Daniel Knaus, Elke Mittendorf, and Peter Schauble. Improving a basic retrieval method by links and passage level evidence. In Third Text REtrieval Conference (TREC-3), 1995. In press.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....categories: 1. Many attempts have been made to adjust the ranking of documents with respect to a query based upon the best passage (or passages) of the document: the belief or similarity of a document is replaced or modified by the belief or similarity of one of its passages. RWJ 95, Cal94, KMS95, SAB93, Wil94] In some cases, the passage similarity has been used as a precision enhancing filter. SB91, SAB93] These efforts have typically been successful in improving the effectiveness of retrieval. 2. Alternately, top ranking documents are assumed to be relevant and they or their passages ....

Daniel Knaus, Elke Mittendorf, and Peter Schauble. Improving a basic retrieval method by links and passage level evidence. In Third Text REtrieval Conference (TREC-3), 1995. In press.

....categories: 1. Many attempts have been made to adjust the ranking of documents with respect to a query based upon the best passage (or passages) of the document: the belief or similarity of a document is replaced or modified by the belief or similarity of one of its passages. RWJ 95, Cal94, KMS95, SAB93, Wil94] In some cases, the passage similarity has been used as a precision enhancing filter. SB91, SAB93] These efforts have typically been successful in improving the effectiveness of retrieval. 2. Alternately, top ranking documents are assumed to be relevant and they or their ....

Daniel Knaus, Elke Mittendorf, and Peter Schauble. Improving a basic retrieval method by links and passage level evidence. In Third Text REtrieval Conference (TREC-3), 1995. In press.

....the possible improvement in performance from these models would be limited. In this article, we examine in detail one such fusion model: the linear combination of scores (LC) The LC model has been used by many IR researchers with varying degrees of success: Bartell et al. 1994] Kantor, 1995] [Knaus et al. 1995], Selberg and Etzioni, 1996] Shaw and Fox, 1995] and [Vogt et al. 1997b] Our analysis of the model reveals what types of systems the model works best with, and explores techniques for training the model. This article has the following format. First we describe the model and the specific ....

Knaus, D., Mittendorf, E., and Sch auble, P. (1995). Improving a basic retrieval method by links and passage level evidence. In [Harman, 1995]. NIST Special Publication 500-226.

.... ; w 2 : given to each IR system: R(W; d; q) X systems w i R i (d; q) Or, for only two IR systems: R(w 1 ; w 2 ; d; q) w 1 R 1 (d; q) w 2 R 2 (d; q) 1) This straightforward approach has been attempted with varying degrees of success by a number of researchers ( Shaw and Fox, 1995] [Knaus et al. 1995], Kantor, 1995] in [Harman, 1995] Bartell et al. 1994] and [Selberg and Etzioni, 1996] and [Vogt et al. 1996] in [Harman, 1997] However, consistent, significant improvement has been elusive. An interesting question is: when is it even possible to improve the performance of two IR systems ....

Knaus, D., Mittendorf, E., and Sch"auble, P. (1995). Improving a basic retrieval method by links and passage level evidence. In

....the possible improvement in performance from these models would be limited. In this article, we examine in detail one such fusion model: the linear combination of scores (LC) The LC model has been used by many IR researchers with varying degrees of success: Bartell et al. 1994] Kantor, 1995] [Knaus et al. 1995], Selberg and Etzioni, 1996] Shaw and Fox, 1995] and [Vogt et al. 1997b] Our analysis of the model reveals what types of systems the model works best with, and explores techniques for training the model. This article has the following format. First we describe the model and the specific ....

Knaus, D., Mittendorf, E., and Sch auble, P. (1995). Improving a basic retrieval method by links and passage level evidence. In [Harman, 1995]. NIST Special Publication 500-226.

....The weight on that system would have to be low in order to account for the remainder (and presumably the majority) of the documents the ones for which the score was inaccurate. Nevertheless, this approach has been used with varying degrees of success by a number of researchers (e.g. 2] 8] [9], 13] 14] and [15] However, consistent, significant improvement has been elusive. An interesting question is: when is it even possible to improve the performance of two IR systems by linearly combining their estimates of relevance One study by Lee [10] has attempted to answer this question. ....

....; GPA ni ; C rel ; rel : The positive coefficient on GPA indicates that this measure should be maximized. In other words, the two systems should generally rank documents in their intersection similarly and the distribution of scores by both systems should be similar to each other. Knaus, et al. [9] have indicated that problems may occur when linearly combining systems that distribute RSV s differently, so this result is not surprising, and is also supported by the positive coefficient on C rel . On the other hand, the negative coefficients on GPA rel and GPAni indicate that these measures ....

Daniel Knaus, Elke Mittendorf, and Peter Sch¨auble. Improving a basic retrieval method by links and passage level evidence. In Harman [6]. NIST Special Publication.

....the best individual expert. All of their experiments, however, are on relatively small collections. Others have successfully used mixture (a.k.a. fusion) approaches on larger collections, including TREC, but they hand pick the model parameters, clearly an undesirable approach (see [Kantor, 1995] [Knaus et al. 1995], Shaw and Fox, 1995] in [Harman, 1995a] Here we show how the mixture technique coupled with automatic parameter adjustment via rank order statistic optimization scales up to the TREC collection. Our results indicate that we have yet to find the best class of model for this task. 2 METHOD 2.1 ....

Knaus, D., Mittendorf, E., and Schauble, P. (1995). Improving a basic retrieval method by links and passage level evidence. In [Harman, 1995a].

....the improvement in performance possible from these models would be limited. In this article, we examine in detail one such fusion model: the linear combination of scores (LC) The LC model has been used by many IR researchers with varying degrees of success [Bartell et al. 1994] Kantor, 1995] [Knaus et al. 1995], Selberg and Etzioni, 1996] Shaw and Fox, 1995] and D R A F T June 29, 1998, 4:28pm D R A F T Fusion Via a Linear Combination of Scores 3 [Vogt et al. 1997b] Our analysis of the model reveals what types of systems the model works best with and explores techniques for training the model. ....

Knaus, D., Mittendorf, E., and Sch auble, P. (1995). Improving a basic retrieval method by links and passage level evidence. In [Harman, 1995]. NIST Special Publication 500-226.

....effectively. With respect to r, the estimation formula (1) is not yet realistic because the constant c does not have the same effect for low and high values of r. The influence of c on the retrieval effectiveness is subject to further investigations. 3 Passage Retrieval In previous experiments [Knaus et al. 1994, Mittendorf Schauble, 1994] we have shown that Hidden Markov Models (HMM) provide a natural and effective method for retrieving relevant passages from documents. In this section we summarize the ideas of HMM based passage retrieval and we describe the refinements and changes we have made in ....

.... get probability distributions which are tractable given a query, we map each basic unit to a value which describes the similarity between the basic unit and the query; for example if a single indexing feature is the basic unit then the similarity is described by ff( i ; q) Delta nidf 2 ( i ) [Knaus et al. 1994], where the normalized inverse document frequency is defined by nidf( i ) 1 Gamma log(df( i ) 1) log(n 1) We assume that there are certain states of a HMM which model the production of the passages that are relevant to the query and there are other states that are responsible for ....

Knaus, D., Mittendorf, E., & Schauble, P. (1994). Improving a Basic Retrieval Method by Links and Passage Level Evidence. In TREC-3 Proceedings, pp. 241--246.

No context found.

D. Knaus, E. Mittendorf and P. Schauble, Improving a Basic Retrieval Method by Links and Passage Level Evidence, Text Retrieval Conference, Washington, D.C., November 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC