Search (3 results, page 1 of 1)

0.0053522647 = product of:
  0.013380662 = sum of:
    0.008595286 = weight(_text_:a in 127) [ClassicSimilarity], result of:
      0.008595286 = score(doc=127,freq=26.0), product of:
        0.053464882 = queryWeight, product of:
          1.153047 = idf(docFreq=37942, maxDocs=44218)
          0.046368346 = queryNorm
        0.16076508 = fieldWeight in 127, product of:
          5.0990195 = tf(freq=26.0), with freq of:
            26.0 = termFreq=26.0
          1.153047 = idf(docFreq=37942, maxDocs=44218)
          0.02734375 = fieldNorm(doc=127)
    0.004785376 = product of:
      0.009570752 = sum of:
        0.009570752 = weight(_text_:information in 127) [ClassicSimilarity], result of:
          0.009570752 = score(doc=127,freq=6.0), product of:
            0.08139861 = queryWeight, product of:
              1.7554779 = idf(docFreq=20772, maxDocs=44218)
              0.046368346 = queryNorm
            0.11757882 = fieldWeight in 127, product of:
              2.4494898 = tf(freq=6.0), with freq of:
                6.0 = termFreq=6.0
              1.7554779 = idf(docFreq=20772, maxDocs=44218)
              0.02734375 = fieldNorm(doc=127)
      0.5 = coord(1/2)
  0.4 = coord(2/5)

Abstract

This thesis is organised as follows: Chapter 2 gives a general introduction to the field of information retrieval, covering its most important aspects. Further, the tasks of distributed and peer-to-peer information retrieval (P2PIR) are introduced, motivating their application and characterising the special challenges that they involve, including a review of existing architectures and search protocols in P2PIR. Finally, chapter 2 presents approaches to evaluating the e ectiveness of both traditional and peer-to-peer IR systems. Chapter 3 contains a detailed account of state-of-the-art information retrieval models and algorithms. This encompasses models for matching queries against document representations, term weighting algorithms, approaches to feedback and associative retrieval as well as distributed retrieval. It thus defines important terminology for the following chapters. The notion of "multi-level association graphs" (MLAGs) is introduced in chapter 4. An MLAG is a simple, graph-based framework that allows to model most of the theoretical and practical approaches to IR presented in chapter 3. Moreover, it provides an easy-to-grasp way of defining and including new entities into IR modeling, such as paragraphs or peers, dividing them conceptually while at the same time connecting them to each other in a meaningful way. This allows for a unified view on many IR tasks, including that of distributed and peer-to-peer search. Starting from related work and a formal defiition of the framework, the possibilities of modeling that it provides are discussed in detail, followed by an experimental section that shows how new insights gained from modeling inside the framework can lead to novel combinations of principles and eventually to improved retrieval effectiveness.
Chapter 5 empirically tackles the first of the two research questions formulated above, namely the question of global collection statistics. More precisely, it studies possibilities of radically simplified results merging. The simplification comes from the attempt - without having knowledge of the complete collection - to equip all peers with the same global statistics, making document scores comparable across peers. Chapter 5 empirically tackles the first of the two research questions formulated above, namely the question of global collection statistics. More precisely, it studies possibilities of radically simplified results merging. The simplification comes from the attempt - without having knowledge of the complete collection - to equip all peers with the same global statistics, making document scores comparable across peers. What is examined, is the question of how we can obtain such global statistics and to what extent their use will lead to a drop in retrieval effectiveness. In chapter 6, the second research question is tackled, namely that of making forwarding decisions for queries, based on profiles of other peers. After a review of related work in that area, the chapter first defines the approaches that will be compared against each other. Then, a novel evaluation framework is introduced, including a new measure for comparing results of a distributed search engine against those of a centralised one. Finally, the actual evaluation is performed using the new framework.

0.004915534 = product of:
  0.012288835 = sum of:
    0.008341924 = weight(_text_:a in 1842) [ClassicSimilarity], result of:
      0.008341924 = score(doc=1842,freq=12.0), product of:
        0.053464882 = queryWeight, product of:
          1.153047 = idf(docFreq=37942, maxDocs=44218)
          0.046368346 = queryNorm
        0.15602624 = fieldWeight in 1842, product of:
          3.4641016 = tf(freq=12.0), with freq of:
            12.0 = termFreq=12.0
          1.153047 = idf(docFreq=37942, maxDocs=44218)
          0.0390625 = fieldNorm(doc=1842)
    0.003946911 = product of:
      0.007893822 = sum of:
        0.007893822 = weight(_text_:information in 1842) [ClassicSimilarity], result of:
          0.007893822 = score(doc=1842,freq=2.0), product of:
            0.08139861 = queryWeight, product of:
              1.7554779 = idf(docFreq=20772, maxDocs=44218)
              0.046368346 = queryNorm
            0.09697737 = fieldWeight in 1842, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.7554779 = idf(docFreq=20772, maxDocs=44218)
              0.0390625 = fieldNorm(doc=1842)
      0.5 = coord(1/2)
  0.4 = coord(2/5)

Abstract

Many terminology engineering processes involve the task of automatic terminology extraction: before the terminology of a given domain can be modelled, organised or standardised, important concepts (or terms) of this domain have to be identified and fed into terminological databases. These serve in further steps as a starting point for compiling dictionaries, thesauri or maybe even terminological ontologies for the domain. For the extraction of the initial concepts, extraction methods are needed that operate on specialised language texts. On the other hand, many machine learning or information retrieval applications require automatic indexing techniques. In Machine Learning applications concerned with the automatic clustering or classification of texts, often feature vectors are needed that describe the contents of a given text briefly but meaningfully. These feature vectors typically consist of a fairly small set of index terms together with weights indicating their importance. Short but meaningful descriptions of document contents as provided by good index terms are also useful to humans: some knowledge management applications (e.g. topic maps) use them as a set of basic concepts (topics). The author believes that the tasks of terminology extraction and automatic indexing have much in common and can thus benefit from the same set of basic algorithms. It is the goal of this paper to outline some methods that may be used in both contexts, but also to find the discriminating factors between the two tasks that call for the variation of parameters or application of different techniques. The discussion of these methods will be based on statistical, syntactical and especially morphological properties of (index) terms. The paper is concluded by the presentation of some qualitative and quantitative results comparing statistical and morphological methods.

Type

a

7.893822E-4 = product of:
  0.003946911 = sum of:
    0.003946911 = product of:
      0.007893822 = sum of:
        0.007893822 = weight(_text_:information in 126) [ClassicSimilarity], result of:
          0.007893822 = score(doc=126,freq=2.0), product of:
            0.08139861 = queryWeight, product of:
              1.7554779 = idf(docFreq=20772, maxDocs=44218)
              0.046368346 = queryNorm
            0.09697737 = fieldWeight in 126, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.7554779 = idf(docFreq=20772, maxDocs=44218)
              0.0390625 = fieldNorm(doc=126)
      0.5 = coord(1/2)
  0.2 = coord(1/5)

Abstract

Die vorliegende Arbeit beschäftigt sich mit einem Teilgebiet des TextMining, versucht also Information (in diesem Fall Fachterminologie) aus natürlichsprachlichem Text zu extrahieren. Die der Arbeit zugrundeliegende These besagt, daß in vielen Gebieten des Text Mining die Kombination verschiedener Methoden sinnvoll sein kann, um dem Facettenreichtum natürlicher Sprache gerecht zu werden. Die bei der Terminologie-Extraktion angewandten Methoden sind statistischer und linguistischer (bzw. musterbasierter) Natur. Um sie herzuleiten, wurden einige Eigenschaften von Fachtermini herausgearbeitet, die für deren Extraktion relevant sind. So läßt sich z.B. die Tatsache, daß viele Fachbegriffe Nominalphrasen einer bestimmten Form sind, direkt für eine Suche nach gewissen POS-Mustern ausnützen, die Verteilung von Termen in Fachtexten führte zu einem statistischen Ansatz - der Differenzanalyse. Zusammen mit einigen weiteren wurden diese Ansätze in ein Verfahren integriert, welches in der Lage ist, aus dem Feedback eines Anwenders zu lernen und in mehreren Schritten die Suche nach Terminologie zu verfeinern. Dabei wurden mehrere Parameter des Verfahrens veränderlich belassen, d.h. der Anwender kann sie beliebig anpassen. Bei der Untersuchung der Ergebnisse anhand von zwei Fachtexten aus unterschiedlichen Domänen wurde deutlich, daß sich zwar die verschiedenen Verfahren gut ergänzen, daß aber die optimalen Werte der veränderbaren Parameter, ja selbst die Auswahl der angewendeten Verfahren text- und domänenabhängig sind.