### Citations

4014 |
Introduction to Modern Information Retrieval
- Salton, McGill
- 1983
(Show Context)
Citation Context ...such a correlation to generate a relevance status value (RSV) for each document and rank documents accordingly. The vector space model is the most well-known model of this type (Salton et al., 1975a; =-=Salton and McGill, 1983-=-; Salton, 1989). In the vector space model, a document and a query are represented as two term vectors in a high-dimensional term space. Each term is assigned a weight that reflects its “importance” t... |

3770 | Indexing by Latent Semantic Analysis
- Dumais, Furnas, et al.
- 1990
(Show Context)
Citation Context ...tent semantic indexing can be applied to reduce the dimension of the term space and to capture the semantic “closeness” among terms, and thus to improve the representation of the documents and query (=-=Deerwester et al., 1990-=-). A document can also be represented by a multinomial distribution over the terms, as in the distribution model of indexing proposed in (Wong and Yao, 1989). In the vector space model, feedback is ty... |

2186 | Term-weighting approaches in automatic text retrieval
- Salton, Buckley
- 1988
(Show Context)
Citation Context ...997; Zhai, 1997). Many heuristics have also been proposed to improve term weighting, but again, no weighting method has been found to be significantly better than the heuristic TF-IDF term weighting (=-=Salton and Buckley, 1988-=-). To address the variances in the length of documents, an effective weighting formula also needs to incorporate document length heuristically (Singhal et al., 1996). Salton et al. introduced the idea... |

1509 |
Automatic Text Processing - The Transformation, Analysis, and Retrieval of Information by Computer
- Salton
- 1989
(Show Context)
Citation Context ...erate a relevance status value (RSV) for each document and rank documents accordingly. The vector space model is the most well-known model of this type (Salton et al., 1975a; Salton and McGill, 1983; =-=Salton, 1989-=-). In the vector space model, a document and a query are represented as two term vectors in a high-dimensional term space. Each term is assigned a weight that reflects its “importance” to the document... |

1373 |
A vector space model for automatic indexing
- Salton, Wong, et al.
- 1975
(Show Context)
Citation Context ...easure that preserves such a correlation to generate a relevance status value (RSV) for each document and rank documents accordingly. The vector space model is the most well-known model of this type (=-=Salton et al., 1975-=-a; Salton and McGill, 1983; Salton, 1989). In the vector space model, a document and a query are represented as two term vectors in a high-dimensional term space. Each term is assigned a weight that r... |

1153 | A language modeling approach to information retrieval
- Ponte, Croft
(Show Context)
Citation Context ...e model. This work provides a relevance-based justification for this new family of probabilistic models based on statistical language modeling. The language modeling approach was first introduced in (=-=Ponte and Croft, 1998-=-) and later explored in (Hiemstra and Kraaij, 1998; Miller et al., 1999; Berger and Lafferty, 1999; Song and Croft, 1999), among others. The estimation of a language model based on a document (i.e., t... |

1092 |
Relevance feedback in information Retrieval
- Rocchio
- 1971
(Show Context)
Citation Context ...chio method, which simply adds the centroid vector of the relevant documents to the query vector and subtracts from it the centroid vector of the non-relevant documents with appropriate coefficients (=-=Rocchio, 1971-=-). In effect, this leads to an expansion of the original query vector, i.e., additional terms are extracted from the known relevant (and non-relevant) documents, and are added to the original query ve... |

961 | A study of smoothing methods for language models applied to information retrieval. - Zhai, Lafferty - 2004 |

755 | Relevance weighting of search terms - Robertson, Jones - 1976 |

755 | Improving retrieval performance by relevance feedback
- Salton, Buckley
- 1990
(Show Context)
Citation Context ...expansion of the original query vector, i.e., additional terms are extracted from the known relevant (and non-relevant) documents, and are added to the original query vector with appropriate weights (=-=Salton and Buckley, 1990-=-). The extended Boolean (p-norm) model is a heuristic extension of the traditional Boolean model to perform document ranking, but it can also be regarded as a special case of the similarity model (Fox... |

