Search (114 results, page 1 of 6)

0.14408328 = product of:
  0.28816655 = sum of:
    0.07204164 = product of:
      0.2161249 = sum of:
        0.2161249 = weight(_text_:3a in 862) [ClassicSimilarity], result of:
          0.2161249 = score(doc=862,freq=2.0), product of:
            0.38455155 = queryWeight, product of:
              8.478011 = idf(docFreq=24, maxDocs=44218)
              0.0453587 = queryNorm
            0.56201804 = fieldWeight in 862, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              8.478011 = idf(docFreq=24, maxDocs=44218)
              0.046875 = fieldNorm(doc=862)
      0.33333334 = coord(1/3)
    0.2161249 = weight(_text_:2f in 862) [ClassicSimilarity], result of:
      0.2161249 = score(doc=862,freq=2.0), product of:
        0.38455155 = queryWeight, product of:
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.0453587 = queryNorm
        0.56201804 = fieldWeight in 862, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.046875 = fieldNorm(doc=862)
  0.5 = coord(2/4)

Source

https%3A%2F%2Farxiv.org%2Fabs%2F2212.06721&usg=AOvVaw3i_9pZm9y_dQWoHi6uv0EN

0.12006941 = product of:
  0.24013881 = sum of:
    0.060034703 = product of:
      0.1801041 = sum of:
        0.1801041 = weight(_text_:3a in 1000) [ClassicSimilarity], result of:
          0.1801041 = score(doc=1000,freq=2.0), product of:
            0.38455155 = queryWeight, product of:
              8.478011 = idf(docFreq=24, maxDocs=44218)
              0.0453587 = queryNorm
            0.46834838 = fieldWeight in 1000, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              8.478011 = idf(docFreq=24, maxDocs=44218)
              0.0390625 = fieldNorm(doc=1000)
      0.33333334 = coord(1/3)
    0.1801041 = weight(_text_:2f in 1000) [ClassicSimilarity], result of:
      0.1801041 = score(doc=1000,freq=2.0), product of:
        0.38455155 = queryWeight, product of:
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.0453587 = queryNorm
        0.46834838 = fieldWeight in 1000, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.0390625 = fieldNorm(doc=1000)
  0.5 = coord(2/4)

Content

Master thesis Master of Science (Library and Information Studies) (MSc), Universität Wien. Advisor: Christoph Steiner. Vgl.: https://www.researchgate.net/publication/371680244_Vergabe_von_DDC-Sachgruppen_mittels_eines_Schlagwort-Thesaurus. DOI: 10.25365/thesis.70030. Vgl. dazu die Präsentation unter: https://www.google.com/url?sa=i&rct=j&q=&esrc=s&source=web&cd=&ved=0CAIQw7AJahcKEwjwoZzzytz_AhUAAAAAHQAAAAAQAg&url=https%3A%2F%2Fwiki.dnb.de%2Fdownload%2Fattachments%2F252121510%2FDA3%2520Workshop-Gabler.pdf%3Fversion%3D1%26modificationDate%3D1671093170000%26api%3Dv2&psig=AOvVaw0szwENK1or3HevgvIDOfjx&ust=1687719410889597&opi=89978449.

0.106243305 = product of:
  0.21248661 = sum of:
    0.19405018 = weight(_text_:assisted in 941) [ClassicSimilarity], result of:
      0.19405018 = score(doc=941,freq=4.0), product of:
        0.30640912 = queryWeight, product of:
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.0453587 = queryNorm
        0.6333042 = fieldWeight in 941, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.046875 = fieldNorm(doc=941)
    0.018436432 = product of:
      0.036872864 = sum of:
        0.036872864 = weight(_text_:22 in 941) [ClassicSimilarity], result of:
          0.036872864 = score(doc=941,freq=2.0), product of:
            0.15883844 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0453587 = queryNorm
            0.23214069 = fieldWeight in 941, 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=941)
      0.5 = coord(1/2)
  0.5 = coord(2/4)

Abstract

Inferring the gender of named entities present in a text has several practical applications in information sciences. Existing approaches toward name gender identification rely exclusively on using the gender distributions from labeled data. In the absence of such labeled data, these methods fail. In this article, we propose a two-stage model that is able to infer the gender of names present in text without requiring explicit name-gender labels. We use coreference resolution as the backbone for our proposed model. To aid coreference resolution where the existing contextual information does not suffice, we use a retrieval-assisted context aggregation framework. We demonstrate that state-of-the-art name gender inference is possible without supervision. Our proposed method matches or outperforms several supervised approaches and commercially used methods on five English language datasets from different domains.

