Search (76 results, page 1 of 4)

0.1257631 = product of:
  0.5030524 = sum of:
    0.5030524 = weight(_text_:3a in 1826) [ClassicSimilarity], result of:
      0.5030524 = score(doc=1826,freq=2.0), product of:
        0.53704935 = queryWeight, product of:
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.06334615 = queryNorm
        0.93669677 = fieldWeight in 1826, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.078125 = fieldNorm(doc=1826)
  0.25 = coord(1/4)

Source

http://www.google.de/url?sa=t&rct=j&q=&esrc=s&source=web&cd=5&ved=0CDQQFjAE&url=http%3A%2F%2Fdigbib.ubka.uni-karlsruhe.de%2Fvolltexte%2Fdocuments%2F3131107&ei=HzFWVYvGMsiNsgGTyoFI&usg=AFQjCNE2FHUeR9oQTQlNC4TPedv4Mo3DaQ&sig2=Rlzpr7a3BLZZkqZCXXN_IA&bvm=bv.93564037,d.bGg&cad=rja

0.06288155 = product of:
  0.2515262 = sum of:
    0.2515262 = weight(_text_:3a in 4388) [ClassicSimilarity], result of:
      0.2515262 = score(doc=4388,freq=2.0), product of:
        0.53704935 = queryWeight, product of:
          8.478011 = idf(docFreq=24, maxDocs=44218)
          0.06334615 = queryNorm
        0.46834838 = fieldWeight in 4388, 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=4388)
  0.25 = coord(1/4)

Footnote

Vgl. unter: https://www.google.de/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=2ahUKEwizweHljdbcAhVS16QKHXcFD9QQFjABegQICRAB&url=https%3A%2F%2Fwww.researchgate.net%2Fpublication%2F271200105_Die_Autonomie_des_Menschen_und_der_Maschine_-_gegenwartige_Definitionen_von_Autonomie_zwischen_philosophischem_Hintergrund_und_technologischer_Umsetzbarkeit_Redigierte_Version_der_Magisterarbeit_Karls&usg=AOvVaw06orrdJmFF2xbCCp_hL26q.

0.05931501 = product of:
  0.23726004 = sum of:
    0.23726004 = weight(_text_:objects in 469) [ClassicSimilarity], result of:
      0.23726004 = score(doc=469,freq=8.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.7046855 = fieldWeight in 469, product of:
          2.828427 = tf(freq=8.0), with freq of:
            8.0 = termFreq=8.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.046875 = fieldNorm(doc=469)
  0.25 = coord(1/4)

Abstract

Current digital preservation is strongly biased towards data objects: digital files of document-style objects, or encapsulated and largely self-contained objects. To provide authenticity and provenance information, comprehensive metadata models are deployed to document information on an object's context. Yet, we claim that simply documenting an objects context may not be sufficient to ensure proper provenance and to fulfill the stated preservation goals. Specifically in e-Science and business settings, capturing, documenting and preserving entire processes may be necessary to meet the preservation goals. We thus present an approach for capturing, documenting and preserving processes, and means to assess their authenticity upon re-execution. We will discuss options as well as limitations and open challenges to achieve sound preservation, speci?cally within scientific processes.

0.042806923 = product of:
  0.1712277 = sum of:
    0.1712277 = weight(_text_:objects in 3372) [ClassicSimilarity], result of:
      0.1712277 = score(doc=3372,freq=6.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.508563 = fieldWeight in 3372, product of:
          2.4494898 = tf(freq=6.0), with freq of:
            6.0 = termFreq=6.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.0390625 = fieldNorm(doc=3372)
  0.25 = coord(1/4)

Abstract

Europeana provides access to more than 54 million cultural heritage objects through its portal Europeana Collections. It is crucial for Europeana to be recognized by search engines as a trusted authoritative repository of cultural heritage objects. Indeed, even though its portal is the main entry point, most Europeana users come to it via search engines. Europeana Collections is fuelled by metadata describing cultural objects, represented in the Europeana Data Model (EDM). This paper presents the research and consequent recommendations for publishing Europeana metadata using the Schema.org vocabulary and best practices. Schema.org html embedded metadata to be consumed by search engines to power rich services (such as Google Knowledge Graph). Schema.org is an open and widely adopted initiative (used by over 12 million domains) backed by Google, Bing, Yahoo!, and Yandex, for sharing metadata across the web It underpins the emergence of new web techniques, such as so called Semantic SEO. Our research addressed the representation of the embedded metadata as part of the Europeana HTML pages and sitemaps so that the re-use of this data can be optimized. The practical objective of our work is to produce a Schema.org representation of Europeana resources described in EDM, being the richest as possible and tailored to Europeana's realities and user needs as well the search engines and their users.

