Search (13 results, page 1 of 1)

0.0036536194 = product of:
  0.010960858 = sum of:
    0.010960858 = product of:
      0.021921717 = sum of:
        0.021921717 = weight(_text_:of in 1139) [ClassicSimilarity], result of:
          0.021921717 = score(doc=1139,freq=14.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.31997898 = fieldWeight in 1139, product of:
              3.7416575 = tf(freq=14.0), with freq of:
                14.0 = termFreq=14.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0546875 = fieldNorm(doc=1139)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

In light of AI (Artificial Intelligence) and NLP (Natural language processing) technologies, this article examines the feasibility of using AI/NLP models to enhance the subject indexing of digital resources. While BERT (Bidirectional Encoder Representations from Transformers) models are widely used in scholarly communities, the authors assess whether BERT models can be used in machine-assisted indexing in the Project Gutenberg collection, through suggesting Library of Congress subject headings filtered by certain Library of Congress Classification subclass labels. The findings of this study are informative for further research on BERT models to assist with automatic subject indexing for digital library collections.

0.0033478998 = product of:
  0.010043699 = sum of:
    0.010043699 = product of:
      0.020087399 = sum of:
        0.020087399 = weight(_text_:of in 412) [ClassicSimilarity], result of:
          0.020087399 = score(doc=412,freq=16.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.2932045 = fieldWeight in 412, product of:
              4.0 = tf(freq=16.0), with freq of:
                16.0 = termFreq=16.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.046875 = fieldNorm(doc=412)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Genre plays an important role in the description, navigation, and discovery of movies, but it is rarely studied at large scale using quantitative methods. This allows an analysis of how genre labels are applied, how genres are composed and how these ingredients change, and how genres compare. We apply unsupervised topic modeling to a large collection of textual movie summaries and then use the model's topic proportions to investigate key questions in genre, including recognizability, mapping, canonicity, and change over time. We find that many genres can be quite easily predicted by their lexical signatures and this defines their position on the genre landscape. We find significant genre composition changes between periods for westerns, science fiction and road movies, reflecting changes in production and consumption values. We show that in terms of canonicity, canonical examples are often at the high end of the topic distribution profile for the genre rather than central as might be predicted by categorization theory.

Source

Journal of the Association for Information Science and Technology. 72(2021) no.12, S.1511-1527

0.0031316737 = product of:
  0.009395021 = sum of:
    0.009395021 = product of:
      0.018790042 = sum of:
        0.018790042 = weight(_text_:of in 720) [ClassicSimilarity], result of:
          0.018790042 = score(doc=720,freq=14.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.2742677 = fieldWeight in 720, product of:
              3.7416575 = tf(freq=14.0), with freq of:
                14.0 = termFreq=14.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.046875 = fieldNorm(doc=720)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

This project brought together undergraduate students in Computer Science with librarians to mine abstracts of articles from the Texas A&M University Libraries' institutional repository, OAKTrust, in order to probe the creation of new metadata to improve discovery and use. The mining operation task consisted simply of classifying the articles into two categories of research type: basic research ("for understanding," "curiosity-based," or "knowledge-based") and applied research ("use-based"). These categories are fundamental especially for funders but are also important to researchers. The mining-to-classification steps took several iterations, but ultimately, we achieved good results with the toolkit BERT (Bidirectional Encoder Representations from Transformers). The project and its workflows represent a preview of what may lie ahead in the future of crafting metadata using text mining techniques to enhance discoverability.

0.0031316737 = product of:
  0.009395021 = sum of:
    0.009395021 = product of:
      0.018790042 = sum of:
        0.018790042 = weight(_text_:of in 723) [ClassicSimilarity], result of:
          0.018790042 = score(doc=723,freq=14.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.2742677 = fieldWeight in 723, product of:
              3.7416575 = tf(freq=14.0), with freq of:
                14.0 = termFreq=14.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.046875 = fieldNorm(doc=723)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Manual subject indexing in libraries is a time-consuming and costly process and the quality of the assigned subjects is affected by the cataloger's knowledge on the specific topics contained in the book. Trying to solve these issues, we exploited the opportunities arising from artificial intelligence to develop Kratt: a prototype of an automatic subject indexing tool. Kratt is able to subject index a book independent of its extent and genre with a set of keywords present in the Estonian Subject Thesaurus. It takes Kratt approximately one minute to subject index a book, outperforming humans 10-15 times. Although the resulting keywords were not considered satisfactory by the catalogers, the ratings of a small sample of regular library users showed more promise. We also argue that the results can be enhanced by including a bigger corpus for training the model and applying more careful preprocessing techniques.