Date

22. 3.2023 12:00:14

0.05659797 = product of:
  0.22639188 = sum of:
    0.22639188 = weight(_text_:assisted in 1139) [ClassicSimilarity], result of:
      0.22639188 = score(doc=1139,freq=4.0), product of:
        0.30640912 = queryWeight, product of:
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.0453587 = queryNorm
        0.7388549 = fieldWeight in 1139, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.0546875 = fieldNorm(doc=1139)
  0.25 = coord(1/4)

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.

0.05188354 = product of:
  0.10376708 = sum of:
    0.09147613 = weight(_text_:assisted in 181) [ClassicSimilarity], result of:
      0.09147613 = score(doc=181,freq=2.0), product of:
        0.30640912 = queryWeight, product of:
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.0453587 = queryNorm
        0.29854244 = fieldWeight in 181, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.03125 = fieldNorm(doc=181)
    0.0122909555 = product of:
      0.024581911 = sum of:
        0.024581911 = weight(_text_:22 in 181) [ClassicSimilarity], result of:
          0.024581911 = score(doc=181,freq=2.0), product of:
            0.15883844 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0453587 = queryNorm
            0.15476047 = fieldWeight in 181, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.03125 = fieldNorm(doc=181)
      0.5 = coord(1/2)
  0.5 = coord(2/4)

Abstract

Purpose This paper aims to explore the information needs and information-seeking behaviours of the new generation of engineering designers. A survey study is used to approach what their information needs are, how these needs change during an engineering design project and how their information-seeking behaviours have been influenced by the newly developed information technologies (ITs). Through an in-depth analysis of the survey results, the key functions have been identified for the next-generation management systems. Design/methodology/approach The paper first proposed four hypotheses on the information needs and information-seeking behaviours of young engineers. Then, a survey study was undertaken to understand their information usage in terms of the information needs and information-seeking behaviours during a complete engineering design process. Through analysing the survey results, several findings were obtained and on this basis, further comparisons were made to discuss and evaluate the hypotheses. Findings The paper has revealed that the engineering designers' information needs will evolve throughout the engineering design project; thus, they should be assisted at several different levels. Although they intend to search information and knowledge on know-what and know-how, what they really require is the know-why knowledge in order to help them complete design tasks. Also, the paper has shown how the newly developed ITs and web-based applications have influenced the engineers' information-seeking practices. Research limitations/implications The research subjects chosen in this study are engineering students in universities who, although not as experienced as engineers in companies, do go through a complete design process with the tasks similar to industrial scenarios. In addition, the focus of this study is to understand the information-seeking behaviours of a new generation of design engineers, so that the development of next-generation information and knowledge management systems can be well informed. In this sense, the results obtained do reveal some new knowledge about the information-seeking behaviours during a general design process. Practical implications This paper first identifies the information needs and information-seeking behaviours of the new generation of engineering designers. On this basis, the varied ways to meet these needs and behaviours are discussed and elaborated. This intends to provide the key characteristics for the development of the next-generation knowledge management system for engineering design projects. Originality/value This paper proposes a novel means of exploring the future engineers' information needs and information-seeking behaviours in a collaborative working environment. It also characterises the key features and functions for the next generation of knowledge management systems for engineering design.

Date

20. 1.2015 18:30:22

0.030677874 = product of:
  0.122711495 = sum of:
    0.122711495 = product of:
      0.24542299 = sum of:
        0.24542299 = weight(_text_:instructions in 5751) [ClassicSimilarity], result of:
          0.24542299 = score(doc=5751,freq=4.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.7692834 = fieldWeight in 5751, product of:
              2.0 = tf(freq=4.0), with freq of:
                4.0 = termFreq=4.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0546875 = fieldNorm(doc=5751)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

This article presents a case study of RIMMF, a software tool developed to improve the orientation and training of catalogers who use Resource Description and Access (RDA) to maintain bibliographic data. The cataloging guidance and instructions of RDA are based on the Functional Requirements conceptual models that are now consolidated in the IFLA Library Reference Model, but many catalogers are applying RDA in systems that have evolved from inventory and text-processing applications developed from older metadata paradigms. The article describes how RIMMF interacts with the RDA Toolkit and RDA Registry to offer cataloger-friendly multilingual data input and editing interfaces.

