Information Access and Retrieval Meets Quantum Mechanics

When facing a problem, everyone needs information and searches out the most relevant information to her needs. Due to the big amount of users, needs and data, computerized methods are unavoidable. Information Retrieval (IR) is... [ view full abstract ]

When facing a problem, everyone needs information and searches out the most relevant information to her needs. Due to the big amount of users, needs and data, computerized methods are unavoidable. Information Retrieval (IR) is the science to design and engineer systems that represent and retrieve all and only relevant information in any context (perfect retrieval). However, what a user assesses as relevant at a certain time, in a certain location or with a certain intent will di er from what is relevant to another user, at another time, in another location, with another intent. Relevance cannot exactly be measured outside context unlike other less context-sensitive observables (e.g. pertinence). If relevance can be measured only in a given context, other observables may interfere. E.g., IR is based on classical probability, thus events are as subsets of a single sample space; this implies that the probability of an event A (e.g. relevance) can never be less than the probability of the conjunction of A with another event B (e.g. pertinence). However, violations of this law have been found in empirical studies, but disappears in quantum probability. For 15 years, Quantum Mechanics (QM) has also been investigated in human cognition, natural language processing and IR among others. Many questions are still outstanding. Can perfect IR be achieved by QM? Should modalities (e.g. image clicking) be modeled addressed by classical probability as done for words? Will context create a major shift in how probability is viewed within IR? Constructing classical probabilistic models involves joint probability distribution over variables: does this joint distribution always exist? These issues are under investigation within an EU MSCA-ITN project coordinated at the Department of Information Engineering of the University of Padova, Italy (www.quartz-itn.eu).