Project PERSonal Information Services (N9MI)
The PERSonal Information Services (PERSIS) project occupies itself with personalizing information services. This is necessary because the available amount of information in text, audio and video increases by the day. It’s necessary for information services to get to know their users first, before they are able to help them find that one pin in the ever growing haystack. The general thought is that once the information service knows you, it can serve you and your specific needs much better.
This 'getting to know you' process can take place in several ways. Examples are making someone grade a webpage that he visited, tracking the amount of time he uses to watch a television program or pop concert online, or making him create a user profile, thus finding out more about him.
People - System interaction
At the information systems that PERSIS develops, an optimal cooperation between person and system is always the goal. To reach this, it is essential to know how people treat information and what they are good and less good at. By studying users as well as psychological theories, the parts where humans are more important in the search process and the parts where a machine can take over are determined. After all, two know more than one. PERSIS in fact takes the step from information retrieval to information interaction.
Social tagging
The human role is not only important in the search process, but also in the information describing part. People can add keywords to video or other information, creating a form of social tagging. Also studied is how these tags added by users can help to improve automatically generated metadata.
Recommendations cut to size
The PERSIS project personalizes information services, for example by applying an automatic recommendation system. For a personal recommendation it's important to know the person you make the recommendation for, as well as the products that you recommend. Therefore the previously mentioned user profiles and metadata are essential. A recommendation system learns what interests a specific person and compares this personal preference to metadata from, for example, a television program to determine whether this program is interesting for this person.
The system also compares people to each other. When two persons have similar profiles, the rate that one person gave to a program probably says something about how interesting this program is to the other person.
Participants
On the PERSIS worktable, researchers from the Telematica Institute, TU Delft and TU Eindhoven are working together with Roessingh Research and Development, IBM, SURFnet, and Fabchannel.
Learning Features (N1)
Multimodal Interaction (N2)
Ambient Multimedia Databases (N3)
Semantic Multimedia Access (N5)
Professional's Dashboard (N6)
Video At Your Fingertips (N7)
E-Culture (N9C)
PERsonal Information Services (N9MI)