Search (29 results, page 1 of 2)

0.0783509 = product of:
  0.11752635 = sum of:
    0.10143948 = weight(_text_:systematic in 2547) [ClassicSimilarity], result of:
      0.10143948 = score(doc=2547,freq=4.0), product of:
        0.28397155 = queryWeight, product of:
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.049684696 = queryNorm
        0.35721707 = fieldWeight in 2547, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.03125 = fieldNorm(doc=2547)
    0.016086869 = product of:
      0.032173738 = sum of:
        0.032173738 = weight(_text_:indexing in 2547) [ClassicSimilarity], result of:
          0.032173738 = score(doc=2547,freq=2.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.16916946 = fieldWeight in 2547, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.03125 = fieldNorm(doc=2547)
      0.5 = coord(1/2)
  0.6666667 = coord(2/3)

Abstract

The discovery service (a search engine and service called WILBERT) used at our library at the Technical University of Applied Sciences Wildau (TUAS Wildau) is comprised of more than 8 million items. If we were to record all licensed publications in this tool to a higher level of articles, including their bibliographic records and full texts, we would have a holding estimated at a hundred million documents. A lot of features, such as ranking, autocompletion, multi-faceted classification, refining opportunities reduce the number of hits. However, it is not enough to give intuitive support for a systematic overview of topics related to documents in the library. John Naisbitt once said: "We are drowning in information, but starving for knowledge." This quote is still very true today. Two years ago, we started to develop micro thesauri for MINT topics in order to develop an advanced indexing of the library stock. We use iQvoc as a vocabulary management system to create the thesaurus. It provides an easy-to-use browser interface that builds a SKOS thesaurus in the background. The purpose of this is to integrate the thesauri in WILBERT in order to offer a better subject-related search. This approach especially supports first-year students by giving them the possibility to browse through a hierarchical alignment of a subject, for instance, logistics or computer science, and thereby discover how the terms are related. It also supports the students with an insight into established abbreviations and alternative labels. Students at the TUAS Wildau were involved in the developmental process of the software regarding the interface and functionality of iQvoc. The first steps have been taken and involve the inclusion of 3000 terms in our discovery tool WILBERT.

0.07317951 = product of:
  0.10976927 = sum of:
    0.08966068 = weight(_text_:systematic in 1078) [ClassicSimilarity], result of:
      0.08966068 = score(doc=1078,freq=2.0), product of:
        0.28397155 = queryWeight, product of:
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.049684696 = queryNorm
        0.31573826 = fieldWeight in 1078, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.0390625 = fieldNorm(doc=1078)
    0.020108584 = product of:
      0.04021717 = sum of:
        0.04021717 = weight(_text_:indexing in 1078) [ClassicSimilarity], result of:
          0.04021717 = score(doc=1078,freq=2.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.21146181 = fieldWeight in 1078, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.0390625 = fieldNorm(doc=1078)
      0.5 = coord(1/2)
  0.6666667 = coord(2/3)

Abstract

In this article and two other articles which conceptualize a future stage of the research program (Leide, Cole, Large, & Beheshti, submitted for publication; Cole, Leide, Large, Beheshti, & Brooks, in preparation), we map-out a domain novice user's encounter with an IR system from beginning to end so that appropriate classification-based visualization schemes can be inserted into the encounter process. This article describes the visualization of a navigation classification scheme only. The navigation classification scheme uses the metaphor of a ship and ship's navigator traveling through charted (but unknown to the user) waters, guided by a series of lighthouses. The lighthouses contain mediation interfaces linking the user to the information store through agents created for each. The user's agent is the cognitive model the user has of the information space, which the system encourages to evolve via interaction with the system's agent. The system's agent is an evolving classification scheme created by professional indexers to represent the structure of the information store. We propose a more systematic, multidimensional approach to creating evolving classification/indexing schemes, based on where the user is and what she is trying to do at that moment during the search session.