0.028681545 = product of:
  0.11472618 = sum of:
    0.11472618 = product of:
      0.34417853 = sum of:
        0.34417853 = weight(_text_:lcsh's in 1141) [ClassicSimilarity], result of:
          0.34417853 = score(doc=1141,freq=2.0), product of:
            0.4492837 = queryWeight, product of:
              9.905128 = idf(docFreq=5, maxDocs=44218)
              0.0453587 = queryNorm
            0.7660606 = fieldWeight in 1141, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              9.905128 = idf(docFreq=5, maxDocs=44218)
              0.0546875 = fieldNorm(doc=1141)
      0.33333334 = coord(1/3)
  0.25 = coord(1/4)

Abstract

Bibliographic records can include information from controlled vocabularies to capture identities about individuals, especially about authors or intended audiences; personal name authority records can also contain information about identity. Employing a systematic analysis of the overlap of the Homosaurus, Library of Congress Subject Headings (LCSH), and Library of Congress Demographic Group Terms (LCDGT), this article explores the extent to which LGBTQ+ identities are represented in the three vocabularies. Despite LCSH's long, iterative history of development and the faceted, post-coordinate nature of LCDGT, neither vocabulary was found to be adequate in covering the complex, LGBTQ+ identities represented in the Homosaurus.

0.02858629 = product of:
  0.11434516 = sum of:
    0.11434516 = weight(_text_:assisted in 468) [ClassicSimilarity], result of:
      0.11434516 = score(doc=468,freq=2.0), product of:
        0.30640912 = queryWeight, product of:
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.0453587 = queryNorm
        0.37317806 = fieldWeight in 468, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          6.7552447 = idf(docFreq=139, maxDocs=44218)
          0.0390625 = fieldNorm(doc=468)
  0.25 = coord(1/4)

Abstract

The rise in artificial intelligence and natural language processing techniques has increased considerably in the last few decades. Historically, the focus has been primarily on texts expressed in prose form, leaving mostly aside figurative or poetic expressions of language due to their rich semantics and syntactic complexity. The creation and analysis of poetry have been commonly carried out by hand, with a few computer-assisted approaches. In the Spanish context, the promise of machine learning is starting to pan out in specific tasks such as metrical annotation and syllabification. However, there is a task that remains unexplored and underdeveloped: stanza classification. This classification of the inner structures of verses in which a poem is built upon is an especially relevant task for poetry studies since it complements the structural information of a poem. In this work, we analyzed different computational approaches to stanza classification in the Spanish poetic tradition. These approaches show that this task continues to be hard for computers systems, both based on classical machine learning approaches as well as statistical language models and cannot compete with traditional computational paradigms based on the knowledge of experts.

0.024791464 = product of:
  0.09916586 = sum of:
    0.09916586 = product of:
      0.19833171 = sum of:
        0.19833171 = weight(_text_:instructions in 5858) [ClassicSimilarity], result of:
          0.19833171 = score(doc=5858,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.62167484 = fieldWeight in 5858, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0625 = fieldNorm(doc=5858)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

In this article we report on an experiment that tested how useful library-based guidelines are for measuring the integrity of information in the era of fake news. We found that the usefulness of these guidelines depends on at least three factors: weighting indicators (criteria), clear instructions, and context-specificity.

0.024791464 = product of:
  0.09916586 = sum of:
    0.09916586 = product of:
      0.19833171 = sum of:
        0.19833171 = weight(_text_:instructions in 113) [ClassicSimilarity], result of:
          0.19833171 = score(doc=113,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.62167484 = fieldWeight in 113, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0625 = fieldNorm(doc=113)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

In this article we report on an experiment that tested how useful library-based guidelines are for measuring the integrity of information in the era of fake news. We found that the usefulness of these guidelines depends on at least three factors: weighting indicators (criteria), clear instructions, and context-specificity.

0.02169253 = product of:
  0.08677012 = sum of:
    0.08677012 = product of:
      0.17354023 = sum of:
        0.17354023 = weight(_text_:instructions in 707) [ClassicSimilarity], result of:
          0.17354023 = score(doc=707,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.54396546 = fieldWeight in 707, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0546875 = fieldNorm(doc=707)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

This paper discusses the basics of creating name authority records for Arabic personal names in accordance with Resource Description and Access instructions and Program for Cooperative Cataloging guidelines. A background into the use of romanization for non-Latin scripts in bibliographic and authority records is provided to establish the context. Issues with romanization that are particular to Arabic are addressed. Separate sections on modern and classical names provide an overview of the major challenges, and strategies to enhance discovery are outlined. The paper concludes with an examination of the possible benefits of identity management and other changes in the authority control landscape for names in non-Latin script.