601 | Okapi at TREC-3
- Robertson
- 1995
(Show Context)
Citation Context ...rmance. Indeed, a simple approximation of the 2-Poisson probabilistic model, which has led to the BM25 retrieval formula used in the Okapi system, has been very effective (Robertson and Walker, 1994; =-=Robertson et al., 1995-=-). The language modeling approaches have also been shown to perform very well (Ponte and Croft, 1998; Hiemstra and Kraaij, 1998; Miller et al., 1999). The BM25 formula is shown below, following the no... |

499 | Naive (Bayes) at Forty: The Independence Assumption in Information Retrieval
- Lewis
- 1998
(Show Context)
Citation Context ...ssical probabilistic model. The BIR model assumes that terms are independently distributed in each of the two relevance models, so it essentially uses the Naïve Bayes classifier for document ranking (=-=Lewis, 1998-=-). 1 The BIR retrieval formula is the following (Robertson and Sparck Jones, 1976; Lafferty and Zhai, 2003): log p(r | D, Q) rank = p(r | D, Q) � p(t | Q, r)(1 − p(t | Q, r)) log (1 − p(t | Q, r))p(t ... |

477 | Pivoted Document Length Normalization
- Singhal, Buckley, et al.
- 1996
(Show Context)
Citation Context ...euristic TF-IDF term weighting (Salton and Buckley, 1988). To address the variances in the length of documents, an effective weighting formula also needs to incorporate document length heuristically (=-=Singhal et al., 1996-=-). Salton et al. introduced the idea of the discrimination value of an indexing term (Salton et al., 1975b). The discrimination value of an indexing term is the increase or the decrease in the mean in... |

459 | Some simple effective approximations to the 2-Poisson model for probabilistic weighted retrieval
- Robertson, Walker
- 1994
(Show Context)
Citation Context ...Robertson et al., 1981). This model has not shown empirical improvement in retrieval performance directly, but an approximation of the model using a simple TF formula turns out to be quite effective (=-=Robertson and Walker, 1994-=-). The heuristic retrieval formula BM25 has been successfully used in City University’s Okapi system and several other TREC systems (Voorhees and Harman, 2001). A different way of introducing the term... |

440 | Relevance-based language models.
- Lavrenko, Croft
- 2001
(Show Context)
Citation Context ...oft and Harper, 1979; Robertson and Walker, 1997). Recently, Lavrenko and Croft made progress in estimating the relevance model without relevance judgments by exploiting language modeling techniques (=-=Lavrenko and Croft, 2001-=-). When explicit relevance judgments are available, the classic models, being based on document generation, have the advantage of being able to improve the estimation of the component probabilistic mo... |

387 | The INQUERY retrieval system
- Callan, Croft, et al.
- 1992
(Show Context)
Citation Context ...ys little about how one can further decompose the general probabilistic model. As a result, operationally, one usually has to set probabilities based on heuristics, as was done in the Inquery system (=-=Callan et al., 1992-=-). Kwok’s network model may also be considered as performing a probabilistic inference (Kwok, 1995), though it is based on spread activation. In general, the probabilistic inference models address the... |

382 | Document language models, query models and risk minimization for information retrieval
- Lafferty, Zhai
- 2001
(Show Context)
Citation Context ..., but rather, relies on a given similarity measure. Second, it is only helpful for selecting indexing terms, but not for the weighting of terms. A new generation of similarity-based retrieval models (=-=Lafferty and Zhai, 2001-=-a; Zhai and Lafferty, 2001a; Lavrenko, 2004) has been proposed based on the idea of representing documents and queries with statistical language models. In one of the most effective methods, both a do... |

