Knowledge engineering

List institute develops non structured data automated analysis and description tools for knowledge extraction and delivery to the user under the form of an exploitable synthesis. We work on texts for translation, summary or social networks scanning purposes. We also work on image search and indexation, on documents semantic processing or media flows.

Our researchers collaborate with specialised industrial companies on IT monitoring, electronic document management, marketing or tourism.

Among our academic partners

LIMSI-CNRS (Saclay, France), Ecole Centrale Paris (France), Mines Paris (France), Carnegie Mellon University (Pittsburg, USA)

Assets

• Ability to process multimedia documents : text, image, video, speech
• Ability to manage cross and multi-linguism (10 languages)
• Choice on technology (algorithms and architecture) to support scale-up (Big Analytics)
• Ability to take into account a domain’s specificities to go beyond market’s generic tools in terms of precision performances.
• Successful participation to international evaluation assessment campaigns among the best laboratories worldwide

Major technologies

Text search engine

Description

AMOSE (All Media One Search Engine) software platform is a semantic search engine based on LIMA (LIST Multilingual Analyzer) language analysis tool. AMOSE offers innovative functionalities for multimedia documents’ search. The tool is able to understand the words semantic in their context inside the request and index, making thus the research results even more relevant. It is what we call a “cross-language” tool which means that you can have a request in a given language and get results from different languages’ sources of information among the 10 programmed languages including Arabic and Chinese. AMOSE offers also a dynamic presentation of the results under the form of grades of relevance that classifies the documents according to their common terms in sub-groups. It can also organize them in theme grades meaning that it will group the documents sharing a same topic.

Applications

Search engines

 

Search results’ clustering

Description

Texts documents like patents or technical annexes are being analysed for automatic and dynamic ranking purposes without any necessity of previous information content or topic indications.

Applications

Documents fast thematic analysis

Image-based objects’ recognition

Description

Image-based objects recognition is a process that consists in automated association of image and object name through photography. List institute has developed ELISE platform, a hybrid objects recognition tool that gathers in the same architecture object categories recognition (ex: planes) and object instances recognition (ex: A380). It is based on computer vision innovative technologies: scalable DeepLearning for object categories and fast matching of points of interest for object instances. ELISE distribution capacities help annotating and searching images among large-scale data bases (100 million images in a very precise and fast way. These capacities have been demonstrated several times on objects search/recognition and vision geolocation international evaluation campaigns.

Applications

Personalised advertising, Web monitoring or e-business