0.02169253 = product of:
  0.08677012 = sum of:
    0.08677012 = product of:
      0.17354023 = sum of:
        0.17354023 = weight(_text_:instructions in 1159) [ClassicSimilarity], result of:
          0.17354023 = score(doc=1159,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.54396546 = fieldWeight in 1159, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0546875 = fieldNorm(doc=1159)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

This study examines users' needs for bibliographic enhancement data (BIBED) in academic library catalogs. Qualitative data were collected through 30 academic users' activity logs and follow-up interviews. These 30 participants were recruited from a public university in the United States that has over 19,000 students enrolled and over 600 full-time faculty members. This study identified 19 types of BIBED useful for supporting the five user tasks proposed in the IFLA Library Reference Model and in seven other contexts, such as enhancing one's understanding, offering search instructions, and providing readers' advisory. Findings suggest that adopting BIBFRAME and Semantic Web technologies may enable academic library catalogs to provide BIBED to better meet user needs in various contexts.

0.018593598 = product of:
  0.07437439 = sum of:
    0.07437439 = product of:
      0.14874879 = sum of:
        0.14874879 = weight(_text_:instructions in 706) [ClassicSimilarity], result of:
          0.14874879 = score(doc=706,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.46625614 = fieldWeight in 706, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.046875 = fieldNorm(doc=706)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

Descriptive cataloging is the process of representing resources by recording their identifying traits and selecting specific names and titles to serve as access points. It is a key component of the larger cataloging process alongside subject cataloging, authority work, and encoding. Descriptive cataloging practices have existed for centuries and, over time, have become standardized through the use of cataloging codes. These documents guide this process by prescribing a consistent set of elements, providing directions on how to record these elements, and offering instructions on how to select and format access points. The goal of descriptive cataloging is not to create perfect representations but to provide data to serve users. The international cataloging standard Resource Description and Access (RDA) is now bringing more institutions under the same set of descriptive practices than ever before. This, along with recent technological developments, promises increased sharing and reuse of descriptive cataloging data.

0.015008676 = product of:
  0.060034703 = sum of:
    0.060034703 = product of:
      0.1801041 = sum of:
        0.1801041 = weight(_text_:3a in 5669) [ClassicSimilarity], result of:
          0.1801041 = score(doc=5669,freq=2.0), product of:
            0.38455155 = queryWeight, product of:
              8.478011 = idf(docFreq=24, maxDocs=44218)
              0.0453587 = queryNorm
            0.46834838 = fieldWeight in 5669, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              8.478011 = idf(docFreq=24, maxDocs=44218)
              0.0390625 = fieldNorm(doc=5669)
      0.33333334 = coord(1/3)
  0.25 = coord(1/4)

Content

"Die Englischsprachige Wikipedia verfügt jetzt über mehr als 6 Millionen Artikel. An zweiter Stelle kommt die deutschsprachige Wikipedia mit 2.3 Millionen Artikeln, an dritter Stelle steht die französischsprachige Wikipedia mit 2.1 Millionen Artikeln (via Researchbuzz: Firehose <https://rbfirehose.com/2020/01/24/techcrunch-wikipedia-now-has-more-than-6-million-articles-in-english/> und Techcrunch <https://techcrunch.com/2020/01/23/wikipedia-english-six-million-articles/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+Techcrunch+%28TechCrunch%29&guccounter=1&guce_referrer=aHR0cHM6Ly9yYmZpcmVob3NlLmNvbS8yMDIwLzAxLzI0L3RlY2hjcnVuY2gtd2lraXBlZGlhLW5vdy1oYXMtbW9yZS10aGFuLTYtbWlsbGlvbi1hcnRpY2xlcy1pbi1lbmdsaXNoLw&guce_referrer_sig=AQAAAK0zHfjdDZ_spFZBF_z-zDjtL5iWvuKDumFTzm4HvQzkUfE2pLXQzGS6FGB_y-VISdMEsUSvkNsg2U_NWQ4lwWSvOo3jvXo1I3GtgHpP8exukVxYAnn5mJspqX50VHIWFADHhs5AerkRn3hMRtf_R3F1qmEbo8EROZXp328HMC-o>). 250120 via digithek ch = #fineBlog s.a.: Angesichts der Veröffentlichung des 6-millionsten Artikels vergangene Woche in der englischsprachigen Wikipedia hat die Community-Zeitungsseite "Wikipedia Signpost" ein Moratorium bei der Veröffentlichung von Unternehmensartikeln gefordert. Das sei kein Vorwurf gegen die Wikimedia Foundation, aber die derzeitigen Maßnahmen, um die Enzyklopädie gegen missbräuchliches undeklariertes Paid Editing zu schützen, funktionierten ganz klar nicht. *"Da die ehrenamtlichen Autoren derzeit von Werbung in Gestalt von Wikipedia-Artikeln überwältigt werden, und da die WMF nicht in der Lage zu sein scheint, dem irgendetwas entgegenzusetzen, wäre der einzige gangbare Weg für die Autoren, fürs erste die Neuanlage von Artikeln über Unternehmen zu untersagen"*, schreibt der Benutzer Smallbones in seinem Editorial <https://en.wikipedia.org/wiki/Wikipedia:Wikipedia_Signpost/2020-01-27/From_the_editor> zur heutigen Ausgabe."