360 | A probabilistic model of information retrieval: development and comparative experiments
- Spärck-Jones, Walker, et al.
- 2000
(Show Context)
Citation Context ...he next section. 3 Probabilistic Relevance Models In a probabilistic relevance model, we are interested in the question “What is the probability that this document is relevant to this query?” (Sparck =-=Jones et al., 2000-=-). Given a query, a document is assumed to be 3seither relevant or non-relevant, but a system can never be sure about the true relevance status of a document, so it has to rely on a probabilistic rele... |

314 | Information retrieval as statistical translation
- Berger, Lafferty
- 1999
(Show Context)
Citation Context ...ic models based on statistical language modeling. The language modeling approach was first introduced in (Ponte and Croft, 1998) and later explored in (Hiemstra and Kraaij, 1998; Miller et al., 1999; =-=Berger and Lafferty, 1999-=-; Song and Croft, 1999), among others. The estimation of a language model based on a document (i.e., the estimation of p(. | D, r)) is the key component in the language modeling approach. Indeed, most... |

263 | Evaluation of an inference network-based retrieval model
- Turtle, Croft
- 1991
(Show Context)
Citation Context ...h representations is in a way outside the model, so there is little guidance on how to choose or how to improve a representation. The inference network model is also based on probabilistic inference (=-=Turtle and Croft, 1991-=-). It is essentially a Bayesian belief network that models the dependency between the satisfaction of a query and the observation of documents. The estimation of relevance is based on the computation ... |

258 |
Extended boolean information retrieval
- Salton, Fox, et al.
- 1983
(Show Context)
Citation Context ...xtended Boolean (p-norm) model is a heuristic extension of the traditional Boolean model to perform document ranking, but it can also be regarded as a special case of the similarity model (Fox, 1983; =-=Salton et al., 1983-=-). The similarity function has a parameter p that controls the “strictness” of satisfying the constraint of a Boolean query, in such a way that it approaches a strict (conjunctive or disjunctive) Bool... |

243 | A general language model for information retrieval
- Song, Croft
- 1999
(Show Context)
Citation Context ...cal language modeling. The language modeling approach was first introduced in (Ponte and Croft, 1998) and later explored in (Hiemstra and Kraaij, 1998; Miller et al., 1999; Berger and Lafferty, 1999; =-=Song and Croft, 1999-=-), among others. The estimation of a language model based on a document (i.e., the estimation of p(. | D, r)) is the key component in the language modeling approach. Indeed, most work in this directio... |

237 |
On relevance, probabilistic indexing, and information retrieval
- Maron, Kuhns
- 1960
(Show Context)
Citation Context ...er who likes document D would use Q as a query. The second component p(r | D) is a prior that can be used to encode a user’s bias on documents. Models based on query generation have been explored in (=-=Maron and Kuhns, 1960-=-), (Fuhr, 1992) and (Lafferty and Zhai, 2001b). The probabilistic indexing model proposed in (Maron and Kuhns, 1960) is the first probabilistic retrieval model, in which the indexing terms assigned to... |

236 | Modern information retrieval: a brief overview.
- Singhal
- 2001
(Show Context)
Citation Context ...m i. Similarly, a query q can be represented by a query vector �q = (y1, y2, ..., yn). The weight is usually computed based on the so-called TF-IDF weighting, which is a combination of three factors (=-=Singhal, 2001-=-): (1) the local frequency of the term (in a document or query); (2) the global frequency of the term in the whole collection; (3) document length. With the cosine measure, we have the following simil... |

225 |
A hidden Markov model information retrieval system.
- Miller, Leek, et al.
- 1999
(Show Context)
Citation Context ...family of probabilistic models based on statistical language modeling. The language modeling approach was first introduced in (Ponte and Croft, 1998) and later explored in (Hiemstra and Kraaij, 1998; =-=Miller et al., 1999-=-; Berger and Lafferty, 1999; Song and Croft, 1999), among others. The estimation of a language model based on a document (i.e., the estimation of p(. | D, r)) is the key component in the language mode... |