0.0031192217 = product of:
  0.009357665 = sum of:
    0.009357665 = product of:
      0.01871533 = sum of:
        0.01871533 = weight(_text_:of in 5816) [ClassicSimilarity], result of:
          0.01871533 = score(doc=5816,freq=20.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.27317715 = fieldWeight in 5816, product of:
              4.472136 = tf(freq=20.0), with freq of:
                20.0 = termFreq=20.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0390625 = fieldNorm(doc=5816)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Millions of messages are produced on microblog platforms every day, leading to the pressing need for automatic identification of key points from the massive texts. To absorb salient content from the vast bulk of microblog posts, this article focuses on the task of microblog keyphrase extraction. In previous work, most efforts treat messages as independent documents and might suffer from the data sparsity problem exhibited in short and informal microblog posts. On the contrary, we propose to enrich contexts via exploiting conversations initialized by target posts and formed by their replies, which are generally centered around relevant topics to the target posts and therefore helpful for keyphrase identification. Concretely, we present a neural keyphrase extraction framework, which has 2 modules: a conversation context encoder and a keyphrase tagger. The conversation context encoder captures indicative representation from their conversation contexts and feeds the representation into the keyphrase tagger, and the keyphrase tagger extracts salient words from target posts. The 2 modules were trained jointly to optimize the conversation context encoding and keyphrase extraction processes. In the conversation context encoder, we leverage hierarchical structures to capture the word-level indicative representation and message-level indicative representation hierarchically. In both of the modules, we apply character-level representations, which enables the model to explore morphological features and deal with the out-of-vocabulary problem caused by the informal language style of microblog messages. Extensive comparison results on real-life data sets indicate that our model outperforms state-of-the-art models from previous studies.

Source

Journal of the Association for Information Science and Technology. 71(2020) no.5, S.553-567

0.0029591531 = product of:
  0.008877459 = sum of:
    0.008877459 = product of:
      0.017754918 = sum of:
        0.017754918 = weight(_text_:of in 341) [ClassicSimilarity], result of:
          0.017754918 = score(doc=341,freq=18.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.25915858 = fieldWeight in 341, product of:
              4.2426405 = tf(freq=18.0), with freq of:
                18.0 = termFreq=18.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0390625 = fieldNorm(doc=341)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Purpose Based on the highlights of The Metropolitan Museum of Art's collection, the purpose of this paper is to examine the similarities and differences between the subject keywords tags assigned by the museum and those produced by three computer vision systems. Design/methodology/approach This paper uses computer vision tools to generate the data and the Getty Research Institute's Art and Architecture Thesaurus (AAT) to compare the subject keyword tags. Findings This paper finds that there are clear opportunities to use computer vision technologies to automatically generate tags that expand the terms used by the museum. This brings a new perspective to the collection that is different from the traditional art historical one. However, the study also surfaces challenges about the accuracy and lack of context within the computer vision results. Practical implications This finding has important implications on how these machine-generated tags complement the current taxonomies and vocabularies inputted in the collection database. In consequence, the museum needs to consider the selection process for choosing which computer vision system to apply to their collection. Furthermore, they also need to think critically about the kind of tags they wish to use, such as colors, materials or objects. Originality/value The study results add to the rapidly evolving field of computer vision within the art information context and provide recommendations of aspects to consider before selecting and implementing these technologies.

Source

Journal of documentation. 77(2021) no.4, S.946-964

0.0029591531 = product of:
  0.008877459 = sum of:
    0.008877459 = product of:
      0.017754918 = sum of:
        0.017754918 = weight(_text_:of in 658) [ClassicSimilarity], result of:
          0.017754918 = score(doc=658,freq=18.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.25915858 = fieldWeight in 658, product of:
              4.2426405 = tf(freq=18.0), with freq of:
                18.0 = termFreq=18.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0390625 = fieldNorm(doc=658)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Automated text classification is an important function for many AI systems relevant to libraries, including automated subject indexing and classification. When implemented using the traditional natural language processing (NLP) paradigm, one key part of the process is the normalization of words using stemming or lemmatization, which reduces the amount of linguistic variation and often improves the quality of classification. In this paper, we compare the output of seven different text lemmatization algorithms as well as two baseline methods. We measure how the choice of method affects the quality of text classification using example corpora in three languages. The experiments have been performed using the open source Annif toolkit for automated subject indexing and classification, but should generalize also to other NLP toolkits and similar text classification tasks. The results show that lemmatization methods in most cases outperform baseline methods in text classification particularly for Finnish and Swedish text, but not English, where baseline methods are most effective. The differences between lemmatization methods are quite small. The systematic comparison will help optimize text classification pipelines and inform the further development of the Annif toolkit to incorporate a wider choice of normalization methods.

