Search (2 results, page 1 of 1)

0.007663213 = product of:
  0.015326426 = sum of:
    0.015326426 = product of:
      0.030652853 = sum of:
        0.030652853 = weight(_text_:web in 6690) [ClassicSimilarity], result of:
          0.030652853 = score(doc=6690,freq=2.0), product of:
            0.17002425 = queryWeight, product of:
              3.2635105 = idf(docFreq=4597, maxDocs=44218)
              0.052098576 = queryNorm
            0.18028519 = fieldWeight in 6690, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.2635105 = idf(docFreq=4597, maxDocs=44218)
              0.0390625 = fieldNorm(doc=6690)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Abstract

In assessing information retrieval systems, it is important to know not only the precision of the retrieved set, but also to compare the number of retrieved relevant items to the total number of relevant items. For large collections, such as the TREC test collections, or the World Wide Web, it is not possible to enumerate the entire set of relevant documents. If the retrieved documents are evaluated, a variant of the statistical "capture-recapture" method can be used to estimate the total number of relevant documents, providing the several retrieval systems used are sufficiently independent. We show that the underlying signal detection model supporting such an analysis can be extended in two ways. First, assuming that there are two distinct performance characteristics (corresponding to the chance of retrieving a relevant, and retrieving a given non-relevant document), we show that if there are three or more independent systems available it is possible to estimate the number of relevant documents without actually having to decide whether each individual document is relevant. We report applications of this 3-system method to the TREC data, leading to the conclusion that the independence assumptions are not satisfied. We then extend the model to a multi-system, multi-problem model, and show that it is possible to include statistical dependencies of all orders in the model, and determine the number of relevant documents for each of the problems in the set. Application to the TREC setting will be presented

0.0038316066 = product of:
  0.007663213 = sum of:
    0.007663213 = product of:
      0.015326426 = sum of:
        0.015326426 = weight(_text_:web in 6299) [ClassicSimilarity], result of:
          0.015326426 = score(doc=6299,freq=2.0), product of:
            0.17002425 = queryWeight, product of:
              3.2635105 = idf(docFreq=4597, maxDocs=44218)
              0.052098576 = queryNorm
            0.09014259 = fieldWeight in 6299, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.2635105 = idf(docFreq=4597, maxDocs=44218)
              0.01953125 = fieldNorm(doc=6299)
      0.5 = coord(1/2)
  0.5 = coord(1/2)

Footnote

Rez. in: Machine translation review: 1999, no.10, S.26-27 (D. Lewis): "Cross Language Information Retrieval (CLIR) addresses the growing need to access large volumes of data across language boundaries. The typical requirement is for the user to input a free form query, usually a brief description of a topic, into a search or retrieval engine which returns a list, in ranked order, of documents or web pages that are relevant to the topic. The search engine matches the terms in the query to indexed terms, usually keywords previously derived from the target documents. Unlike monolingual information retrieval, CLIR requires query terms in one language to be matched to indexed terms in another. Matching can be done by bilingual dictionary lookup, full machine translation, or by applying statistical methods. A query's success is measured in terms of recall (how many potentially relevant target documents are found) and precision (what proportion of documents found are relevant). Issues in CLIR are how to translate query terms into index terms, how to eliminate alternative translations (e.g. to decide that French 'traitement' in a query means 'treatment' and not 'salary'), and how to rank or weight translation alternatives that are retained (e.g. how to order the French terms 'aventure', 'business', 'affaire', and 'liaison' as relevant translations of English 'affair'). Grefenstette provides a lucid and useful overview of the field and the problems. The volume brings together a number of experiments and projects in CLIR. Mark Davies (New Mexico State University) describes Recuerdo, a Spanish retrieval engine which reduces translation ambiguities by scanning indexes for parallel texts; it also uses either a bilingual dictionary or direct equivalents from a parallel corpus in order to compare results for queries on parallel texts. Lisa Ballesteros and Bruce Croft (University of Massachusetts) use a 'local feedback' technique which automatically enhances a query by adding extra terms to it both before and after translation; such terms can be derived from documents known to be relevant to the query.