0.04194205 = product of:
  0.1677682 = sum of:
    0.1677682 = weight(_text_:objects in 3884) [ClassicSimilarity], result of:
      0.1677682 = score(doc=3884,freq=4.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.49828792 = fieldWeight in 3884, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.046875 = fieldNorm(doc=3884)
  0.25 = coord(1/4)

Abstract

Techniques for multidimensional scaling visualize objects as points in a low-dimensional metric map. As a result, the visualizations are subject to the fundamental limitations of metric spaces. These limitations prevent multidimensional scaling from faithfully representing non-metric similarity data such as word associations or event co-occurrences. In particular, multidimensional scaling cannot faithfully represent intransitive pairwise similarities in a visualization, and it cannot faithfully visualize "central" objects. In this paper, we present an extension of a recently proposed multidimensional scaling technique called t-SNE. The extension aims to address the problems of traditional multidimensional scaling techniques when these techniques are used to visualize non-metric similarities. The new technique, called multiple maps t-SNE, alleviates these problems by constructing a collection of maps that reveal complementary structure in the similarity data. We apply multiple maps t-SNE to a large data set of word association data and to a data set of NIPS co-authorships, demonstrating its ability to successfully visualize non-metric similarities.

0.03954334 = product of:
  0.15817337 = sum of:
    0.15817337 = weight(_text_:objects in 1700) [ClassicSimilarity], result of:
      0.15817337 = score(doc=1700,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.46979034 = fieldWeight in 1700, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.0625 = fieldNorm(doc=1700)
  0.25 = coord(1/4)

Abstract

The search giant is automatically building Knowledge Vault, a massive database that could give us unprecedented access to the world's facts GOOGLE is building the largest store of knowledge in human history - and it's doing so without any human help. Instead, Knowledge Vault autonomously gathers and merges information from across the web into a single base of facts about the world, and the people and objects in it.

0.034951705 = product of:
  0.13980682 = sum of:
    0.13980682 = weight(_text_:objects in 470) [ClassicSimilarity], result of:
      0.13980682 = score(doc=470,freq=4.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.41523993 = fieldWeight in 470, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.0390625 = fieldNorm(doc=470)
  0.25 = coord(1/4)

Abstract

With the rapidly increasing pace at which Web content is evolving, particularly social media, preserving the Web and its evolution over time becomes an important challenge. Meaningful analysis of Web content lends itself to an entity-centric view to organise Web resources according to the information objects related to them. Therefore, the crucial challenge is to extract, detect and correlate entities from a vast number of heterogeneous Web resources where the nature and quality of the content may vary heavily. While a wealth of information extraction tools aid this process, we believe that, the consolidation of automatically extracted data has to be treated as an equally important step in order to ensure high quality and non-ambiguity of generated data. In this paper we present an approach which is based on an iterative cycle exploiting Web data for (1) targeted archiving/crawling of Web objects, (2) entity extraction, and detection, and (3) entity correlation. The long-term goal is to preserve Web content over time and allow its navigation and analysis based on well-formed structured RDF data about entities.

0.03460042 = product of:
  0.13840169 = sum of:
    0.13840169 = weight(_text_:objects in 4836) [ClassicSimilarity], result of:
      0.13840169 = score(doc=4836,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.41106653 = fieldWeight in 4836, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.0546875 = fieldNorm(doc=4836)
  0.25 = coord(1/4)

Abstract

We present the DOCMA method (DOCument and Model for Action) focused to Socio-Semantic web applications in large communities of interest. DOCMA is dedicated to end-users without any knowledge in Information Science. Community Members can elicit, structure and index shared business items emerging from their inquiry (such as projects, actors, products, geographically situated objects of interest.). We apply DOCMA to an experiment in the field of Sustainable Development: the Cartodd-Map21 collaborative Web portal.

