Multi-Document Summarization systems must produce
summaries about different texts keeping track of redundant
information but also contradictory information and also additional
information. We are particularly interested in finding redundant
information as well as contradictory information using techniques
such as text assymetric similarity and textual entailment.
Multimedia Multimodal Information Retrieval
Information Retrieval is an important trend in research. We are particularly interested in Spatio-Temporal Image Retrieval, which objective is to retrieve
web images for a text query with a spatio-temporal intent. For that purpose, multimodal (text and image) characteristics are used.
Efficient Computing of Non-Contiguous Multiword
Computing statistics of non-contiguous substrings
can be overwhelming when gigabytes of texts may be processed. For this purpose, specific algorithms must be proposed to improve the computation
efficiency of multiword units. For that purpose, we study specific text algorithms which may handle contiguous and non-contiguous strings.
Mobile Adaptative Web
interfaces in the context of Human Computer Interaction is
important with new mobile devices. In this context, we are
particularly interested in studying personalized and collaborative web search using micro-ontologies.
Automatic Construction of Ontologies
Terminological ontologies are built to allow reasoning about texts. In this context, we are
particularly interested in studying the automatic construction of ontologies based on a pretopological framework with different types of similarity measures.
Haptic Access to Web Pages
Visually Impaired People (VIP) are info-included. In the context of search engines, we are
particularly interested in finding new haptic ways to interact with web pages. The overall idea is to allow blind or disabled users to understand the web
page layouts without seeing them.
Search Engines for Kids
The WWW is full of prohibited contents for children. In this context, we are particularly
interested in developping methodologies that correctly identify prohibited contents across domains.