What is QUARTZ
QUARTZ (Quantum Information Access and Retrieval Theory) is an Innovative Training Network (ITN) that aims to educate its Early Stage Researchers to adopt a novel theoretically and empirically motivated approach to Information Access and Retrieval (IAR) based on the quantum mechanical framework that gives up the notions of unimodal features and classical ranking models disconnected from context.
The current state of the art of IAR is not sufficient to address the challenges of a dynamic, adaptive and context-aware user-machine interaction and to make a major breakthrough in the overall effectiveness of retrieval systems, and that a genuine theoretical breakthrough is on the contrary necessary.
We believe that this breakthrough can be provided by quantum theory which can integrate abstract vector spaces, probability spaces and logic in a single theoretical framework which extend and generalize the classical vector, probability and logic spaces utilised in IAR.
Who is participating
University of Padova, Italy
Prof. Massimo Melucci (coordinator)
The Open University, The United Kingdom
University of Bedfordshire, The United Kingdom
Vrije Universiteit Brussel, Belgium
University of Copenhagen, Denmark
Linnæus University, Sweden
What we do in QUARTZ
QUARTZ consists of 13 individual ESR projects to investigate theoretical issues, evaluate methods and prototypes for adaptive IAR systems managing large data collections and meeting the end user's information needs in a dynamic context.
Moreover, QUARTZ consists of training activities and ESR research projects which investigate theoretical issues and evaluate methods and prototypes for adaptive IAR systems managing large data collections and meeting the end user's information needs in a dynamic context.
In addition, QUARTZ complements the doctoral schools offered by the organisations by a number of additional network training activities to help the ESRs to achieve the best scientific and technological preparation in IAR and the quantum mechanical framework, international relationships with leading research groups in Europe and outside Europe, transversal skills useful to disseminate their own research work at the highest level possible, and to get a deep understanding of the strategic priority of “Open Science”.
In particular, secondments are offered and strongly suggested to every ESR at some participants. All ESRs should attend the QUARTZ school, several workshops where they can introduce their research, and various courses to achieve important transversal skills.
How QUARTZ works
The 13 individual research projects are inter-related across the following Work Packages (WPs) toward a common, big objective:
WP1: Geometry and Deep Learning of Vector Spaces for Multimodal IAR
The main objective of this work-package is to define abstract vector spaces to build multimodal complex features, operators for interactive search of multimodal information in abstract vector spaces, a quantum measurement-inspired ranking function in abstract vector spaces, an integrated theoretical framework and computational methods for learning an adaptive representation of the multimodal content and context, and an adaptive retrieval function in the multimodal context. Prototype software libraries will be implemented to assess the proposed computational methods
This work-package aims at providing evidence that the quantum theoretical framework can describe and process rich interaction models considering vague and evolving information needs based on Information Foraging Theory and a polyrepresentative model that is capable to react to different forms of user interaction. It also investigates and applies the dependencies between different representations and sources of evidence. An information access language with formal syntax and semantics based on modal logic and quantum logic will be developed. For a computer-based reasoning, an appropriate calculus needs to be developed with proven soundness and completeness.
WP3: Quantum Probability and Ranking in IAR
This work-package defines the quantum-like properties when the users' relevance measurements are uncertain, interfered by the context factors, or evolving with the user's progress in searching and interacting with information. These findings will be used to predict the user's cognitive status and relevance measurements in context sensitive search scenarios. Another result will be mathematical tools for considering and quantifying the degree of uncertainty in probability estimates of document features that are used for ranking. Also, ranking decision algorithms when the input arguments may be fraught with uncertainty will be designed. Another objective is to develop a mathematical framework for: switching between linguistically naive and linguistically more robust representations of document semantics; integrating a formal modeling of this decision to switch into ranking; producing a new family of ranking models that are closer to human cognition than presently possible.
In this work-package, we address all the evaluation issues and the efforts made in the individual research projects. Every project will address evaluation and industrial applications to some extent depending on the characteristics of the project itself or the subject thereof. All Early Stage Researchers should take into account the real user requirements, apply and evaluate their methods in relevant non-academic partners' real world application scenarios. Some theoretical project may require a kind of evaluation (e.g., simulation) different from the evaluation needed for a more application-oriented project (e.g., test collections or user studies). (ESR-x is an ESR funded by a partner organisation.)