Text Simplification of Patent Documents

Résumé

This paper represents an automatic text simplification system for patent documents. The simplification system is embedded in the broader context of an information retrieval system which extracts IDM related knowledge from patent documents. Extracting elements of IDM ontology from patents involves training machine-learning model. However, an accuracy of the model is compromised when the given text is too long, hence the need of simplifying the texts to improve machine learning. There have been precedent studies on automatic text simplification based on hand-written rules or statistical approach. However, few researches addressed simplifying patent documents. Patent document has its particularity in its lengthy sentences and multiword expression terminology, which often hinder accurate parsing. Therefore, in this research, we present our method to automatically simplify texts of patent documents and scientific papers by analyzing their syntactic and lexical patterns.

BibTeX

@inproceedings{kang:hal-02279758,
  TITLE = {{Text Simplification of Patent Documents}},
  AUTHOR = {Kang, Jeongwoo and Souili, Achille and Cavallucci, Denis},
  URL = {https://inria.hal.science/hal-02279758},
  NOTE = {Part 6: TRIZ and Patenting},
  BOOKTITLE = {{IFIP Advances in Information and Communication Technology}},
  ADDRESS = {Strasbourg, France},
  EDITOR = {Denis Cavallucci and Roland De Guio and Sebastian Koziolek},
  PUBLISHER = {{Springer International Publishing}},
  SERIES = {Automated Invention for Smart Industries},
  VOLUME = {AICT-541},
  PAGES = {225-237},
  YEAR = {2018},
  MONTH = Oct,
  DOI = {10.1007/978-3-030-02456-7\_19},
  KEYWORDS = {Text mining ; Syntactic analysis ; Text simplification ; Information extraction ; TRIZ ; Inventive Design Method},
  PDF = {https://inria.hal.science/hal-02279758v1/file/474537_1_En_19_Chapter.pdf},
  HAL_ID = {hal-02279758},
  HAL_VERSION = {v1},
}