S. E. Robertson, "On term selection for query Expansion", Journal of Documentation, 46, 4, pp.359--364, 1990  Home/Search   Document Not in Database   Summary   Related Articles   Check

....highly match the seamher s query as a means of showing the searcher what kind of information has been retrieved. This system is intended to present the searcher with information on the whole retrieval result rather than just the top ten pages. A second system uses a form of relevance feedback, [8], to automatically update the information that the interface displays to the searcher. In this case the system uses implicit information information captured from the searchef s interaction with the interface to estimate what information may be of use to the searcher. This system is intended ....

....taken from the content of the assumed relevant summaries. Each time the system believes the searcher has identified a useful summary the content of all useful summaries is used to generate a list of possible query expansion terms. The function we used for this purpose is Robertsoh s wpq formula [8], equation 1. The top 6 terms, from the wpq ranking of terms, are added to the original query, and the new query is used to re rank the list of top ranking sentences. This system, then, dynamically updates the list of top ranking sentences each time the system assumes that the searcher has found ....

S. Robertson. On term selection for query expansion. Journal of Documentation. 46. pp 359-364. 1990.

....query generation method. The problem of retrieving documents that are relevant to a user s information need has been the core focus of the information retrieval (IR) field [27] Although our problem is different in nature, we exploit state of the art term weighting and query expansion results [24] from IR in the design of one variant of our system (Section 2.4.2) Alternatively, the characterization of the useful documents could be viewed as a traditional classification problem. We explore a number of machine learning techniques [8, 18] in the design of other variants of our system ....

....two machine learning text classifiers. Finally, we present a hybrid query generation technique that combines the learned queries from all of the above methods. Okapi: As a first query generation strategy, we exploit a state of the art term weighting scheme from IR, from the Okapi retrieval system [24]. While there are many promising alternatives to this weighting scheme in the IR literature (e.g. 29, 26] we chose Okapi because it has been demonstrated to perform well, is naturally well suited to our task, and is relatively straightforward to implement. Incorporating alternative information ....

[Article contains additional citation context not shown here]

S. Robertson. On term selection for query expansion. In Journal of Documentation, volume 46, 1990.

....QP as the number of relevant documents, and consider the number of the remaining #Question, Answer# pairs where ## # appears in the Answer as non relevant, and apply the formula in Equation 1. We then compute the term selection weights, ### # , for each candidate transform ## # , as described in [14] in the context of selecting terms for automatic query expansion as: ### # # ### # # # ### # (2) where ### # is the co occurrence count of ## # with QP,and# ### # is the relevance based term weight of ## # computed with respect to QP. This term ranking strategy exploits both co occurrence ....

S. Robertson. On term selection for query expansion. In Journal of Documentation, volume 46, pages 359--364, 1990.

....QP as the number of relevant documents, and consider the number of the remaining Question, Answer pairs where tr i appears in the Answer as non relevant, and apply the formula in Equation 1. We then compute the term selection weights, wtr i , for each candidate transform tr i , as described in [14] in the context of selecting terms for automatic query expansion as: wtr i = qtf i w (1) i (2) where qtf i is the co occurrence count of tr i with QP, and w (1) i is the relevance based term weight of tr i computed with respect to QP. This term ranking strategy exploits both co occurrence ....

S. Robertson. On term selection for query expansion. In Journal of Documentation, volume 46, pages 359--364, 1990.

....may be a valid evaluation methodology. However, as shown in the second section, this first impression is erroneous. Evaluation of Learning Schemes Used in Information Retrieval 13 Of course, variations on the previously described relevance feedback scheme have been suggested. For example, Robertson (1990) suggests adopting different procedures for the selection of new search terms and for the weighting of search keywords, see also (Buckley Salton 1995) For Harman (1992) the number of terms to be included in the expanded request can depend on the average document size of the underlying ....

Robertson, S. E. (1990). On term selection for query expansion. Journal of Documentation. 46(4), 359-364.