196 | A non-classical logic for information retrieval - Rijsbergen - 1986 |

184 |
Using probabilistic models of document retrieval without relevance information
- Croft, Harper
- 1979
(Show Context)
Citation Context ...n no explicit relevance information is available. Typically, p(t | Q, r) is set to a constant and p(t | Q, r) is estimated under the assumption that the whole collection of documents is non-relevant (=-=Croft and Harper, 1979-=-; Robertson and Walker, 1997). Recently, Lavrenko and Croft made progress in estimating the relevance model without relevance judgments by exploiting language modeling techniques (Lavrenko and Croft, ... |

179 |
Representation and Learning in Information Retrieval
- Lewis
- 1992
(Show Context)
Citation Context ...s and the weighting of the indexing terms. The choice of different indexing units has been extensively studied, but no significant improvement has been achieved over the simplest word-based indexing (=-=Lewis, 1992-=-), though some more recent evaluation has shown more promising improvement on average through using linguistic phrases (Evans and Zhai, 1996; Strzalkowski, 1997; Zhai, 1997). Many heuristics have also... |

169 | Cluster-based retrieval using language models
- Liu, Croft
- 2004
(Show Context)
Citation Context ...improved through improving the estimation of both the query language model (Zhai and Lafferty, 2001a; Lavrenko and Croft, 2001; Shen et al., 2005; Tao and Zhai, 2006) and the document language model (=-=Liu and Croft, 2004-=-; Kurland and Lee, 2004; Tao et al., ). Instead of imposing a strict document generation or query generation decomposition of p(D, Q | R), one can also “generate” a document-query pair simultaneously.... |

150 | Context-sensitive information retrieval using implicit feedback
- SHEN, TAN, et al.
- 2005
(Show Context)
Citation Context ...l (Lavrenko and Croft, 2001)). In general, retrieval performance can be improved through improving the estimation of both the query language model (Zhai and Lafferty, 2001a; Lavrenko and Croft, 2001; =-=Shen et al., 2005-=-; Tao and Zhai, 2006) and the document language model (Liu and Croft, 2004; Kurland and Lee, 2004; Tao et al., ). Instead of imposing a strict document generation or query generation decomposition of ... |

120 | Probabilistic Models in Information Retrieval”,
- Fuhr
- 1992
(Show Context)
Citation Context ...relevant documents. Documents are then ranked according to the posterior probability of relevance. Most classic probabilistic retrieval models (Robertson and Sparck Jones, 1976; van Rijsbergen, 1979; =-=Fuhr, 1992-=-) are based on document generation. The Binary Independence Retrieval (BIR) model (Robertson 4 .sand Sparck Jones, 1976; Fuhr, 1992) is perhaps the most well known classical probabilistic model. The B... |

118 |
A probabilistic approach to automatic keyword indexing: Part i. on the distribution of specialty words in a technical literature.
- Harter
- 1975
(Show Context)
Citation Context ...the vector space model makes it easy to incorporate different indexing models. For example, the 2-Poisson probabilistic indexing model can be used to select indexing terms and/or assign term weights (=-=Harter, 1975-=-; Bookstein and Swanson, 1975). Latent semantic indexing can be applied to reduce the dimension of the term space and to capture the semantic “closeness” among terms, and thus to improve the represent... |

118 |
C.T.: A theory of term importance in automatic text analysis.
- Salton, Yang, et al.
- 1975
(Show Context)
Citation Context ...easure that preserves such a correlation to generate a relevance status value (RSV) for each document and rank documents accordingly. The vector space model is the most well-known model of this type (=-=Salton et al., 1975-=-a; Salton and McGill, 1983; Salton, 1989). In the vector space model, a document and a query are represented as two term vectors in a high-dimensional term space. Each term is assigned a weight that r... |

