Document (#41109)

Author
Wolfangel, E.
Title
¬Die Grenzen der künstlichen Intelligenz
Issue
[03.05.2016].
Source
http://www.spektrum.de/news/die-grenzen-der-kuenstlichen-intelligenz/1409149
Year
2016
Series
Algorithmen
Abstract
Der Sieg einer Google-Software über den Weltmeister im Brettspiel Go hat der Technologie der künstlichen Intelligenz gesellschaftlichen Auftrieb verschafft. Die Mühen der Ebene zeigen aber, dass die Algorithmen des maschinellen Lernens allerlei Fallen bergen - auch für ihre Entwickler.
Content
"Wie lernen Maschinen überhaupt? Eine zentrale Unterscheidung betrifft die Art des Lernens: Algorithmen können überwacht oder unüberwacht lernen. Ersteres wird unter anderem für Klassifikationsaufgaben genutzt: Ist beispielsweise ein Mensch auf einem Foto oder nicht? Grundlage dafür sind Trainingsdaten, anhand derer der Algorithmus auf Vorgabe eines Menschen lernt, was das richtige Ergebnis ist - auf diesen 1000 Bildern ist ein Mensch, auf diesen 1000 nicht. Hat das System für alle eventuell vorkommenden Fälle genügend Trainingsdaten, so die Idee, lernt es daraus selbst, bislang unbekannte Bilder zu klassifizieren. Alphago lernte beispielsweise unter anderem anhand von Millionen menschlicher Go-Spielzüge.
Auch überwachtes Lernen ist kaum zu kontrollieren Dabei führt der Begriff überwachtes Lernen in die Irre: Dieses Lernen ist weit weniger zu kontrollieren, als der Begriff suggeriert. Der Algorithmus entscheidet schließlich auf eigene Faust, welche Kriterien wichtig sind für die Unterscheidung. "Deshalb ist es zentral, dass der Trainingsdatensatz repräsentativ ist für die Art von Daten, die man vorhersagen will", sagt Fred Hamprecht, Professor für Bildverarbeitung an der Universität Heidelberg. Das kann allerdings kniffelig sein. So kursiert in Forscherkreisen das Beispiel eines Systems, das darauf trainiert wurde, Panzer auf Bildern zu erkennen. Der Trainingsdatensatz bestand aus Werbebildern der Herstellerfirmen von Panzern und beliebigen anderen Bildern, auf denen kein Panzer zu sehen war. Aber das System funktionierte in der Anwendung nicht - es erkannte Panzer nicht, sondern filterte stattdessen Bilder heraus, auf denen die Sonne schien. Das Problem war schnell erkannt: Auf den Werbebildern hatte ebenfalls stets die Sonne geschienen. Das Netz hatte das als Kriterium angenommen. "Falls das Beispiel nicht wahr ist, ist es zumindest schön erfunden", sagt Hamprecht. Aber nicht alle Fehler sind so einfach zu finden. "Die Frage ist immer, woher die Daten kommen", sagt Hamprecht. Ein Sensor beispielsweise altert oder verschmutzt, Bilder werden eventuell mit der Zeit dunkler. Wer kein falsches Ergebnis haben möchte, muss diese "Datenalterung" mit einrechnen - und sich ihrer dafür erstmal bewusst sein. Auch ein falsches Ergebnis wird nicht unbedingt so einfach entdeckt: Schließlich entscheiden Algorithmen nicht nur über für Menschen offensichtlich zu erkennende Dinge wie die, ob auf einem Bild ein Panzer ist.
Neuronale Netze lernen aus Erfahrungen Angesichts immer größerer Computer und wachsender Masse an Trainingsdaten gewinnen bei der Bilderkennung so genannte neuronale Netze immer mehr an Bedeutung. "Sie sind heute die leistungsfähigsten Mustererkennungsverfahren", sagt Hamprecht. Dabei wird die Funktionsweise des menschlichen Gehirns lose nachgeahmt: Die Netze bestehen aus mehreren Lagen mit einer festzulegenden Anzahl an Neuronen, deren Verbindungen sich verstärken oder abschwächen abhängig von den "Erfahrungen", die sie machen. Solche Erfahrungen sind beispielsweise die Trainingsdaten aus dem überwachten Lernen und das Feedback, ob zu einem Trainingsdatum die richtige oder falsche Vorhersage gemacht wurde. Dank der vielen Übungsdaten lassen sich heute sehr viel größere und tiefere Netze trainieren als noch vor einigen Jahren. Während früher ein berühmter Computer-Vision-Datensatz aus 256 Bildern und sein Nachfolger aus 1000 Bildern bestand, gibt es heute Datensätze mit einer Million gelabelter Bilder - also solche, auf denen Menschen markiert haben, was darauf zu sehen ist. Aber die Netze haben auch entscheidende Haken: "Man kann bei neuronalen Netzen schwer nachvollziehen, wie sie zu einer Entscheidung kamen", sagt Hamprecht. Zudem beruhe der Entwurf neuronaler Netze auf einer großen Willkür: Bei der Entscheidung, wie viele Lagen mit wie vielen Neuronen genutzt werden sollten, beruhe vieles auf Bauchgefühl oder auf Ausprobieren. Die Entwickler testen verschiedene Varianten und schauen, wann das beste Ergebnis entsteht. Erklären können sie ihre Entscheidung nicht. "Dieses Rumprobieren gefällt mir nicht", sagt Hamprecht, "ich gehe den Dingen lieber auf den Grund.""
Field
Informatik

