What internet user has never dreamt of easily finding the answers to existing questions that are close to those they are asking using their own words? On the web there are many areas (FAQ, forums, etc.) where one can find the answers to questions that have already been asked by other people. However, in order to process these answers, the question closest to the one we are asking still needs to be found. Today though we must still ask questions using words that are already indexed by search engines. But maybe not for long…
In effect, semantic similarity techniques enable us to overcome this hurdle by replacing words with their context of occurrence. Frédéric Herledan, a project manager at Orange, is working on the question, notably through two practical applications: the improvement of searches on technical assistance forums, and the offer of a new way to be informed and to learn via radio programmes by processing audio fonds. Applications that should enable the enhancement of Orange’s assistance forum and the optimisation of usage via the Djingo personal assistant, for example. A project that was presented at the latest Orange Salon de la recherche from 5 to 7 December 2017.
Listen to the podcast