115 | Twenty-one at trec-7: Ad-hoc and cross-language track. In:
- Hiemstra, Kraaij
- 1999
(Show Context)
Citation Context ...justification for this new family of probabilistic models based on statistical language modeling. The language modeling approach was first introduced in (Ponte and Croft, 1998) and later explored in (=-=Hiemstra and Kraaij, 1998-=-; Miller et al., 1999; Berger and Lafferty, 1999; Song and Croft, 1999), among others. The estimation of a language model based on a document (i.e., the estimation of p(. | D, r)) is the key component... |

105 | A formal study of information retrieval heuristics.
- Fang, Tao, et al.
- 2004
(Show Context)
Citation Context ...ically well-founded frameworks can provide a roadmap for applying statistical language models to retrieval. (2) a new axiomatic approach to developing retrieval models has been proposed and explored (=-=Fang et al., 2004-=-; Fang and Zhai, 2005; Fang and Zhai, 2006). This new approach attempts to model relevance more directly with term-level heuristic constraints. As a result, it provides several important benefits, inc... |

105 |
On modeling information retrieval with probabilistic inference
- Wong, Yao
- 1995
(Show Context)
Citation Context ...y in relevance, van Rijsbergen introduced a logic for probabilistic inference, in which the probability 7sof a conditional, such as p → q, can be estimated based on the notion of possible worlds. In (=-=Wong and Yao, 1995-=-), Wong and Yao extended the probabilistic inference model and developed a general probabilistic inference model which subsumes several other retrieval models such as Boolean, vector space, and the cl... |

103 | A probabilistic learning approach for document indexing.
- Fuhr, Buckley
- 1991
(Show Context)
Citation Context ...n model was probably first introduced with some success by Fox (Fox, 1983), where features such as term frequency, authorship, and co-citation were combined using linear regression. Fuhr and Buckley (=-=Fuhr and Buckley, 1991-=-) used polynomial regression to approximate relevance. Gey used logistic regression involving information such as query term frequency, document term frequency, IDF, and relative term frequency in the... |

92 | Noun-phrase analysis in unrestricted text for information retrieval.
- Evans, Zhai
- 1996
(Show Context)
Citation Context ...improvement has been achieved over the simplest word-based indexing (Lewis, 1992), though some more recent evaluation has shown more promising improvement on average through using linguistic phrases (=-=Evans and Zhai, 1996-=-; Strzalkowski, 1997; Zhai, 1997). Many heuristics have also been proposed to improve term weighting, but again, no weighting method has been found to be significantly better than the heuristic TF-IDF... |

88 | Probabilistic Relevance Models Based on Document and Query Generation, chapter 1
- Lafferty, Zhai
- 2002
(Show Context)
Citation Context ...ures. A regression model thus provides only limited guidance for extending a retrieval model. Alternatively, p(R = r | D, Q) can be estimated indirectly using a generative model in the following way (=-=Lafferty and Zhai, 2003-=-): p(R = r | D, Q) = p(D, Q | R = r) p(R = r) p(D, Q) Equivalently, we may use the following log-odds ratio to rank documents: log p(r | D, Q) p(r | D, Q) = log p(D, Q | r) p(r) p(D, Q | r) p(r) . The... |

78 |
On relevance weights with little relevance information.
- Robertson, Walker
- 1997
(Show Context)
Citation Context ...information is available. Typically, p(t | Q, r) is set to a constant and p(t | Q, r) is estimated under the assumption that the whole collection of documents is non-relevant (Croft and Harper, 1979; =-=Robertson and Walker, 1997-=-). Recently, Lavrenko and Croft made progress in estimating the relevance model without relevance judgments by exploiting language modeling techniques (Lavrenko and Croft, 2001). When explicit relevan... |

74 | Corpus structure, language models, and ad hoc information retrieval.
- Kurland, Lee
- 2004
(Show Context)
Citation Context ...oving the estimation of both the query language model (Zhai and Lafferty, 2001a; Lavrenko and Croft, 2001; Shen et al., 2005; Tao and Zhai, 2006) and the document language model (Liu and Croft, 2004; =-=Kurland and Lee, 2004-=-; Tao et al., ). Instead of imposing a strict document generation or query generation decomposition of p(D, Q | R), one can also “generate” a document-query pair simultaneously. Mittendorf & Schauble ... |

