Artificial Intelligence

Research activities:

Natural language processing provides technologies for the analysis and processing of text documents.  These techniques can be exploited in applications such as question answering, information extraction from documents, and text summarization and translation.

Core research is being conducted in the following areas:

  • natural language parsing
  • semantic parsing
  • part-of-speech tagging

People: Giorgio Satta (contact person)

Preferences are a central concept of decision making. Preference models are needed in decision-support systems such as web-based recommender systems, in automated problem solvers such as configurators, and in multi-agent scenarios where we need to combine preferences given by several agents. Preferences are studied in many areas of artificial intelligence such as knowledge representation, multi-agent systems, computational social choice, constraint satisfaction, preference elicitation, decision making, planning, scheduling, timetabling, resource allocation, and stable matching problems.

The research in this area focuses on the following lines of research:

  • Design of models to handle in a compact way preferences on combinatorial domains when there is incomplete preference information.
  • Preference elicitation procedures to extract only relevant missing preferences in scenarios with incomplete preference information.
  • Algorithmic techniques to solve problems with preferences and constraints such as, scheduling, timetabling, and resource allocation problems.
  • Design of preference aggregation methods in multi-agent scenarios where each agent expresses his preferences over a large number of alternatives and some preferences are missing.
  • Study of the computational complexity of manipulation, influence, and bribery for voting rules over a large number of alternatives when some agent preferences are missing or when the agents express their preferences in a compact way.
  • Design of methods for finding stable matchings between two groups of elements, where each element of one group expresses preferences over all the members of the other group.

 People: Maria Silvia Pini (contact person)

  • Models of integration of quantitative and qualitative imperfect temporal information, based on the FCSP (Fuzzy Constraint Satisfaction Problem) paradigm;
  • Development of automated systems capable of representing and reasoning about temporal knowledge in presence of uncertainty and vagueness; application to planning and scheduling problems and to medical diagnosis;
  • Probabilistic reasoning with Bayesian Networks; study of how to build a Bayesian Network by hybridizing complete search with indipendence tests; Bayesian Network formalism as a means for identifying causal relations between a set of random variables.

People: Silvana Badaloni (contact person)