A study of smoothing methods for language models applied to Ad Hoc information retrieval

By Chengxiang Zhai, John D. Lafferty. 2001

1,367 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems, Information Retrieval & Search

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Information Systems

In Information Retrieval & Search

Citations in the corpus

Representation Learning Using Multi-Task Deep Neural Networks for Semantic Classification and Information Retrieval (2015, neutral): “Our model satisﬁes both with Models NDCG@1 NDCG@3 NDCG@10 TF-IDF model (BM25) 0.305 0.328 0.388 Unigram Language Model (Zhai and Lafferty, 2001) 0.304 0.327 0.385 PLSA(Topic=100) (Hofmann, 1999; Gao et al., 2011) 0.305 0.335 0.402 PLSA(Topic=500) (Hofmann, 1999; Gao et al., 2011) 0.308 0.337 0.402…”
A Deep Relevance Matching Model for Ad-hoc Retrieval (2016, background): “[31] is one of the best performing language models.”
Task-Oriented Query Reformulation with Reinforcement Learning (2017, methodology): “We use a Dirichlet smoothed language model [21] to compute a language model from a document d ∈ D0:”
Multi-step Entity-centric Information Retrieval for Multi-Hop Question Answering (2019, positive): “Following prior work (Nogueira and Cho, 2017), we use query likelihood retrieval model with Dirichlet prior smoothing (Zhai and Lafferty, 2001) for first retrieval run.”
An Information Retrieval Approach to Short Text Conversation (2014, background): “…any one of their words t to a di ﬀ erent but semantically related query word w with a non-zero probability. α is the Jelinek-Mercer smoothing factor (Zhai and La ﬀ erty, 2001). β is the interpolation parameter between the post model and the response model. γ is the parameter to balance between…”
Semantic Matching by Non-Linear Word Transportation for Information Retrieval (2016, methodology): “[32] is one of the best performing language models.”
Retrieval models for question and answer archives (2008, methodology): “Typically, unigram language models with the maximum likelihood estimator are used to estimate document language models that are smoothed by background collections with the Dirichlet smoothing technique [16].”
Estimation of statistical translation models based on mutual information for ad hoc information retrieval (2010, methodology): “The methods used for experiments in the following sections are: BL (baseline), i.e., either Dirichlet prior smoothing or JM smoothing [42], TM-MI (translation language model with mutual information for word-to-word translation probabilities), TM-SYN (translation language model with synthetic…”
Character N-Gram Tokenization for European Language Text Retrieval (2004, methodology): “For a treatment of smoothing techniques for the language modeling approach to retrieval, see Zhai and Lafferty (2001).”
Statistical Language Models for Information Retrieval: A Critical Review (2008, background): “Foundations and Trends R© in Information Retrieval, 2008, Volume 2, 4 issues.”
Regularized estimation of mixture models for robust pseudo-relevance feedback (2006, methodology): “The background language model θB is estimated using the whole collection of documents C [19], i.”
Find-similar: similarity browsing as a search tool (2006, background): “Dirichlet prior smoothing [42]: P (w|MD) = D(w)+mP (w|C) |D|+m , where P (w|C) is the MLE model of the collection, and m is the Dirichlet prior smoothing parameter.”
Generalizing Translation Models in the Probabilistic Relevance Framework (2016, methodology): “However, this not being the focus of this study, we use Dirichlet smoothing [38], as many others have done, successfully, before us [12,34,41].”
Supervised Text-based Geolocation Using Language Models on an Adaptive Grid (2012, methodology): “Following previous work in document geolocation, particularly Serdyukov et al. (2009) (hence-forth SMvZ) and Wing and Baldridge (2011) (henceforth W&B), we geolocate texts using a language modeling approach to information retrieval (Ponte and Croft, 1998; Zhai and Lafferty, 2001).”
LETOR: Benchmark Dataset for Research on Learning to Rank for Information Retrieval (2007, positive): “High-level features include the outputs of BM25 [29] and LMIR [35] algorithms.”
Query Expansion Using Word Embeddings (2016, background): “Documents are ranked by the cross entropy between a query model and their Dirichlet smoothed language models [19].”
Overview of Touché 2022: Argument Retrieval (2022, neutral): “Pearl [85] also proposed a two-stage retrieval pipeline using DirichletLM [86] and DPH [87] models to retrieve argumentative sentences.”
Diffusion Kernels on Statistical Manifolds (2005, background): “We remark that the perspective of associating multinomial models with individual documents has recently been explored in information retrieval, with promising results (Ponte and Croft, 1998, Zhai and Laﬀerty, 2001).”