Fast Disjunctive Candidate Generation Using Live Block Filtering

Mallia, Antonio; Siedlaczek, Michał; Suel, Torsten

doi:10.1145/3437963.3441813

Cited by 6 publications

(2 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Efficiency depends on which type of system it is focused on (multi-thread or mono-thread per query retrieval) and also depends heavily on the machine used to perform the measures [15]. For example, efficient sparse retrieval is a domain in itself, with a rich diversity of methods [4,8,9,22,24,30,32,46] that have been proposed in order to improve their retrieval times. Note that the "best method" depends on various factors [22,27] and there is no one method that is better than all the others.…”

Section: Previous Workmentioning

confidence: 99%

“…For instance, a standard dense bi-encoder may rely on multiple CPUs to perform the search while some systems rely only on a single core and others on multi-core implementations. An advantage of sparse retrieval models is the vast literature [4,8,9,22,24,30,32,46] in optimizing retrieval with inverted indices. Furthermore, these works achieve impressive mono-cpu retrieval numbers for traditional sparse retrieval models [26], making it simple to improve the scalability of the system (one just needs to add more cpus).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Efficiency Study for SPLADE Models

Lassance¹,

Clinchant²

2022

Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval

View full text Add to dashboard Cite

Latency and efficiency issues are often overlooked when evaluating IR models based on Pretrained Language Models (PLMs) in reason of multiple hardware and software testing scenarios. Nevertheless, efficiency is an important part of such systems and should not be overlooked.In this paper, we focus on improving the efficiency of the SPLADE model since it has achieved state-of-the-art zero-shot performance and competitive results on TREC collections. SPLADE efficiency can be controlled via a regularization factor, but solely controlling this regularization has been shown to not be efficient enough. In order to reduce the latency gap between SPLADE and traditional retrieval systems, we propose several techniques including L1 regularization for queries, a separation of document/query encoders, a FLOPSregularized middle-training, and the use of faster query encoders. Our benchmark demonstrates that we can drastically improve the efficiency of these models while increasing the performance metrics on in-domain data. To our knowledge, we propose the first neural models that, under the same computing constraints, achieve similar latency (less than 4ms difference) as traditional BM25, while having similar performance (less than 10% MRR@10 reduction) as the stateof-the-art single-stage neural rankers on in-domain data.

show abstract