0.012395732 = product of:
  0.04958293 = sum of:
    0.04958293 = product of:
      0.09916586 = sum of:
        0.09916586 = weight(_text_:instructions in 872) [ClassicSimilarity], result of:
          0.09916586 = score(doc=872,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.31083742 = fieldWeight in 872, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.03125 = fieldNorm(doc=872)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

0.010846265 = product of:
  0.04338506 = sum of:
    0.04338506 = product of:
      0.08677012 = sum of:
        0.08677012 = weight(_text_:instructions in 923) [ClassicSimilarity], result of:
          0.08677012 = score(doc=923,freq=2.0), product of:
            0.31902805 = queryWeight, product of:
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.0453587 = queryNorm
            0.27198273 = fieldWeight in 923, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              7.033448 = idf(docFreq=105, maxDocs=44218)
              0.02734375 = fieldNorm(doc=923)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Abstract

They might appear intelligent, but LLMs are nothing of the sort. They don't understand the meanings of the words they are using, nor the concepts expressed within the sentences they create. When asked how to bring a cow back to life, earlier versions of ChatGPT, for example, which ran on a souped-up version of GPT-3, would confidently provide a list of instructions. So-called hallucinations like this happen because language models have no concept of what a "cow" is or that "death" is a non-reversible state of being. LLMs do not have minds that can think about objects in the world and how they relate to each other. All they "know" is how likely it is that some sets of words will follow other sets of words, having calculated those probabilities from their training data. To make sense of all this, I spoke with Gary Marcus, an emeritus professor of psychology and neural science at New York University, for "Babbage", our science and technology podcast. Last year, as the world was transfixed by the sudden appearance of ChatGPT, he made some fascinating predictions about GPT-4.

0.010754585 = product of:
  0.04301834 = sum of:
    0.04301834 = product of:
      0.08603668 = sum of:
        0.08603668 = weight(_text_:22 in 1079) [ClassicSimilarity], result of:
          0.08603668 = score(doc=1079,freq=2.0), product of:
            0.15883844 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0453587 = queryNorm
            0.5416616 = fieldWeight in 1079, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.109375 = fieldNorm(doc=1079)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Date

27. 1.2023 16:22:55

0.009218216 = product of:
  0.036872864 = sum of:
    0.036872864 = product of:
      0.07374573 = sum of:
        0.07374573 = weight(_text_:22 in 4156) [ClassicSimilarity], result of:
          0.07374573 = score(doc=4156,freq=2.0), product of:
            0.15883844 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0453587 = queryNorm
            0.46428138 = fieldWeight in 4156, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.09375 = fieldNorm(doc=4156)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Date

2. 3.2020 14:08:22

0.009218216 = product of:
  0.036872864 = sum of:
    0.036872864 = product of:
      0.07374573 = sum of:
        0.07374573 = weight(_text_:22 in 1203) [ClassicSimilarity], result of:
          0.07374573 = score(doc=1203,freq=2.0), product of:
            0.15883844 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0453587 = queryNorm
            0.46428138 = fieldWeight in 1203, product of:
              1.4142135 = tf(freq=2.0), with freq of:
                2.0 = termFreq=2.0
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.09375 = fieldNorm(doc=1203)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Source

Spektrum der Wissenschaft. 2023, H.10, S.22-24

0.007681847 = product of:
  0.030727388 = sum of:
    0.030727388 = product of:
      0.061454777 = sum of:
        0.061454777 = weight(_text_:22 in 5723) [ClassicSimilarity], result of:
          0.061454777 = score(doc=5723,freq=2.0), product of:
            0.15883844 = queryWeight, product of:
              3.5018296 = idf(docFreq=3622, maxDocs=44218)
              0.0453587 = queryNorm
            0.38690117 = fieldWeight in 5723, 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=5723)
      0.5 = coord(1/2)
  0.25 = coord(1/4)

Date

17. 1.2020 22:15:11