Cerca nel sito
ICube   >   Agenda : Thèse : Inventive Solutions Retrieval from Patent documents via Natural Language Processing

Thèse : Inventive Solutions Retrieval from Patent documents via Natural Language Processing

17 dicembre 2021

Thèse : Inventive Solutions Retrieval from Patent documents via Natural Language Processing

Équipe :

Le 17 décembre 2021 10h dans l'amphiteatre de l'INSAS (24 BD de la Victoire, 67000 Strasbourg), aura lieu la soutenance de thèse de Xin NI un doctorant de l'équipe .


Innovation is a key factor for companies developing products and engaging in continuous progress in a highly competitive market. In recent years, in the context of this growing concern for engineering innovation, the demand for inventive engineering solutions has been increasing rapidly in companies. Besides, a large number of published patent documents from wider domains tend to contain the latest inventive knowledge in the world. Mining this sort of knowledge is a significant way to enable industrial innovation. It is also an important alternative to brace the complex manufacturing challenges.


Nevertheless, it is always a significant challenge for engineers without a broad

understanding of different domain knowledge to make full use of the inventive knowledge contained in patent documents. Especially, exploring several patents by an expert rapidly turns to be an arduous task. Theory of Inventive Problem Solving (TRIZ) was proposed to provide a logical approach to enhance creativity. However, its lack of formalization and complex principles generate a huge obstacle to implementing it, even for engineers to understand it. 


In order to address the aforementioned challenges, in the thesis, we aim to automate the entire inventive problem-solving process by using patent documents based on Natural Language Processing (NLP) techniques. In particular, we propose four main contributions: 1. two similar problem retrieval models called IDM-Similar based on Word2vec neural networks and SAM-IDM based on LSTM neural networks are proposed to retrieve similar problems from different domain patents; 2. a problem-solution matching model named IDM-Matching according to XLNet neural networks is proposed to build connections between problems and solutions in patent documents; 3. an inventive solutions ranking model called PatRIS based on multiple criteria decision analysis approach is proposed to rank potential inventive solutions; 4. a software prototype named PatentSolver combining aforementioned models is developed to provide engineers with a real tool to prepare inventive solutions from patent documents. These models have been evaluated on both benchmark and real-world patent datasets.


Les membres du Jury sont :

  • Directeur de thèse: Prof. Denis Cavallucci, Professeur, INSA Strasbourg 
  • Encadrant de thèse: Dr. Ahmed Samet, Maître de conférences, INSA Strasbourg
  • Rapporteur externe: Prof. Cecilia Zanni-Merk, Professeur, INSA Rouen 
  • Rapporteur externe: Prof. Hervé Panetto, Professeur, Université de Lorraine
  • Examinateur: Prof. Pierre Collet, Professeur, Université de Strasbourg
  • Examinateur: Dr. Florence le Ber, Maître de conférences HDR, ENGEES
  • Examinateur: Dr. Magali Barbaroux, Responsable de programmes de recherche, Sartorius
  • Invité: Dr. Marion Moliner, Data science manager, Data Excellence Factory, Hager Group


À la une

Le dépôt des candidatures pour les postes d’enseignants-chercheur est ouvert. Les offres sont...

Flux RSS

Flux RSS