0.03460042 = product of:
  0.13840169 = sum of:
    0.13840169 = weight(_text_:objects in 3886) [ClassicSimilarity], result of:
      0.13840169 = score(doc=3886,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.41106653 = fieldWeight in 3886, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.0546875 = fieldNorm(doc=3886)
  0.25 = coord(1/4)

Abstract

The paper investigates the acceleration of t-SNE-an embedding technique that is commonly used for the visualization of high-dimensional data in scatter plots-using two tree-based algorithms. In particular, the paper develops variants of the Barnes-Hut algorithm and of the dual-tree algorithm that approximate the gradient used for learning t-SNE embeddings in O(N*logN). Our experiments show that the resulting algorithms substantially accelerate t-SNE, and that they make it possible to learn embeddings of data sets with millions of objects. Somewhat counterintuitively, the Barnes-Hut variant of t-SNE appears to outperform the dual-tree variant.

0.030343821 = product of:
  0.121375285 = sum of:
    0.121375285 = weight(_text_:22 in 3582) [ClassicSimilarity], result of:
      0.121375285 = score(doc=3582,freq=4.0), product of:
        0.22182742 = queryWeight, product of:
          3.5018296 = idf(docFreq=3622, maxDocs=44218)
          0.06334615 = queryNorm
        0.54716086 = fieldWeight in 3582, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.5018296 = idf(docFreq=3622, maxDocs=44218)
          0.078125 = fieldNorm(doc=3582)
  0.25 = coord(1/4)

Date

22. 4.2017 10:42:05
22. 4.2017 10:48:38

0.03003885 = product of:
  0.1201554 = sum of:
    0.1201554 = weight(_text_:22 in 8365) [ClassicSimilarity], result of:
      0.1201554 = score(doc=8365,freq=2.0), product of:
        0.22182742 = queryWeight, product of:
          3.5018296 = idf(docFreq=3622, maxDocs=44218)
          0.06334615 = queryNorm
        0.5416616 = fieldWeight in 8365, 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=8365)
  0.25 = coord(1/4)

Date

22. 6.2015 16:08:38

0.029964847 = product of:
  0.11985939 = sum of:
    0.11985939 = weight(_text_:objects in 1874) [ClassicSimilarity], result of:
      0.11985939 = score(doc=1874,freq=6.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.3559941 = fieldWeight in 1874, product of:
          2.4494898 = tf(freq=6.0), with freq of:
            6.0 = termFreq=6.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.02734375 = fieldNorm(doc=1874)
  0.25 = coord(1/4)

Content

"People can summarize a complex scene in a few words without thinking twice. It's much more difficult for computers. But we've just gotten a bit closer -- we've developed a machine-learning system that can automatically produce captions (like the three above) to accurately describe images the first time it sees them. This kind of system could eventually help visually impaired people understand pictures, provide alternate text for images in parts of the world where mobile connections are slow, and make it easier for everyone to search on Google for images. Recent research has greatly improved object detection, classification, and labeling. But accurately describing a complex scene requires a deeper representation of what's going on in the scene, capturing how the various objects relate to one another and translating it all into natural-sounding language. Many efforts to construct computer-generated natural descriptions of images propose combining current state-of-the-art techniques in both computer vision and natural language processing to form a complete image description approach. But what if we instead merged recent computer vision and language models into a single jointly trained system, taking an image and directly producing a human readable sequence of words to describe it? This idea comes from recent advances in machine translation between languages, where a Recurrent Neural Network (RNN) transforms, say, a French sentence into a vector representation, and a second RNN uses that vector representation to generate a target sentence in German. Now, what if we replaced that first RNN and its input words with a deep Convolutional Neural Network (CNN) trained to classify objects in images? Normally, the CNN's last layer is used in a final Softmax among known classes of objects, assigning a probability that each object might be in the image. But if we remove that final layer, we can instead feed the CNN's rich encoding of the image into a RNN designed to produce phrases. We can then train the whole system directly on images and their captions, so it maximizes the likelihood that descriptions it produces best match the training descriptions for each image.