0.002609728 = product of:
  0.007829184 = sum of:
    0.007829184 = product of:
      0.015658367 = sum of:
        0.015658367 = weight(_text_:of in 1024) [ClassicSimilarity], result of:
          0.015658367 = score(doc=1024,freq=14.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.22855641 = fieldWeight in 1024, product of:
              3.7416575 = tf(freq=14.0), with freq of:
                14.0 = termFreq=14.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0390625 = fieldNorm(doc=1024)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

There are several ways to employ machine learning for automating subject indexing. One popular strategy is to utilize a supervised learning algorithm to train a model on a set of documents that have been manually indexed by subject matter using a standard vocabulary. The resulting model can then predict the subject of new and previously unseen documents by identifying patterns learned from the training data. To do this, the first step is to gather a large dataset of documents and manually assign each document a set of subject keywords/descriptors from a controlled vocabulary (e.g., from Agrovoc). Next, the dataset (obtained from Agris) can be divided into - i) a training dataset, and ii) a test dataset. The training dataset is used to train the model, while the test dataset is used to evaluate the model's performance. Machine learning can be a powerful tool for automating the process of subject indexing. This research is an attempt to apply Annif (http://annif. org/), an open-source AI/ML framework, to autogenerate subject keywords/descriptors for documentary resources in the domain of agriculture. The training dataset is obtained from Agris, which applies the Agrovoc thesaurus as a vocabulary tool (https://www.fao.org/agris/download).

Source

¬SRELS Journal of Information Management. 60(2023) no.2, S.85-95

0.0019529418 = product of:
  0.005858825 = sum of:
    0.005858825 = product of:
      0.01171765 = sum of:
        0.01171765 = weight(_text_:of in 710) [ClassicSimilarity], result of:
          0.01171765 = score(doc=710,freq=4.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.17103596 = fieldWeight in 710, product of:
              2.0 = tf(freq=4.0), with freq of:
                4.0 = termFreq=4.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0546875 = fieldNorm(doc=710)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Subject headings and genre terms are notoriously difficult to apply, yet are important for fiction. The current project functions as a proof of concept, using a text-mining methodology to identify affective information (emotion and tone) about fiction titles from professional book reviews as a potential first step in automating the subject analysis process. Findings are presented and discussed, comparing results to the range of aboutness and isness information in library cataloging records. The methodology is likewise presented, and how future work might expand on the current project to enhance catalog records through text-mining is explored.

0.0019529418 = product of:
  0.005858825 = sum of:
    0.005858825 = product of:
      0.01171765 = sum of:
        0.01171765 = weight(_text_:of in 718) [ClassicSimilarity], result of:
          0.01171765 = score(doc=718,freq=4.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.17103596 = fieldWeight in 718, product of:
              2.0 = tf(freq=4.0), with freq of:
                4.0 = termFreq=4.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0546875 = fieldNorm(doc=718)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

In the face of the ever-increasing document volume, libraries around the globe are more and more exploring (semi-) automated approaches to subject indexing. This helps sustain bibliographic objectives, enrich metadata, and establish more connections across documents from various collections, effectively leading to improved information retrieval and access. However, generally accepted automated approaches that are functional in operative systems are lacking. This article aims to provide an overview of basic principles used for automated subject indexing, major approaches in relation to their possible application in actual library systems, existing working examples, as well as related challenges calling for further research.

