Re-examining the effects of adding relevance information in a relevance feedback environment

Wong, Woon K.; Luk, Robert W. P.; Leong, Hong Va; Ho, Kei Shiu; Lee, Dik Lun

doi:10.1016/j.ipm.2007.12.002

Cited by 17 publications

(17 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The rationale is that terms that were correctly ranked higher (because more relevant to the specific query at hand) will be downweighted if their relevance value with respect to the entire collection being searched is low. This observation has been confirmed in several experiments where the use of a distribution difference-based scoring function for both query expansion and reweighting achieved the best retrieval effectiveness, not only for English (e.g., , [Wong et al 2008]) but also for other european ] and asian languages [Savoy 2005]. Even a simple reweighting scheme based on an inverse function of term ranks may produce good results (e.g., [Carpineto et al 2002], [Hu et al 2006]).…”

Section: Query Reformulationmentioning

confidence: 82%

“…Some research has been carried out on the optimum number of features to include and there are differing suggestions ranging from five-ten features ( [Chang et al 2006], ]) to a few hundred ( [Buckley et al 1995], [Wong et al 2008], [Bernardini and Carpineto 2008]). On the other hand, the performance decrease associated with non-optimal values is usually modest , and most experimental studies agree that the number of expansion features is of low relevance.…”

Section: · 25mentioning

confidence: 99%

“…Several simple normalization schemes, discussed in [Wong et al 2008], have been proposed; usually they produce comparable results, although more powerful methods that not only scale data into the same range but also increase its uniformity could be more effective [Montague and Aslam 2001].…”

Section: Query Reformulationmentioning

confidence: 99%

“…The list in Table I is not exhaustive. Other term-scoring functions, including variants of those reported in this paper, are considered in [Efthimiadis 1993], , and [Wong et al 2008].…”

mentioning

confidence: 99%

“…To compute the probability of occurrence of a term t in X (whether the set of pseudo-relevant documents R or the whole collection C), the maximum likelihood criterion is often adopted, i.e., the ratio between the number of occurrences of t in X, treated as a long sequence of terms, and the number of terms in X. A different probability estimate is to use the fraction of documents in X that contain the term t. This latter criterion, generally used to compute the binary independence model (BIM) and Robertson selection value (RSV) functions, has also been applied to Chi-square and Kullback-Leibler distance (KLD) in a recent experimental study [Wong et al 2008], with very good …”

mentioning

confidence: 99%

See 4 more Smart Citations

A Survey of Automatic Query Expansion in Information Retrieval

2012

View full text Add to dashboard Cite

The relative ineffectiveness of information retrieval systems is largely caused by the inaccuracy with which a query formed by a few keywords models the actual user information need. One well known method to overcome this limitation is automatic query expansion (AQE), whereby the user's original query is augmented by new features with a similar meaning. AQE has a long history in the information retrieval community but it is only in the last years that it has reached a level of scientific and experimental maturity, especially in laboratory settings such as TREC. This survey presents a unified view of a large number of recent approaches to AQE that leverage various data sources and employ very different principles and techniques. The following questions are addressed: Why is query expansion so important to improve search effectiveness? What are the main steps involved in the design and implementation of an AQE component? What approaches to AQE are available and how do they compare? Which issues must still be resolved before AQE becomes a standard component of large operational information retrieval systems (e.g., search engines)?

show abstract

Section: Query Reformulationmentioning

confidence: 82%

Section: · 25mentioning

confidence: 99%

Section: Query Reformulationmentioning

confidence: 99%

“…The list in Table I is not exhaustive. Other term-scoring functions, including variants of those reported in this paper, are considered in [Efthimiadis 1993], , and [Wong et al 2008].…”

mentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

A Survey of Automatic Query Expansion in Information Retrieval

2012

View full text Add to dashboard Cite

show abstract

A new context‐dependent term weight computed by boost and discount using relevance information

Dang

Luk

Allan

et al. 2010

J. Am. Soc. Inf. Sci.

Self Cite

View full text Add to dashboard Cite

We studied the effectiveness of a new class of contextdependent term weights for information retrieval. Unlike the traditional term frequency-inverse document frequency (TF-IDF ), the new weighting of a term t in a document d depends not only on the occurrence statistics of t alone but also on the terms found within a text window (or "document-context") centered on t. We introduce a Boost and Discount (B&D) procedure which utilizes partial relevance information to compute the contextdependent term weights of query terms according to a logistic regression model. We investigate the effectiveness of the new term weights compared with the contextindependent BM25 weights in the setting of relevance feedback. We performed experiments with title queries of the TREC-6, -7, -8, and 2005 collections, comparing the residual Mean Average Precision (MAP) measures obtained using B&D term weights and those obtained by a baseline using BM25 weights. Given either 10 or 20 relevance judgments of the top retrieved documents, using the new term weights yields improvement over the baseline for all collections tested. The MAP obtained with the new weights has relative improvement over the baseline by 3.3 to 15.2%, with statistical significance at the 95% confidence level across all four collections.

show abstract