Last week, Eneko Sabaté and Roger Canal attended the SISAP 2026 international conference, held in Brno (Czech Republic), to present two short papers based on research they had carried out.
SISAP is an international conference specialising in similarity search, an area of computer science that focuses on finding similar elements within large datasets. This technology enables, for example, the identification of documents with similar content, the finding of similar images or the retrieval of results related to a given search.
During the conference, researchers, doctoral students and professors shared the results of their work and the latest advances in this field, providing a space to discover new lines of research and exchange knowledge with the scientific community.
The two papers presented were accepted as part of the conference’s Indexing Challenge, to which they were submitted last term:
- Eneko Sabaté – NMF-IVF: Learning Latent Semantic Topics for Approximate Retrieval. The work proposes an index to accelerate the search for sparse vectors, based on non-negative matrix factorisation (NMF). The aim is to identify latent variables that allow grouping the different vectors and reducing the search space, enabling more efficient approximate retrieval.
- Roger Canal – Inverted Block Indexing for Learned Sparse MIPS. This research presents a variation of SEISMIC based on an inverted index organised by blocks. The proposal allows for parameterised proximity searches with the aim of reducing the number of comparisons required.
Eneko and Roger spent the whole week in Brno, where they were able to present their research, attend the various conference sessions and learn first-hand about the latest research directions in the field of similarity search.
This experience has enabled them to share the results of their work with the international scientific community and to learn about new perspectives and advances in a constantly evolving field of research.