68 | Probabilistic models of indexing and searching - Robertson, Rijsbergen, et al. - 1980 |

66 | A generative theory of relevance.
- Lavrenko
- 2004
(Show Context)
Citation Context ...Second, it is only helpful for selecting indexing terms, but not for the weighting of terms. A new generation of similarity-based retrieval models (Lafferty and Zhai, 2001a; Zhai and Lafferty, 2001a; =-=Lavrenko, 2004-=-) has been proposed based on the idea of representing documents and queries with statistical language models. In one of the most effective methods, both a document and a query are represented as a uni... |

62 |
Model-based feedback in the kl-divergence retrieval model.
- Zhai, Lafferty
- 2001
(Show Context)
Citation Context ...given similarity measure. Second, it is only helpful for selecting indexing terms, but not for the weighting of terms. A new generation of similarity-based retrieval models (Lafferty and Zhai, 2001a; =-=Zhai and Lafferty, 2001-=-a; Lavrenko, 2004) has been proposed based on the idea of representing documents and queries with statistical language models. In one of the most effective methods, both a document and a query are rep... |

56 | An evaluation of feedback in document retrieval using co-occurrence data - HARPER, J - 1978 |

56 | Document and passage retrieval based on hidden markov models
- Mittendorf, Schauble
- 1994
(Show Context)
Citation Context ...one can also “generate” a document-query pair simultaneously. Mittendorf & Schauble explored a passagebased generative model using the Hidden Markov Model (HMM), which can be regarded as such a case (=-=Mittendorf and Schauble, 1994-=-). In this work, a document query pair is represented as a sequence of symbols, each corresponding to a term in a particular position of the document. All term tokens are clustered in terms of the sim... |

56 |
Regularized estimation of mixture models for robust pseudorelevance feedback. In:
- Tao, Zhai
- 2006
(Show Context)
Citation Context ...ft, 2001)). In general, retrieval performance can be improved through improving the estimation of both the query language model (Zhai and Lafferty, 2001a; Lavrenko and Croft, 2001; Shen et al., 2005; =-=Tao and Zhai, 2006-=-) and the document language model (Liu and Croft, 2004; Kurland and Lee, 2004; Tao et al., ). Instead of imposing a strict document generation or query generation decomposition of p(D, Q | R), one can... |

55 | Language model information retrieval with document expansion - Tao, Wang, et al. - 2006 |

53 |
Applying bayesian networks to information retrieval.
- Fung, Favero
- 1995
(Show Context)
Citation Context ...sed on the computation of the conditional probability that the query is satisfied given that the document is observed. Other similar uses of the Bayesian belief network in retrieval are presented in (=-=Fung and Favero, 1995-=-; Ribeiro and Muntz, 1996; Ribeiro-Neto et al., 2000). The inference network model is a much more general formalism than most of the models that we have discussed above. With different ways to realize... |

48 |
Extending the Boolean and vector space models of information retrieval with P-norm queries and multiple concept types.
- FOX
- 1983
(Show Context)
Citation Context ...990). The extended Boolean (p-norm) model is a heuristic extension of the traditional Boolean model to perform document ranking, but it can also be regarded as a special case of the similarity model (=-=Fox, 1983-=-; Salton et al., 1983). The similarity function has a parameter p that controls the “strictness” of satisfying the constraint of a Boolean query, in such a way that it approaches a strict (conjunctive... |

47 | Inferring probability of relevance using the method of logistic regression
- Gey
- 1994
(Show Context)
Citation Context ...rmation such as query term frequency, document term frequency, IDF, and relative term frequency in the whole collection, and this model shows promising performance in three small testing collections (=-=Gey, 1994-=-). Regression models provide a principled way of exploring heuristic features and ideas. One important advantage of regression models is their ability to learn from all the past relevance judgments, i... |