0.029657505 = product of:
  0.11863002 = sum of:
    0.11863002 = weight(_text_:objects in 2144) [ClassicSimilarity], result of:
      0.11863002 = score(doc=2144,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.35234275 = fieldWeight in 2144, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.046875 = fieldNorm(doc=2144)
  0.25 = coord(1/4)

Abstract

The complexity and diversity of archival resources make constructing rich metadata records time consuming and expensive, which in turn limits access to these valuable materials. However, significant automation of the metadata creation process would dramatically reduce the cost of providing access points, improve access to individual resources, and establish connections between resources that would otherwise remain unknown. Using a case study at Oregon Health & Science University as a lens to examine the conceptual and technical challenges associated with automated extraction of access points, we discuss using publically accessible API's to extract entities (i.e. people, places, concepts, etc.) from digital and digitized objects. We describe why Linked Open Data is not well suited for a use case such as ours. We conclude with recommendations about how this method can be used in archives as well as for other library applications.

0.029657505 = product of:
  0.11863002 = sum of:
    0.11863002 = weight(_text_:objects in 3893) [ClassicSimilarity], result of:
      0.11863002 = score(doc=3893,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.35234275 = fieldWeight in 3893, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.046875 = fieldNorm(doc=3893)
  0.25 = coord(1/4)

Abstract

In 2017, the University of Virginia (U.Va.) will launch a two year initiative to celebrate the bicentennial anniversary of the University's founding in 1819. The U.Va. Library is participating in this event by digitizing some 20,000 photographs and negatives that document student life on the U.Va. grounds in the 1960s and 1970s. Metadata librarians and archivists are well-versed in the challenges associated with generating digital content and accompanying description within the context of limited resources. This paper describes how technology and new approaches to metadata design have enabled the University of Virginia's Metadata Analysis and Design Department to rapidly and successfully generate accurate description for these digital objects. Python's pandas module improves efficiency by cleaning and repurposing data recorded at digitization, while the lxml module builds MODS XML programmatically from CSV tables. A simplified technique for subject heading selection and assignment in Google Sheets provides a collaborative environment for streamlined metadata creation and data quality control.

0.029657505 = product of:
  0.11863002 = sum of:
    0.11863002 = weight(_text_:objects in 3896) [ClassicSimilarity], result of:
      0.11863002 = score(doc=3896,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.35234275 = fieldWeight in 3896, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.046875 = fieldNorm(doc=3896)
  0.25 = coord(1/4)

Abstract

The Avalon Media System (Avalon) provides access and management for digital audio and video collections in libraries and archives. The open source project is led by the libraries of Indiana University Bloomington and Northwestern University and is funded in part by grants from The Andrew W. Mellon Foundation and Institute of Museum and Library Services. Avalon is based on the Samvera Community (formerly Hydra Project) software stack and uses Fedora as the digital repository back end. The Avalon project team is in the process of migrating digital repositories from Fedora 3 to Fedora 4 and incorporating metadata statements using the Resource Description Framework (RDF) instead of XML files accompanying the digital objects in the repository. The Avalon team has worked on the migration path for technical metadata and is now working on the migration paths for structural metadata (PCDM) and descriptive metadata (from MODS XML to RDF). This paper covers the decisions made to begin using RDF for software development and offers a window into how Semantic Web technology functions in the real world.