0.029886894 = product of:
  0.08966068 = sum of:
    0.08966068 = weight(_text_:systematic in 4837) [ClassicSimilarity], result of:
      0.08966068 = score(doc=4837,freq=2.0), product of:
        0.28397155 = queryWeight, product of:
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.049684696 = queryNorm
        0.31573826 = fieldWeight in 4837, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.0390625 = fieldNorm(doc=4837)
  0.33333334 = coord(1/3)

Abstract

Icon system could represent an efficient solution for collective iconic categorization of knowledge by providing graphical interpretation. Their pictorial characters assist visualizing the structure of text to become more understandable beyond vocabulary obstacle. In this paper we are proposing a Knowledge Engineering (KM) based iconic representation approach. We assume that these systematic icons improve collective knowledge management. Meanwhile, text (constructed under our knowledge management model - Hypertopic) helps to reduce the diversity of graphical understanding belonging to different users. This "position paper" also prepares to demonstrate our hypothesis by an "iconic social tagging" experiment which is to be accomplished in 2011 with UTT students. We describe the "socio semantic web" information portal involved in this project, and a part of the icons already designed for this experiment in Sustainability field. We have reviewed existing theoretical works on icons from various origins, which can be used to lay the foundation of robust "icons systems".

0.029886894 = product of:
  0.08966068 = sum of:
    0.08966068 = weight(_text_:systematic in 1214) [ClassicSimilarity], result of:
      0.08966068 = score(doc=1214,freq=2.0), product of:
        0.28397155 = queryWeight, product of:
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.049684696 = queryNorm
        0.31573826 = fieldWeight in 1214, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.715473 = idf(docFreq=395, maxDocs=44218)
          0.0390625 = fieldNorm(doc=1214)
  0.33333334 = coord(1/3)

Abstract

Complex cognitive activities, such as analytical reasoning, problem solving, and sense making, are often performed through the mediation of interactive computational tools. Examples include visual analytics, decision support, and educational tools. Through interaction with visual representations of information at the visual interface of these tools, a joint, coordinated cognitive system is formed. This partnership results in a number of relational properties-those depending on both humans and tools-that researchers and designers must be aware of if such tools are to effectively support the performance of complex cognitive activities. This article presents 10 properties of interactive visual representations that are essential and relational and whose values can be adjusted through interaction. By adjusting the values of these properties, better coordination between humans and tools can be effected, leading to higher quality performance of complex cognitive activities. This article examines how the values of these properties affect cognitive processing and visual reasoning and demonstrates the necessity of making their values adjustable-all of which is situated within a broader theoretical framework concerned with human-information interaction in complex cognitive activities. This framework can facilitate systematic research, design, and evaluation in numerous fields including information visualization, health informatics, visual analytics, and educational technology.

0.020983277 = product of:
  0.06294983 = sum of:
    0.06294983 = product of:
      0.12589966 = sum of:
        0.12589966 = weight(_text_:indexing in 1528) [ClassicSimilarity], result of:
          0.12589966 = score(doc=1528,freq=10.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.6619802 = fieldWeight in 1528, product of:
              3.1622777 = tf(freq=10.0), with freq of:
                10.0 = termFreq=10.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.0546875 = fieldNorm(doc=1528)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Statistical methods for automated document indexing are becoming an alternative to the manual assignment of keywords. We argue that the quality of the thesaurus used as a basis for indexing in regard to its ability to adequately cover the contents to be indexed and as a basis for the specific indexing method used is of crucial importance in automatic indexing. We present an interactive tool for thesaurus evaluation that is based on a combination of statistical measures and appropriate visualisation techniques that supports the detection of potential problems in a thesaurus. We describe the methods used and show that the tool supports the detection and correction of errors, leading to a better indexing result.

