SEARCH AND RECOMMENDATION
Human-centric search and recommendation interfaces are key to enabling a world of ambient intelligence
Related Content
In a world where location and context are understood, digital technologies will proactively propose and recommend activities, places and things, helping people to interact and navigate in the physical environment. This digital recommendation and guidance should be as seamless and natural as possible. Our main objective is to translate this ambition into the design of context-aware, personalised and anticipatory search and recommendation modules.
New user interfaces and sensors such as voice-based interfaces, touch screens, accelerometers and geo-localisation have largely modified how people interact with search and recommendation engines. Our main research direction aims at exploiting this new information on user context, behaviour and intention, to build a representation that captures it. We then design novel families of adaptive search and recommendation components to dramatically improve personalisation. We’re also developing prediction models to better anticipate future needs and evolving preferences.
We consider that search and recommendation processes are not one-shot, single-interaction problems. We envision them as multi-stakeholder, multi-round games (continuous learning loops), where the engine and the user interplay collaboratively with the different types of feedback they each provide. Facilitating communication between the computer and the user naturally leads to the design of multimodal strategies where text in any language, images, sound and speech can be handled jointly and consistently. With the same communication objective in mind, and in particular to establish mutual confidence, we focus on developing accountable algorithms that can explain the advice they provide in a transparent and understandable form, while preserving fairness and diversity.
Our research is carried out with the NAVER search group who are responsible for the world’s 5th biggest search engine. This provides research opportunities that go far beyond the traditional Information Retrieval framework.
Examples of the projects we’re working on are:
- POI Search and Recommendation in a Location-Based Social Network mixing geospatial information with social network analysis
- Personalised News Recommendation that anticipates time-varying user preferences, and preserves diversity to avoid any echo chamber
- Cross-modal (text-image) Search based on cross-media similarities and an adaptive re-ranking strategy.
Search and Recommendation team:
