Projects > Search engines > Mobile Spoken Question Answering

Mobile Spoken Question Answering project

Development of a mobile search engine that can answer spoken questions 

The project was part of an integrated project (POSH) in the Center for TeleInFrastruktur (CTIF) at Aalborg University, Denmark. The CEO of Legind Technologies developed as professor of computer science at the university and leader of the CTIF Wireless Computing and Security group the knowledge-based question understanding and answering system with a team of colleagues and master students, while another CTIF group had focus on the speech recogniser.

In the project, we developed a prototype of knowledge-based mobile search engine that can answer spoken queries in a specified domain. For the prototype, we selected the domain of football (soccer). The system can understand football queries in Danish and provide answers from resources in Danish.

Central to the knowledge-based question understanding and answering was (1) the ontology representation of the football domain, (2) the ongoing monitoring of football relevant news and extraction of information from these news, (3) the interpretation of question in terms of the ontology, the stored information extracted and the current context, and (4) providing cooperative answers to the questions.

For instance, a user may ask (spoken question) “Who won in Esbjerg?”. In the context that an important football game recently was played in Esbjerg, the system interprets the question into “Who won the last soccer game in the national premier league played in Esbjerg?” and answers cooperatively “Esbjerg won 3‐2 over Kolding. The goals were: (23 min, to 1‐0) scored by Jakob Ankersen, …” thus providing useful information, instead of only the direct answer “Esbjerg”.

A paper from this project