0.011609698 = product of:
  0.03482909 = sum of:
    0.03482909 = product of:
      0.06965818 = sum of:
        0.06965818 = weight(_text_:indexing in 3222) [ClassicSimilarity], result of:
          0.06965818 = score(doc=3222,freq=6.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.3662626 = fieldWeight in 3222, product of:
              2.4494898 = tf(freq=6.0), with freq of:
                6.0 = termFreq=6.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.0390625 = fieldNorm(doc=3222)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

The use of thesaurus-based indexing is a common approach for increasing the performance of information retrieval. In this thesis, we examine the suitability of a thesaurus for a given set of information and evaluate improvements of existing thesauri to get better search results. On this area, we focus on two aspects: 1. We demonstrate an analysis of the indexing results achieved by an automatic document indexer and the involved thesaurus. 2. We propose a method for thesaurus evaluation which is based on a combination of statistical measures and appropriate visualization techniques that support the detection of potential problems in a thesaurus. In this chapter, we give an overview of the context of our work. Next, we briefly outline the basics of thesaurus-based information retrieval and describe the Collexis Engine that was used for our experiments. In Chapter 3, we describe two experiments in automatically indexing documents in the areas of medicine and economics with corresponding thesauri and compare the results to available manual annotations. Chapter 4 describes methods for assessing thesauri and visualizing the result in terms of a treemap. We depict examples of interesting observations supported by the method and show that we actually find critical problems. We conclude with a discussion of open questions and future research in Chapter 5.

0.011219318 = product of:
  0.033657953 = sum of:
    0.033657953 = product of:
      0.06731591 = sum of:
        0.06731591 = weight(_text_:22 in 3406) [ClassicSimilarity], result of:
          0.06731591 = score(doc=3406,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.38690117 = fieldWeight in 3406, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.078125 = fieldNorm(doc=3406)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

30. 5.2010 16:22:35

0.011219318 = product of:
  0.033657953 = sum of:
    0.033657953 = product of:
      0.06731591 = sum of:
        0.06731591 = weight(_text_:22 in 2755) [ClassicSimilarity], result of:
          0.06731591 = score(doc=2755,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.38690117 = fieldWeight in 2755, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.078125 = fieldNorm(doc=2755)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

1. 2.2016 18:25:22

0.009519908 = product of:
  0.028559722 = sum of:
    0.028559722 = product of:
      0.057119444 = sum of:
        0.057119444 = weight(_text_:22 in 3355) [ClassicSimilarity], result of:
          0.057119444 = score(doc=3355,freq=4.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.32829654 = fieldWeight in 3355, product of:
              2.0 = tf(freq=4.0), with freq of:
                4.0 = termFreq=4.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.046875 = fieldNorm(doc=3355)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

22. 1.2017 16:54:03
22. 1.2017 17:10:56

0.009287758 = product of:
  0.027863273 = sum of:
    0.027863273 = product of:
      0.055726547 = sum of:
        0.055726547 = weight(_text_:indexing in 2643) [ClassicSimilarity], result of:
          0.055726547 = score(doc=2643,freq=6.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.2930101 = fieldWeight in 2643, product of:
              2.4494898 = tf(freq=6.0), with freq of:
                6.0 = termFreq=6.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.03125 = fieldNorm(doc=2643)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Content

1. Introduction Web search engines and digital libraries usually expect the users to use search terms that most accurately represent their information needs. Finding the most appropriate search terms to represent an information need is an age old problem in information retrieval. Keyword or phrase search may produce good search results as long as the search terms or phrase(s) match those used by the authors and have been chosen for indexing by the concerned information retrieval system. Since this does not always happen, a large number of false drops are produced by information retrieval systems. The retrieval results become worse in very large systems that deal with millions of records, such as the Web search engines and digital libraries. Vocabulary control tools are used to improve the performance of text retrieval systems. Thesauri, the most common type of vocabulary control tool used in information retrieval, appeared in the late fifties, designed for use with the emerging post-coordinate indexing systems of that time. They are used to exert terminology control in indexing, and to aid in searching by allowing the searcher to select appropriate search terms. A large volume of literature exists describing the design features, and experiments with the use, of thesauri in various types of information retrieval systems (see for example, Furnas et.al., 1987; Bates, 1986, 1998; Milstead, 1997, and Shiri et al., 2002).

