Towards Cross-Domain Performance Modeling and Prediction: IR/RecSys/NLP

Summary

Information systems, which manage, access, extract and process non-structured information, typically deal with vague and implicit information needs, natural language and complex user tasks. Examples of such systems are information retrieval (IR) systems, recommender systems (RecSys), and applications of natural language processing (NLP) such as e.g. machine translation, document classification, sentiment analysis or search engines. The discipline behind these systems differs from other areas of computer science, and other fields of science and engineering in general, due to the lack of models that allow us to predict system performances in a specific operational context and to design systems ahead to achieve a desired level of effectiveness. In the type of information systems we want to look at, we deal with domains characterized by complex algorithms, dependent on many parameters and confronted with uncertainty both in the information to be processed and the needs to be addressed, where the lack of predictive models is somehow bypassed by massive trials of as many combinations as possible.

These approaches relying on massive experimentation, construction of testbeds, and heuristics are neither indefinitely scaled as the complexity of systems and tasks increases nor applicable outside the context of big Internet companies, which still have the resources to cope with them.

The workshop was organized as follows. The first day was devoted to plenary talks focused on providing a general introduction to IR, RecSys, and NLP and on digging into some specific issues in performance modeling and prediction in these three domains. The second day, participants split into three groups - IR, RecSys, and NLP - and explored performance modeling and prediction issues and challenges within each domain; the working groups then reconvened to present the output of their discussion in a plenary session in order to cross-fertilize across disciplines and to identify cross-discipline themes to be further investigated. The third day, participant split into groups which explored these themes - namely measures, performance analysis, documenting and understanding assumptions, application features, and modeling performance - and reported back in plenary sessions to keep all the participants aligned with the ongoing discussions. The fourth and fifth days have been devoted to the drafting of this report and the manifesto originated from the workshop.

Acknowledgements. We thank Schloss Dagstuhl for hosting us.

References

1. N. Ferro, N. Fuhr, G. Grefenstette, J. A. Konstan, P. Castells, E. M. Daly, T. Declerck, M. D. Ekstrand, W. Geyer, J. Gonzalo, T. Kuflik, K. Lindén, B. Magnini, J.-Y. Nie, R. Perego, B. Shapira, I. Soboroff, N. Tintarev, K. Verspoor, M. C. Willemsen, and J. Zobel. Manifesto from Dagstuhl Perspectives Workshop 17442 – Towards Performance Modeling and Performance Prediction across IR/RecSys/NLP. Dagstuhl Manifestos, Schloss Dagstuhl–Leibniz-Zentrum für Informatik, Germany, 7(1), 2018.