learn and show the learning error on abbreviations
  4. First tagging of capitalized words
  5. Learn and show the learning error on capitalized words
  6. Determine Cap. words according to Abbreviations and conversely
  7. Apply learned model on full corpus to segment sentences
• [JAVA1.4] Download plugin for Gate.