....from a variety of algorithms to execute the algebraic operations as e#cient as possible. Similar to the use of indices in relational databases, we may increase performance by the definition of access structures on 2 Even when the user enters a short query, the processing of relevance feedback [Rob90] HC93] and techniques like local context analysis [XC96] add many terms to an initial query. 18 On the integration of IR and databases submitted to DS8 the BATs. When using a parallel machine, the IR code automatically benefits from the implementation of parallel kernel algorithms. The ....

S.E. Robertson. On term selection for query expansion. Journal of documentation, 46(4):359--364, 1990.

....several of these ideas. The topic of this paper is not to discuss the different approaches or evaluate them, but to show an implementation of it to speed up IR systems. Many research groups have used Parts of Speech Tagging in Information Retrieval tasks [Lu et al. 97] Pederson et al. 97] [Robertson 90]. One idea has been to add noun phrases to all terms in an effort to better represent what the document is about and thus improve precision recall. Adding noun phrases to the index of all terms has been shown to improve precision recall [Zhai 97] However, Crestani97] showed that indexing noun ....

Robertson, S.E. On term selection for query expansion. Journal of Documentation 46. 4, 359-364

....of terms for query expansion were: wpq, emim, porter, r lohi and r hilo. These algorithms are described briefly below. 2.2. 1 The wpq algorithm This algorithm is based on an independence assumption that holds between a query expansion term and the terms in the entire previous search formulation (Robertson, 1990) . According to the relevance weighting theory, the inclusion of term t in the search formulation with weight w t will increase the effectiveness of retrieval by wpq = w t (p t Gamma q t ) 3) where, w t is a weighting function, which in this case is the w f4 ; p t is the probability of term t ....

Robertson, S.E. (1990) On term selection for query expansion.

....a great deal across differing document collections, and that there was more potential for improving retrieval where the initial performance was poor, than where output on the first pass was satisfactory. Another, very popular method of query expansion and reweighting is the probabilistic method [40], which will be described thoroughly in the next section. In general, methods for relevance feedback must be scored by their ability to rank unassessed relevant documents more highly than was possible before the feedback step. Therefore, the initially retrieved documents, which are seen and ....

....each term t i observed in the set of cut off messages could be calculated using the standard equation w i = log r i (N Gamma n i Gamma R r i ) R Gamma r i ) n i Gamma r i ) where the notation is that introduced in Section 3.4. The weighted terms were now ranked by their Offer Weight [40], simply defined as O i = r i w i ; and the top q terms selected for inclusion in the new query. Messages were now retrieved from the residual of the message collection using the newly formulated query. The entire retrieval output, based on the initial and fed back queries, was ranked by ....

S. E. Robertson. On Term Selection for Query Expansion. Journal of Documentation, 46:359--364, 1990.

....sophistication of span scoring. After construction of the initial query for each topic a number of partial queries were extracted, to be used in term implication runs explained below. The number of partial queries used for a topic ranged from one to five. 4. 4 Query Augmentation Robertson [10] argues that the best methods for selecting terms for query expansion and for weighting selected terms are not necessarily the same. In distance based queries, as formulated here, this difficulty is avoided because individual terms are not weighted. However, a new complication is introduced. When ....

S. E. Robertson. On term selection for query expansion. Journal of Documentation, 46(4):359--364, 1990.

....references to relevance feedback in fact refer to pseudo relevance feedback as there was no human involvement in the feedback process. Instead, highly ranked documents retrieved by an initial query were assumed to be sufficiently relevant as to constitute a useful source of additional query terms. Robertson [1990] argued that the weights used to select terms to be added to a query should, in general, be different from the document term weights used when processing the query. This approach has been taken here. 2.2.1 Method of Term Selection Instead of mining complete document text for new terms, only the ....

....initial query were not considered. All other terms were stemmed and stored in a hash table and their frequencies of occurrence within the hotspots were accumulated. Once all hotspots had been mined, selection values for each term in the hash table were computed according to the formula given by Robertson [1990]: a t = w t (p t Gamma q t ) In Robertson s work the p t and q t were the probabilities that a relevant and a non relevant document, respectively, contained the term t. Here, p t and q t are the probabilities that any particular term in a hotspot and not in a hotspot, respectively, is the ....