0.0017084682 = product of:
  0.0051254043 = sum of:
    0.0051254043 = product of:
      0.010250809 = sum of:
        0.010250809 = weight(_text_:of in 63) [ClassicSimilarity], result of:
          0.010250809 = score(doc=63,freq=6.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.1496253 = fieldWeight in 63, product of:
              2.4494898 = tf(freq=6.0), with freq of:
                6.0 = termFreq=6.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0390625 = fieldNorm(doc=63)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Conventional rule-based approaches use exact template matching to capture linguistic information and necessarily need to enumerate all variations. We propose a novel flexible template generation and matching scheme called the principle-based approach (PBA) based on sequence alignment, and employ it for reference metadata extraction (RME) to demonstrate its effectiveness. The main contributions of this research are threefold. First, we propose an automatic template generation that can capture prominent patterns using the dominating set algorithm. Second, we devise an alignment-based template-matching technique that uses a logistic regression model, which makes it more general and flexible than pure rule-based approaches. Last, we apply PBA to RME on extensive cross-domain corpora and demonstrate its robustness and generality. Experiments reveal that the same set of templates produced by the PBA framework not only deliver consistent performance on various unseen domains, but also surpass hand-crafted knowledge (templates). We use four independent journal style test sets and one conference style test set in the experiments. When compared to renowned machine learning methods, such as conditional random fields (CRF), as well as recent deep learning methods (i.e., bi-directional long short-term memory with a CRF layer, Bi-LSTM-CRF), PBA has the best performance for all datasets.

Source

Journal of the Association for Information Science and Technology. 72(2021) no.1, S.32-45

0.0015782153 = product of:
  0.0047346456 = sum of:
    0.0047346456 = product of:
      0.009469291 = sum of:
        0.009469291 = weight(_text_:of in 722) [ClassicSimilarity], result of:
          0.009469291 = score(doc=722,freq=2.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.13821793 = fieldWeight in 722, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.0625 = fieldNorm(doc=722)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

This article provides a conclusion to the special issue on Artificial Intelligence (AI) and Automated Processes for Subject Access. The authors who contributed to this special issue have provoked interesting questions as well as bringing attention to important issues. This concluding article looks at common themes and highlights some of the questions raised.

7.8910764E-4 = product of:
  0.0023673228 = sum of:
    0.0023673228 = product of:
      0.0047346456 = sum of:
        0.0047346456 = weight(_text_:of in 935) [ClassicSimilarity], result of:
          0.0047346456 = score(doc=935,freq=2.0), product of:
            0.06850986 = queryWeight, product of:
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.043811057 = queryNorm
            0.06910896 = fieldWeight in 935, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              1.5637573 = idf(docFreq=25162, maxDocs=44218)
              0.03125 = fieldNorm(doc=935)
      0.5 = coord(1/2)
  0.33333334 = coord(1/3)

Abstract

Die ZBW - Leibniz-Informationszentrum Wirtschaft betreibt seit 2016 eigene angewandte Forschung im Bereich Machine Learning mit dem Zweck, praktikable Lösungen für eine automatisierte oder maschinell unterstützte Inhaltserschließung zu entwickeln. 2020 begann ein Team an der ZBW die Konzeption und Implementierung einer Softwarearchitektur, die es ermöglichte, diese prototypischen Lösungen in einen produktiven Dienst zu überführen und mit den bestehenden Nachweis- und Informationssystemen zu verzahnen. Sowohl die angewandte Forschung als auch die für dieses Vorhaben ("AutoSE") notwendige Softwareentwicklung sind direkt im Bibliotheksbereich der ZBW angesiedelt, werden kontinuierlich anhand des State of the Art vorangetrieben und profitieren von einem engen Austausch mit den Verantwortlichen für die intellektuelle Inhaltserschließung. Dieser Beitrag zeigt die Meilensteine auf, die das AutoSE-Team in zwei Jahren in Bezug auf den Aufbau und die Integration der Software erreicht hat, und skizziert, welche bis zum Ende der Pilotphase (2024) noch ausstehen. Die Architektur basiert auf Open-Source-Software und die eingesetzten Machine-Learning-Komponenten werden im Rahmen einer internationalen Zusammenarbeit im engen Austausch mit der Finnischen Nationalbibliothek (NLF) weiterentwickelt und zur Nachnutzung in dem von der NLF entwickelten Open-Source-Werkzeugkasten Annif aufbereitet. Das Betriebsmodell des AutoSE-Dienstes sieht regelmäßige Überprüfungen sowohl einzelner Komponenten als auch des Produktionsworkflows als Ganzes vor und erlaubt eine fortlaufende Weiterentwicklung der Architektur. Eines der Ergebnisse, das bis zum Ende der Pilotphase vorliegen soll, ist die Dokumentation der Anforderungen an einen dauerhaften produktiven Betrieb des Dienstes, damit die Ressourcen dafür im Rahmen eines tragfähigen Modells langfristig gesichert werden können. Aus diesem Praxisbeispiel lässt sich ableiten, welche Bedingungen gegeben sein müssen, um Machine-Learning-Lösungen wie die in Annif enthaltenen erfolgreich an einer Institution für die Inhaltserschließung einsetzen zu können.