GatfordCentre for Interactive Systems ResearchDepartment of Information

1996

Subfields: Human-Computer Interaction

Citations in the corpus

• Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering (2020, methodology): “On NaturalQuestions, the closed book T5 model obtains 36.6% accuracy with 11B parameters, while our approach obtains 44.1% with 770M parameters plus Wikipedia with BM25 retrieval.”
• Few-shot Learning with Retrieval Augmented Language Models (2022, methodology): “Historically, this information retrieval step was implemented using term-matching methods, such as TF-IDF or BM25 (Jones, 1972; Robertson et al., 1995).”
• Toolformer: Language Models Can Teach Themselves to Use Tools (2023, methodology): “As our search engine, we use a BM25 retriever (Robertson et al., 1995; Baeza-Yates et al., 1999) that indexes the Wikipedia dump from KILT (Petroni et al., 2021).”
• Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategies (2021, methodology): “• With context : We use BM25 (Robertson et al., 1995) to retrieve context from our corpus, while removing stop words from all queries.”
• ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT (2020, methodology): “To tackle this problem, recent work has started exploring using Natural Language Understanding (NLU) techniques to augment traditional retrieval models like BM25 [32].”
• Improving Wikipedia verifiability with AI (2022, background): “For example, verification requires the ability to detect logical entailment in natural language and to convert claims and their context to the best search term for finding evidence—two long-standing problems that have been primarily investigated in somewhat synthetic settings (Bowman et al., 2015;…”
• Unsupervised Dense Information Retrieval with Contrastive Learning (2021, background): “Traditionally, retrieval systems, or retrievers, leverage lexical similarities to match queries and documents, using, for instance, TF-IDF or BM25 weighting (Robertson & Zaragoza, 2009).”
• A Memory Efficient Baseline for Open Domain Question Answering (2020, methodology): “Traditional information retrieval is based on sparse vector representations, with hand crafted term weighting (Jones, 1972; Robertson et al., 1995).”
• PEER: A Collaborative Language Model (2022, methodology): “To make the task of citing the correct document challenging, we obtain the other two documents in Dt by applying the BM25 (Robertson et al., 1995) and DPR (Karpukhin et al., 2020) variants of Petroni et al. (2022) to find the best match in Sphere (Piktus et al., 2021), respectively.”
• Document Ranking with a Pretrained Sequence-to-Sequence Model (2020, methodology): “of a keyword search based on a “classic” IR scoring function such as BM25 (Robertson et al., 1994).”
• Relevance-guided Supervision for OpenQA with ColBERT (2020, methodology): “Mainstream approaches use sparse retrieval models (e.g., BM25; Robertson and Zaragoza 2009) and various heuristics (e.g., traversal of Wikipedia hyperlinks) to retrieve a set of k passages that are relevant to the question q .”
• Variations of the Similarity Function of TextRank for Automated Summarization (2016, methodology): “BM25 is a variation of the TF-IDF model using a probabilistic model [22].”
• Effective site finding using link anchor information (2001, methodology): “TREC systems (including at least one based on the Okapi BM25 formula [10], used in the present study) were found to be highly effective in answering queries of this type, even in comparison to commercial search engines [6].”
• Retrieving Textual Evidence for Knowledge Graph Facts (2019, methodology): “For exact matching we use the OKAPI BM25 [22] retrieval model.”
• Semantic Matching in Search (2014, methodology): “The relevance of the document with respect to the query is represented as the matching degree between the two, calculated using an IR model (matching model) such as BM25 or language models for information retrieval (LM4IR).”
• Ranking of text documents using TF-IDF weighting and association rules mining (2018, background): “• Rank-BM25: Ranking documents based on vector space model (bm25 measure) [23].”
• ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction (2021, methodology): “For each query, a positive d+ is human-annotated, and each negative d− is sampled from unannotated BM25-retrieved passages.”
• Relevance-Based Language Models (2001, methodology): “OKAPI at TREC-3.”