0.027961366 = product of:
  0.11184546 = sum of:
    0.11184546 = weight(_text_:objects in 4449) [ClassicSimilarity], result of:
      0.11184546 = score(doc=4449,freq=4.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.33219194 = fieldWeight in 4449, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.03125 = fieldNorm(doc=4449)
  0.25 = coord(1/4)

Abstract

In a new study, computer scientists found that artificial intelligence systems fail a vision test a child could accomplish with ease. "It's a clever and important study that reminds us that 'deep learning' isn't really that deep," said Gary Marcus , a neuroscientist at New York University who was not affiliated with the work. The result takes place in the field of computer vision, where artificial intelligence systems attempt to detect and categorize objects. They might try to find all the pedestrians in a street scene, or just distinguish a bird from a bicycle (which is a notoriously difficult task). The stakes are high: As computers take over critical tasks like automated surveillance and autonomous driving, we'll want their visual processing to be at least as good as the human eyes they're replacing. It won't be easy. The new work accentuates the sophistication of human vision - and the challenge of building systems that mimic it. In the study, the researchers presented a computer vision system with a living room scene. The system processed it well. It correctly identified a chair, a person, books on a shelf. Then the researchers introduced an anomalous object into the scene - an image of elephant. The elephant's mere presence caused the system to forget itself: Suddenly it started calling a chair a couch and the elephant a chair, while turning completely blind to other objects it had previously seen. Researchers are still trying to understand exactly why computer vision systems get tripped up so easily, but they have a good guess. It has to do with an ability humans have that AI lacks: the ability to understand when a scene is confusing and thus go back for a second glance.

0.025747584 = product of:
  0.10299034 = sum of:
    0.10299034 = weight(_text_:22 in 1552) [ClassicSimilarity], result of:
      0.10299034 = score(doc=1552,freq=2.0), product of:
        0.22182742 = queryWeight, product of:
          3.5018296 = idf(docFreq=3622, maxDocs=44218)
          0.06334615 = queryNorm
        0.46428138 = fieldWeight in 1552, 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=1552)
  0.25 = coord(1/4)

Date

22. 6.2018 11:04:35

0.02471459 = product of:
  0.09885836 = sum of:
    0.09885836 = weight(_text_:objects in 704) [ClassicSimilarity], result of:
      0.09885836 = score(doc=704,freq=2.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.29361898 = fieldWeight in 704, product of:
          1.4142135 = tf(freq=2.0), with freq of:
            2.0 = termFreq=2.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.0390625 = fieldNorm(doc=704)
  0.25 = coord(1/4)

Abstract

We present Broccoli, a fast and easy-to-use search engine forwhat we call semantic full-text search. Semantic full-textsearch combines the capabilities of standard full-text searchand ontology search. The search operates on four kinds ofobjects: ordinary words (e.g., edible), classes (e.g., plants), instances (e.g.,Broccoli), and relations (e.g., occurs-with or native-to). Queries are trees, where nodes are arbitrary bags of these objects, and arcs are relations. The user interface guides the user in incrementally constructing such trees by instant (search-as-you-type) suggestions of words, classes, instances, or relations that lead to good hits. Both standard full-text search and pure ontology search are included as special cases. In this paper, we describe the query language of Broccoli, a new kind of index that enables fast processing of queries from that language as well as fast query suggestion, the natural language processing required, and the user interface. We evaluated query times and result quality on the full version of the English Wikipedia (32 GB XML dump) combined with the YAGO ontology (26 million facts). We have implemented a fully functional prototype based on our ideas, see: http://broccoli.informatik.uni-freiburg.de.

0.02471459 = product of:
  0.09885836 = sum of:
    0.09885836 = weight(_text_:objects in 1875) [ClassicSimilarity], result of:
      0.09885836 = score(doc=1875,freq=8.0), product of:
        0.33668926 = queryWeight, product of:
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.06334615 = queryNorm
        0.29361898 = fieldWeight in 1875, product of:
          2.828427 = tf(freq=8.0), with freq of:
            8.0 = termFreq=8.0
          5.315071 = idf(docFreq=590, maxDocs=44218)
          0.01953125 = fieldNorm(doc=1875)
  0.25 = coord(1/4)