0.0080434345 = product of:
  0.024130303 = sum of:
    0.024130303 = product of:
      0.048260607 = sum of:
        0.048260607 = weight(_text_:indexing in 2131) [ClassicSimilarity], result of:
          0.048260607 = score(doc=2131,freq=2.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.2537542 = fieldWeight in 2131, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.046875 = fieldNorm(doc=2131)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

This is a preprint to be published in the Extensions & Corrections to the UDC. The paper explains the basic functions of browsing and searching that need to be supported in relation to analytico-synthetic classifications such as Universal Decimal Classification (UDC), irrespective of any specific, real-life implementation. UDC is an example of a semi-faceted system that can be used, for instance, for both post-coordinate searching and hierarchical/facet browsing. The advantages of using a classification for IR, however, depend on the strength of the GUI, which should provide a user-friendly interface to classification browsing and searching. The power of this interface is in supporting visualisation that will 'convert' what is potentially a user-unfriendly indexing language based on symbols, to a subject presentation that is easy to understand, search and navigate. A summary of the basic functions of searching and browsing a classification that may be provided on a user-friendly interface is given and examples of classification browsing interfaces are provided.

0.007853523 = product of:
  0.023560567 = sum of:
    0.023560567 = product of:
      0.047121134 = sum of:
        0.047121134 = weight(_text_:22 in 2500) [ClassicSimilarity], result of:
          0.047121134 = score(doc=2500,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.2708308 = fieldWeight in 2500, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0546875 = fieldNorm(doc=2500)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

30. 1.2007 18:22:41

0.0075834226 = product of:
  0.022750268 = sum of:
    0.022750268 = product of:
      0.045500536 = sum of:
        0.045500536 = weight(_text_:indexing in 3646) [ClassicSimilarity], result of:
          0.045500536 = score(doc=3646,freq=4.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.23924173 = fieldWeight in 3646, product of:
              2.0 = tf(freq=4.0), with freq of:
                4.0 = termFreq=4.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.03125 = fieldNorm(doc=3646)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

The use of diagrams to express relationships in classification is not new. Many classificationists have used this approach, but usually in a minor display to make a point or for part of a difficult relational situation. Ranganathan, for example, used diagrams for some of his more elusive concepts. The thesaurus in particular and subject headings in general, with direct and indirect crossreferences or equivalents, need many more diagrams than normally are included to make relationships and even semantics clear. A picture very often is worth a thousand words. Rolling has used directed graphs (arrowgraphs) to join terms as a practical method for rendering relationships between indexing terms lucid. He has succeeded very weIl in this endeavor. Four diagrams in this selection are all that one needs to explain how to employ the system; from initial listing to completed arrowgraph. The samples of his work include illustration of off-page connectors between arrowgraphs. The great advantage to using diagrams like this is that they present relations between individual terms in a format that is easy to comprehend. But of even greater value is the fact that one can use his arrowgraphs as schematics for making three-dimensional wire-and-ball models, in which the relationships may be seen even more clearly. In fact, errors or gaps in relations are much easier to find with this methodology. One also can get across the notion of the threedimensionality of classification systems with such models. Pettee's "hand reaching up and over" (q.v.) is not a figment of the imagination. While the actual hand is a wire or stick, the concept visualized is helpful in illuminating the three-dimensional figure that is latent in all systems that have cross-references or "broader," "narrower," or, especially, "related" terms. Classification schedules, being hemmed in by the dimensions of the printed page, also benefit from such physical illustrations. Rolling, an engineer by conviction, was the developer of information systems for the Cobalt Institute, the European Atomic Energy Community, and European Coal and Steel Community. He also developed and promoted computer-aided translation at the Commission of the European Communities in Luxembourg. One of his objectives has always been to increase the efficiency of mono- and multilingual thesauri for use in multinational information systems.

