Similarity Search and Applications 15th International Conference, SISAP 2022, Bologna, Italy, October 5–7, 2022, Proceedings

This book constitutes the refereed proceedings of the 15th International Conference on Similarity Search and Applications, SISAP 2022, held in Bologna, Italy in October 2022.

SISAP 2022 is an annual international conference for researchers focusing on similarity search challenges and related theoretical/practical problems, as well as the design of content-based similarity search applications. The 15 full papers presented together with 8 short and 2 doctoral symposium papers were carefully reviewed and selected from 34 submissions. They were organized in topical sections as follows: Applications; Foundations; Indexing and Clustering; Learning; Doctoral Symposium.

Applications.- Numerical Data Imputation: Choose KNN over Deep Learning.- COSINER: COntext SImilarity data augmentation for Named Entity Recognition.- An Application of Learned Multi-Modal Product Similarity to e-Commerce.- Deep Vision-Language Model for Efficient Multi-modal Similarity Search in Fashion Retrieval.- Stable Anchors for Matching Unlabelled Point Clouds.- Visual Exploration of Human Motion Data.- Foundations.- On Projections to Linear Subspaces.- Concept of Relational Similarity Search.- On the Expected Exclusion Power of Binary Partitions for Metric Search.- Similarity Search with the Distance Density Model.- Generalized Relative Neighborhood Graph (GRNG) for Similarity Search.- A Ptolemaic Partitioning Mechanism.- HubHSP graph: effective data sampling for pivot-based representation strategies.- Indexing and Clustering.- Compacted search tree for graph edit distance computation.- Clustering by Direct Optimization of the Medoid Silhouette.- Automatic Indexing for Similarity Search in ELKI.- Approximate Nearest Neighbor Search on Standard Search Engines.- Evaluation of LID-Aware Graph Embedding Methods for Node Clustering.- Similarity-based Unsupervised Evaluation of Outlier Detection.- Learning.- FastHebb: Scaling Hebbian Training of Deep Neural Networks to ImageNet Level.- Causal Disentanglement with Network Information for Debiased Recommendations.- Causal Disentanglement with Network Information for Debiased Recommendations.- Self-supervised Information Retrieval Trained from Self-generated Sets of Queries and Relevant Documents.- Doctoral Symposium.- Discovering Knowledge Graphs Via Attention-Driven Graph Generation.- Visual Recommendation and Visual Search for Fashion e-commerce.