Content

"Until now, so-called computer vision has largely been limited to recognizing individual objects. The new software, described on Monday by researchers at Google and at Stanford University, teaches itself to identify entire scenes: a group of young men playing Frisbee, for example, or a herd of elephants marching on a grassy plain. The software then writes a caption in English describing the picture. Compared with human observations, the researchers found, the computer-written descriptions are surprisingly accurate. The advances may make it possible to better catalog and search for the billions of images and hours of video available online, which are often poorly described and archived. At the moment, search engines like Google rely largely on written language accompanying an image or video to ascertain what it contains. "I consider the pixel data in images and video to be the dark matter of the Internet," said Fei-Fei Li, director of the Stanford Artificial Intelligence Laboratory, who led the research with Andrej Karpathy, a graduate student. "We are now starting to illuminate it." Dr. Li and Mr. Karpathy published their research as a Stanford University technical report. The Google team published their paper on arXiv.org, an open source site hosted by Cornell University.
In the longer term, the new research may lead to technology that helps the blind and robots navigate natural environments. But it also raises chilling possibilities for surveillance. During the past 15 years, video cameras have been placed in a vast number of public and private spaces. In the future, the software operating the cameras will not only be able to identify particular humans via facial recognition, experts say, but also identify certain types of behavior, perhaps even automatically alerting authorities. Two years ago Google researchers created image-recognition software and presented it with 10 million images taken from YouTube videos. Without human guidance, the program trained itself to recognize cats - a testament to the number of cat videos on YouTube. Current artificial intelligence programs in new cars already can identify pedestrians and bicyclists from cameras positioned atop the windshield and can stop the car automatically if the driver does not take action to avoid a collision. But "just single object recognition is not very beneficial," said Ali Farhadi, a computer scientist at the University of Washington who has published research on software that generates sentences from digital pictures. "We've focused on objects, and we've ignored verbs," he said, adding that these programs do not grasp what is going on in an image. Both the Google and Stanford groups tackled the problem by refining software programs known as neural networks, inspired by our understanding of how the brain works. Neural networks can "train" themselves to discover similarities and patterns in data, even when their human creators do not know the patterns exist.
In living organisms, webs of neurons in the brain vastly outperform even the best computer-based networks in perception and pattern recognition. But by adopting some of the same architecture, computers are catching up, learning to identify patterns in speech and imagery with increasing accuracy. The advances are apparent to consumers who use Apple's Siri personal assistant, for example, or Google's image search. Both groups of researchers employed similar approaches, weaving together two types of neural networks, one focused on recognizing images and the other on human language. In both cases the researchers trained the software with relatively small sets of digital images that had been annotated with descriptive sentences by humans. After the software programs "learned" to see patterns in the pictures and description, the researchers turned them on previously unseen images. The programs were able to identify objects and actions with roughly double the accuracy of earlier efforts, although still nowhere near human perception capabilities. "I was amazed that even with the small amount of training data that we were able to do so well," said Oriol Vinyals, a Google computer scientist who wrote the paper with Alexander Toshev, Samy Bengio and Dumitru Erhan, members of the Google Brain project. "The field is just starting, and we will see a lot of increases."
Computer vision specialists said that despite the improvements, these software systems had made only limited progress toward the goal of digitally duplicating human vision and, even more elusive, understanding. "I don't know that I would say this is 'understanding' in the sense we want," said John R. Smith, a senior manager at I.B.M.'s T.J. Watson Research Center in Yorktown Heights, N.Y. "I think even the ability to generate language here is very limited." But the Google and Stanford teams said that they expect to see significant increases in accuracy as they improve their software and train these programs with larger sets of annotated images. A research group led by Tamara L. Berg, a computer scientist at the University of North Carolina at Chapel Hill, is training a neural network with one million images annotated by humans. "You're trying to tell the story behind the image," she said. "A natural scene will be very complex, and you want to pick out the most important objects in the image.""

0.024275057 = product of:
  0.09710023 = sum of:
    0.09710023 = weight(_text_:22 in 4313) [ClassicSimilarity], result of:
      0.09710023 = score(doc=4313,freq=4.0), product of:
        0.22182742 = queryWeight, product of:
          3.5018296 = idf(docFreq=3622, maxDocs=44218)
          0.06334615 = queryNorm
        0.4377287 = fieldWeight in 4313, product of:
          2.0 = tf(freq=4.0), with freq of:
            4.0 = termFreq=4.0
          3.5018296 = idf(docFreq=3622, maxDocs=44218)
          0.0625 = fieldNorm(doc=4313)
  0.25 = coord(1/4)

Date

22. 6.2018 14:22:02