0.0067315903 = product of:
  0.02019477 = sum of:
    0.02019477 = product of:
      0.04038954 = sum of:
        0.04038954 = weight(_text_:22 in 1289) [ClassicSimilarity], result of:
          0.04038954 = score(doc=1289,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.23214069 = fieldWeight in 1289, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.046875 = fieldNorm(doc=1289)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Content

Vortrag anlässlich des Workshops: "Extending the multilingual capacity of The European Library in the EDL project Stockholm, Swedish National Library, 22-23 November 2007".

0.0067315903 = product of:
  0.02019477 = sum of:
    0.02019477 = product of:
      0.04038954 = sum of:
        0.04038954 = weight(_text_:22 in 1781) [ClassicSimilarity], result of:
          0.04038954 = score(doc=1781,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.23214069 = fieldWeight in 1781, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.046875 = fieldNorm(doc=1781)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

22. 3.2008 14:35:21

0.0067315903 = product of:
  0.02019477 = sum of:
    0.02019477 = product of:
      0.04038954 = sum of:
        0.04038954 = weight(_text_:22 in 3693) [ClassicSimilarity], result of:
          0.04038954 = score(doc=3693,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.23214069 = fieldWeight in 3693, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.046875 = fieldNorm(doc=3693)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

22. 7.2010 19:36:46

0.0067315903 = product of:
  0.02019477 = sum of:
    0.02019477 = product of:
      0.04038954 = sum of:
        0.04038954 = weight(_text_:22 in 1709) [ClassicSimilarity], result of:
          0.04038954 = score(doc=1709,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.23214069 = fieldWeight in 1709, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.046875 = fieldNorm(doc=1709)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

4. 2.2015 9:22:39

0.0067028617 = product of:
  0.020108584 = sum of:
    0.020108584 = product of:
      0.04021717 = sum of:
        0.04021717 = weight(_text_:indexing in 82) [ClassicSimilarity], result of:
          0.04021717 = score(doc=82,freq=2.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.21146181 = fieldWeight in 82, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.0390625 = fieldNorm(doc=82)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Theme

Citation indexing

0.0067028617 = product of:
  0.020108584 = sum of:
    0.020108584 = product of:
      0.04021717 = sum of:
        0.04021717 = weight(_text_:indexing in 10) [ClassicSimilarity], result of:
          0.04021717 = score(doc=10,freq=2.0), product of:
            0.19018644 = queryWeight, product of:
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.049684696 = queryNorm
            0.21146181 = fieldWeight in 10, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.8278677 = idf(docFreq=2614, maxDocs=44218)
              0.0390625 = fieldNorm(doc=10)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Recent efforts in biomedical visual question answering (VQA) research rely on combined information gathered from the image content and surrounding text supporting the figure. Biomedical journals are a rich source of information for such multimodal content indexing. For multipanel figures in these journals, it is critical to develop automatic figure panel splitting and label recognition algorithms to associate individual panels with text metadata in the figure caption and the body of the article. Challenges in this task include large variations in figure panel layout, label location, size, contrast to background, and so on. In this work, we propose a deep convolutional neural network, which splits the panels and recognizes the panel labels in a single step. Visual features are extracted from several layers at various depths of the backbone neural network and organized to form a feature pyramid. These features are fed into classification and regression networks to generate candidates of panels and their labels. These candidates are merged to create the final panel segmentation result through a beam search algorithm. We evaluated the proposed algorithm on the ImageCLEF data set and achieved better performance than the results reported in the literature. In order to thoroughly investigate the proposed algorithm, we also collected and annotated our own data set of 10,642 figures. The experiments, trained on 9,642 figures and evaluated on the remaining 1,000 figures, show that combining panel splitting and panel label recognition mutually benefit each other.

0.005609659 = product of:
  0.016828977 = sum of:
    0.016828977 = product of:
      0.033657953 = sum of:
        0.033657953 = weight(_text_:22 in 1397) [ClassicSimilarity], result of:
          0.033657953 = score(doc=1397,freq=2.0), product of:
            0.17398734 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.049684696 = queryNorm
            0.19345059 = fieldWeight in 1397, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0390625 = fieldNorm(doc=1397)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Date

22. 3.2008 14:29:25