Efficient Document-at-a-Time and Score-at-a-Time Query Evaluation for Learned Sparse Representations

Abstract

Researchers have had much recent success with ranking models based on so-called learned sparse representations generated by transformers. One crucial advantage of this approach is that such models can exploit inverted indexes for top- k retrieval, thereby leveraging decades of work on efficient query evaluation. Yet, there remain many open questions about how these learned representations fit within the existing literature, which our work aims to tackle using four representative learned sparse models. We find that impact weights generated by transformers appear to greatly reduce opportunities for skipping and early exiting optimizations in well-studied document-at-a-time ( DaaT ) approaches. Similarly, “off-the-shelf” application of score-at-a-time ( SaaT ) processing exhibits a mismatch between these weights and assumptions behind accumulator management strategies. Building on these observations, we present solutions to address deficiencies with both DaaT and SaaT approaches, yielding substantial speedups in query evaluation. Our detailed empirical analysis demonstrates that both methods lie on the effectiveness–efficiency Pareto frontier, indicating that the optimal choice for deployment depends on operational constraints.

A Unified Framework for Learned Sparse Retrieval

“…LSR is compatible with many techniques from sparse retrieval, such as inverted indexing and accompanying query processing algorithms. However, differences in LSR weights can mean that existing query processing optimizations become much less helpful, motivating new optimizations [21,22,24].…”

Pemrosesan Query dan Pemeringkatan Hasil dalam Information Retrieval: Sebuah Kajian Literatur

“…Deep-QPP menggunakan arsitektur yang terdiri dari beberapa lapisan konvolusi 2D diikuti lapisan parameter feed-forward. [23] Penelitian ini [24] menghasilkan peningkatan substansial dalam evaluasi kueri untuk mengatasi kekurangan pada pendekatan Document-at-a-time (DAAT) dan Score-at-a-time (SAAT). Dengan analisis empiris terperinci menunjukkan bahwa DAAT dan SAAT memiliki permasalahan pada pembatasan nilai pareto efektivitas dan efisiensi.…”