arXiv subjects

Matthias Petri

Publications and source records attributed to Matthias Petri.

8 recordsLinked to original sources

arXiv2024

Improved Learned Sparse Retrieval with Corpus-Specific Vocabularies

We explore leveraging corpus-specific vocabularies that improve both efficiency and effectiveness of learned sparse retrieval systems. We find that pre-training the underlying BERT model on the target corpus, specifically targeting different vocabulary sizes incorporated into the document expansion process, improves retrieval quality by up to 12% while in some scenarios decreasing latency by up to 50%. Our experiments show that adopting corpus-specific vocabulary and increasing vocabulary size decreases average postings list length which in turn reduces latency. Ablation studies show interesting interactions between custom vocabularies, document expansion techniques, and sparsification objectives of sparse models. Both effectiveness and efficiency improvements transfer to different retrieval approaches such as uniCOIL and SPLADE and offer a simple yet effective approach to providing new efficiency-effectiveness trade-offs for learned sparse retrieval systems.

cs.IR↗

arXiv2022

A Sensitivity Analysis of the MSMARCO Passage Collection

The recent MSMARCO passage retrieval collection has allowed researchers to develop highly tuned retrieval systems. One aspect of this data set that makes it distinctive compared to traditional corpora is that most of the topics only have a single answer passage marked relevant. Here we carry out a "what if" sensitivity study, asking whether a set of systems would still have the same relative performance if more passages per topic were deemed to be "relevant", exploring several mechanisms for identifying sets of passages to be so categorized. Our results show that, in general, while run scores can vary markedly if additional plausible passages are presumed to be relevant, the derived system ordering is relatively insensitive to additional relevance, providing support for the methodology that was used at the time the MSMARCO passage collection was created.

cs.IR↗

arXiv2021

Anytime Ranking on Document-Ordered Indexes

Inverted indexes continue to be a mainstay of text search engines, allowing efficient querying of large document collections. While there are a number of possible organizations, document-ordered indexes are the most common, since they are amenable to various query types, support index updates, and allow for efficient dynamic pruning operations. One disadvantage with document-ordered indexes is that high-scoring documents can be distributed across the document identifier space, meaning that index traversal algorithms that terminate early might put search effectiveness at risk. The alternative is impact-ordered indexes, which primarily support top-k disjunctions, but also allow for anytime query processing, where the search can be terminated at any time, with search quality improving as processing latency increases. Anytime query processing can be used to effectively reduce high-percentile tail latency which is essential for operational scenarios in which a service level agreement (SLA) imposes response time requirements. In this work, we show how document-ordered indexes can be organized such that they can be queried in an anytime fashion, enabling strict latency control with effective early termination. Our experiments show that processing document-ordered topical segments selected by a simple score estimator outperforms existing anytime algorithms, and allows query runtimes to be accurately limited in order to comply with SLA requirements.

cs.IR↗

arXiv2016

Fast, Small and Exact: Infinite-order Language Modelling with Compressed Suffix Trees

Efficient methods for storing and querying are critical for scaling high-order n-gram language models to large corpora. We propose a language model based on compressed suffix trees, a representation that is highly compact and can be easily held in memory, while supporting queries needed in computing language model probabilities on-the-fly. We present several optimisations which improve query runtimes up to 2500x, despite only incurring a modest increase in construction time and memory usage. For large corpora and high Markov orders, our method is highly competitive with the state-of-the-art KenLM package. It imposes much lower memory requirements, often by orders of magnitude, and has runtimes that are either similar (for training) or comparable (for querying).

cs.CL↗

arXiv2016

CSA++: Fast Pattern Search for Large Alphabets

Indexed pattern search in text has been studied for many decades. For small alphabets, the FM-Index provides unmatched performance, in terms of both space required and search speed. For large alphabets -- for example, when the tokens are words -- the situation is more complex, and FM-Index representations are compact, but potentially slow. In this paper we apply recent innovations from the field of inverted indexing and document retrieval to compressed pattern search, including for alphabets into the millions. Commencing with the practical compressed suffix array structure developed by Sadakane, we show that the Elias-Fano code-based approach to document indexing can be adapted to provide new tradeoff options in indexed pattern search, and offers significantly faster pattern processing compared to previous implementations, as well as reduced space requirements. We report a detailed experimental evaluation that demonstrates the relative advantages of the new approach, using the standard Pizza&Chili methodology and files, as well as applied use-cases derived from large-scale data compression, and from natural language processing. For large alphabets, the new structure gives rise to space requirements that are close to those of the most highly-compressed FM-Index variants, in conjunction with unparalleled search throughput rates.

cs.DS↗

arXiv2016

Access Time Tradeoffs in Archive Compression

Web archives, query and proxy logs, and so on, can all be very large and highly repetitive; and are accessed only sporadically and partially, rather than continually and holistically. This type of data is ideal for compression-based archiving, provided that random-access to small fragments of the original data can be achieved without needing to decompress everything. The recent RLZ (relative Lempel Ziv) compression approach uses a semi-static model extracted from the text to be compressed, together with a greedy factorization of the whole text encoded using static integer codes. Here we demonstrate more precisely than before the scenarios in which RLZ excels. We contrast RLZ with alternatives based on block-based adaptive methods, including approaches that "prime" the encoding for each block, and measure a range of implementation options using both hard-disk (HDD) and solid-state disk (SSD) drives. For HDD, the dominant factor affecting access speed is the compression rate achieved, even when this involves larger dictionaries and larger blocks. When the data is on SSD the same effects are present, but not as markedly, and more complex trade-offs apply.

cs.IT↗

arXiv2014

Compact Indexes for Flexible Top-k Retrieval

We engineer a self-index based retrieval system capable of rank-safe evaluation of top-k queries. The framework generalizes the GREEDY approach of Culpepper et al. (ESA 2010) to handle multi-term queries, including over phrases. We propose two techniques which significantly reduce the ranking time for a wide range of popular Information Retrieval (IR) relevance measures, such as TFxIDF and BM25. First, we reorder elements in the document array according to document weight. Second, we introduce the repetition array, which generalizes Sadakane's (JDA 2007) document frequency structure to document subsets. Combining document and repetition array, we achieve attractive functionality-space trade-offs. We provide an extensive evaluation of our system on terabyte-sized IR collections.

cs.DS↗

arXiv2013

From Theory to Practice: Plug and Play with Succinct Data Structures

Engineering efficient implementations of compact and succinct structures is a time-consuming and challenging task, since there is no standard library of easy-to- use, highly optimized, and composable components. One consequence is that measuring the practical impact of new theoretical proposals is a difficult task, since older base- line implementations may not rely on the same basic components, and reimplementing from scratch can be very time-consuming. In this paper we present a framework for experimentation with succinct data structures, providing a large set of configurable components, together with tests, benchmarks, and tools to analyze resource requirements. We demonstrate the functionality of the framework by recomposing succinct solutions for document retrieval.

cs.DS↗