Robertson, S. E. 1990. On term selection for query expansion. Journal of Documentation 46, 4, 359--364.

No context found.

S E Robertson. On term selection for query expansion. Journal of Documentation, 46:359--364, 1990.

....of different scoring measures were tried. None performed better than the TREC average precision, and this was the only measure used for TREC 5. However, a number of new selection algorithms were tried. For all selection algorithms, the available terms were first arranged in descending RSV 3 [9] order to form a term pool of a given size. An initial query (sometimes empty) was formed from terms at the top of the pool, then terms 2 During term selection a bonus was generally given to topic terms by using a non zero value of k 3 in equation 3 3 Note that RSV stands for Robertson ....

Robertson S E. On term selection for query expansion. Journal of Documentation 46 Dec 1990 p359--364.

....will adding this term to the request benefit the overall performance of the search formulation . In particular, a very rare term, even though it is a strong indicator of relevance when it occurs, is not likely to have much overall effect. A specific model for the overall effect is discussed in Robertson (1990). For each candidate expansion term, this considers the distributions of the scores for relevant and non relevant documents, with the term in question present or absent. The model leads to a formula for a selection value for term t i , indicating the strength of its overall effect, which may ....

Robertson, S.E. (1990) On term selection for query expansion. Journal of Documentation, 46, 359-364.

....items are ranked according to some selection value which is intended to measure how useful they would be if added to the query. The formula usually used for this purpose (and in particular, the one used in TREC 1 and TREC 2) is the Robertson Selection Value (RSV) based on the argument in [8]. The formula given in that reference is w(p Gamma q) where w is the weight to be assigned to the term, p is the probability of the term occurring in a relevant document, and q is the probability that it occurs in a nonrelevant document. For RSV, w is interpreted as the usual Robertson Sparck ....

Robertson S E. On term selection for query expansion. Journal of Documentation 46 Dec 1990 p359-- 364.

.... blind expansion, where there are no explicit judgments of relevance. Further detail is given below. Term ranking for selection In [8] there is a brief discussion of some alternative ways of ranking potential expansion terms. It appeared that no method was superior to the method proposed in [15] by which terms are ranked in decreasing order of TSV = r:w (1) In line with the nonrelevance version of w (1) equation 3) Boughanem has proposed the more general function TSV = r=R Gamma ffs=S) w (1) 5) where ff 2 [0; 1] and r, R, s and S are as above. Passage determination and ....

Robertson, S.E. On term selection for query expansion. Journal of Documentation 46, Dec 1990, p359--364.

No context found.

Robertson, S.E. `On term selection for query expansion', Journal of Documentation, 46, 1990, 359-364.

.... blind expansion, where there are no explicit judgments of relevance. Further detail is given below. Term ranking for selection In [8] there is a brief discussion of some alternative ways of ranking potential expansion terms. It appeared that no method was superior to the method proposed in [12] by which terms are ranked in decreasing order of TSV = r:w (1) In line with the nonrelevance version of w (1) equation 3) Boughanem has proposed the more general function TSV = r=R Gamma ffs=S) w (1) 5) where ff 2 [0; 1] and r, R, s and S are as above. Passage determination and ....

Robertson, S.E. On term selection for query expansion. Journal of Documentation 46, Dec 1990, p359--364.

No context found.

S. E. Robertson, "On term selection for query Expansion", Journal of Documentation, 46, 4, pp.359--364, 1990

No context found.

Robertson, S. E. (1990). On term selection for query expansion. Journal of Documentation 46, 359--364.

No context found.

Robertson, S. E. On term selection for query expansion. Journal of Documentation, 46(4):359-364, 1990.

No context found.

Robertson, S.E., On Term Selection for Query Expansion. Journal of Documentation 46. 4. (1990) 359-364

No context found.

Robertson, S.E. `On term selection for query expansion'. Journal of Documentation. 46. 4, 359-364. 1990.

No context found.

S. Robertson, On term selection for query expansion, Journal of Documentation 46 (1990) 359 -- 364.

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