46 |
Muntz: A belief network model for IR
- Ribeiro-Neto, R
- 1996
(Show Context)
Citation Context ...of the conditional probability that the query is satisfied given that the document is observed. Other similar uses of the Bayesian belief network in retrieval are presented in (Fung and Favero, 1995; =-=Ribeiro and Muntz, 1996-=-; Ribeiro-Neto et al., 2000). The inference network model is a much more general formalism than most of the models that we have discussed above. With different ways to realize the probabilistic relati... |

43 |
A network approach to probabilistic information retrieval
- Kwok
- 1995
(Show Context)
Citation Context ...one usually has to set probabilities based on heuristics, as was done in the Inquery system (Callan et al., 1992). Kwok’s network model may also be considered as performing a probabilistic inference (=-=Kwok, 1995-=-), though it is based on spread activation. In general, the probabilistic inference models address the issue of relevance in a very general way. In some sense, the lack of a commitment to specific ass... |

42 | An exploration of axiomatic approaches to information retrieval.
- Fang, Zhai
- 2005
(Show Context)
Citation Context ... frameworks can provide a roadmap for applying statistical language models to retrieval. (2) a new axiomatic approach to developing retrieval models has been proposed and explored (Fang et al., 2004; =-=Fang and Zhai, 2005-=-; Fang and Zhai, 2006). This new approach attempts to model relevance more directly with term-level heuristic constraints. As a result, it provides several important benefits, including making it poss... |

40 |
Some inconsistencies and misnomers in probabilistic information retrieval.
- Cooper
- 1991
(Show Context)
Citation Context ...he content of D. The Binary Independence Indexing (BII) model proposed in (Fuhr, 1992) is another 1 The required underlying independence assumption for the final retrieval formula is actually weaker (=-=Cooper, 1991-=-). 5sspecial case of the query generation model. It allows the description of a document (with weighted terms) to be estimated based on arbitrary queries, but the specific parameterization makes it ha... |

40 | Semantic Term Matching in Axiomatic Approaches to Information Retrieval.
- Fang, Zhai
- 2006
(Show Context)
Citation Context ...de a roadmap for applying statistical language models to retrieval. (2) a new axiomatic approach to developing retrieval models has been proposed and explored (Fang et al., 2004; Fang and Zhai, 2005; =-=Fang and Zhai, 2006-=-). This new approach attempts to model relevance more directly with term-level heuristic constraints. As a result, it provides several important benefits, including making it possible to analytically ... |

36 | Fast statistical parsing of noun phrases for document indexing
- Zhai
- 1997
(Show Context)
Citation Context ...plest word-based indexing (Lewis, 1992), though some more recent evaluation has shown more promising improvement on average through using linguistic phrases (Evans and Zhai, 1996; Strzalkowski, 1997; =-=Zhai, 1997-=-). Many heuristics have also been proposed to improve term weighting, but again, no weighting method has been found to be significantly better than the heuristic TF-IDF term weighting (Salton and Buck... |

35 | A new probabilistic model of text classification and retrieval
- Kalt
- 1996
(Show Context)
Citation Context ... of introducing the term frequency into the model, not directly proposed but implied by a lot of work in text categorization, is to regard a document as being generated from a unigram language model (=-=Kalt, 1996-=-; McCallum and Nigam, 1998). With query generation, p(D, Q | R) = p(Q | D, R)p(D | R), so we end up with the following ranking formula: log p(r | D, Q) p(Q | D, r) p(r | D) = log + log p(r | D, Q) p(Q... |

21 |
A probability distribution model for information retrieval
- Wong, Yao
- 1989
(Show Context)
Citation Context ...esentation of the documents and query (Deerwester et al., 1990). A document can also be represented by a multinomial distribution over the terms, as in the distribution model of indexing proposed in (=-=Wong and Yao, 1989-=-). In the vector space model, feedback is typically treated as query vector updating. A well-known approach is the Rocchio method, which simply adds the centroid vector of the relevant documents to th... |

20 |
A decision theoretic foundation for indexing.
- Bookstein, Swanson
- 1975
(Show Context)
Citation Context ...ce model makes it easy to incorporate different indexing models. For example, the 2-Poisson probabilistic indexing model can be used to select indexing terms and/or assign term weights (Harter, 1975; =-=Bookstein and Swanson, 1975-=-). Latent semantic indexing can be applied to reduce the dimension of the term space and to capture the semantic “closeness” among terms, and thus to improve the representation of the documents and qu... |

19 | A belief network model for ir - Ribeiro, Muntz - 1996 |

18 |
Document representation in probabilistic models of information retrieval
- Croft
- 1981
(Show Context)
Citation Context ...parameters is a problem in practice (Harper and van Rijsbergen, 1978). Croft investigated how the heuristic term significance weight can be incorporated into probabilistic models in a principled way (=-=Croft, 1981-=-). Another effort to improve document representation involves introducing the term frequency directly into the model by using a multiple 2-Poisson mixture representation of documents (Robertson et al.... |

14 |
NLP track at TREC-5. In
- Strzalkowski
- 1997
(Show Context)
Citation Context ...chieved over the simplest word-based indexing (Lewis, 1992), though some more recent evaluation has shown more promising improvement on average through using linguistic phrases (Evans and Zhai, 1996; =-=Strzalkowski, 1997-=-; Zhai, 1997). Many heuristics have also been proposed to improve term weighting, but again, no weighting method has been found to be significantly better than the heuristic TF-IDF term weighting (Sal... |

11 | C.J.: A theoretical basis for theuse of co-occurrence data in information retrieval. - Rijbergen - 1977 |

10 |
Language models and uncertain inference in information retrieval
- Fuhr
- 2001
(Show Context)
Citation Context ...space, and the classic probabilistic models. Fuhr shows that some particular form of the language modeling approach can also be derived as a special case of the general probabilistic inference model (=-=Fuhr, 2001-=-). While theoretically interesting, the probabilistic inference models all must rely on further assumptions about the representation of documents and queries in order to obtain an operational retrieva... |

9 |
A comparison of event models for nave bayes text classification. In: AAAI-1998 Learning for Text Categorization Workshop
- McCallum, Nigam
- 1998
(Show Context)
Citation Context ...ing the term frequency into the model, not directly proposed but implied by a lot of work in text categorization, is to regard a document as being generated from a unigram language model (Kalt, 1996; =-=McCallum and Nigam, 1998-=-). With query generation, p(D, Q | R) = p(Q | D, R)p(D | R), so we end up with the following ranking formula: log p(r | D, Q) p(Q | D, r) p(r | D) = log + log p(r | D, Q) p(Q | D, r) p(r | D) Under th... |

9 |
Bayesian network models for information retrieval
- Ribeiro-Neto, Silva, et al.
- 2000
(Show Context)
Citation Context ...ility that the query is satisfied given that the document is observed. Other similar uses of the Bayesian belief network in retrieval are presented in (Fung and Favero, 1995; Ribeiro and Muntz, 1996; =-=Ribeiro-Neto et al., 2000-=-). The inference network model is a much more general formalism than most of the models that we have discussed above. With different ways to realize the probabilistic relationship between the observat... |

7 | editors (2001 - Voorhees, Harman |

4 | Probabilistic IR models based on query and document generation - Lafferty, Zhai - 2001 |

2 |
Extended Boolean retrieval: a heuristic approach
- Rousseau
- 1990
(Show Context)
Citation Context ...egular vector space similarity measure as p becomes smaller. However, the model must rely on some assumptions about the Boolean structure of a query, and has some undesirable mathematical properties (=-=Rousseau, 1990-=-). There has also been little, if any, large-scale evaluation of the model. The vector space model is by far the most popular retrieval model due to its simplicity and effectiveness. The following is ... |