Similar documents (author)

  1. Wolfangel, E.: Ich verstehe (2017) 6.18 
    6.176928 = sum of:
  6.176928 = weight(author_txt:wolfangel in 5441) [ClassicSimilarity], result of:
    6.176928 = score(doc=5441,freq=1.0), product of:
      0.99999994 = queryWeight, product of:
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.101182975 = queryNorm
      6.1769285 = fieldWeight in 5441, product of:
        1.0 = tf(freq=1.0), with freq of:
          1.0 = termFreq=1.0
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.625 = fieldNorm(doc=5441)
  2. Wolfangel, E.: Immer stur dem Navi nach (2017) 6.18 
    6.176928 = sum of:
  6.176928 = weight(author_txt:wolfangel in 162) [ClassicSimilarity], result of:
    6.176928 = score(doc=162,freq=1.0), product of:
      0.99999994 = queryWeight, product of:
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.101182975 = queryNorm
      6.1769285 = fieldWeight in 162, product of:
        1.0 = tf(freq=1.0), with freq of:
          1.0 = termFreq=1.0
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.625 = fieldNorm(doc=162)
  3. Wolfangel, E.: Oh, wie so trügerisch (2018) 6.18 
    6.176928 = sum of:
  6.176928 = weight(author_txt:wolfangel in 163) [ClassicSimilarity], result of:
    6.176928 = score(doc=163,freq=1.0), product of:
      0.99999994 = queryWeight, product of:
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.101182975 = queryNorm
      6.1769285 = fieldWeight in 163, product of:
        1.0 = tf(freq=1.0), with freq of:
          1.0 = termFreq=1.0
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.625 = fieldNorm(doc=163)
  4. Wolfangel, E.: DeepMind will Problem der Proteinfaltung gelöst haben (2020) 6.18 
    6.176928 = sum of:
  6.176928 = weight(author_txt:wolfangel in 636) [ClassicSimilarity], result of:
    6.176928 = score(doc=636,freq=1.0), product of:
      0.99999994 = queryWeight, product of:
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.101182975 = queryNorm
      6.1769285 = fieldWeight in 636, product of:
        1.0 = tf(freq=1.0), with freq of:
          1.0 = termFreq=1.0
        9.883085 = idf(docFreq=5, maxDocs=43254)
        0.625 = fieldNorm(doc=636)

Similar documents (content)

  1. Gillitzer, B.: Yewno (2017) 0.24 
    0.24205321 = sum of:
  0.24205321 = product of:
    0.67237 = sum of:
      0.013669514 = weight(abstract_txt:auch in 4912) [ClassicSimilarity], result of:
        0.013669514 = score(doc=4912,freq=2.0), product of:
          0.046985842 = queryWeight, product of:
            3.76169 = idf(docFreq=2732, maxDocs=43254)
            0.012490621 = queryNorm
          0.29092836 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            3.76169 = idf(docFreq=2732, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.015174456 = weight(abstract_txt:einer in 4912) [ClassicSimilarity], result of:
        0.015174456 = score(doc=4912,freq=2.0), product of:
          0.050374072 = queryWeight, product of:
            1.0354283 = boost
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.012490621 = queryNorm
          0.30123544 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.019842485 = weight(abstract_txt:über in 4912) [ClassicSimilarity], result of:
        0.019842485 = score(doc=4912,freq=3.0), product of:
          0.052621745 = queryWeight, product of:
            1.0582764 = boost
            3.9809077 = idf(docFreq=2194, maxDocs=43254)
            0.012490621 = queryNorm
          0.37707767 = fieldWeight in 4912, product of:
            1.7320508 = tf(freq=3.0), with freq of:
              3.0 = termFreq=3.0
            3.9809077 = idf(docFreq=2194, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.014964363 = weight(abstract_txt:software in 4912) [ClassicSimilarity], result of:
        0.014964363 = score(doc=4912,freq=1.0), product of:
          0.06288018 = queryWeight, product of:
            1.1568402 = boost
            4.351674 = idf(docFreq=1514, maxDocs=43254)
            0.012490621 = queryNorm
          0.23798218 = fieldWeight in 4912, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            4.351674 = idf(docFreq=1514, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.02930596 = weight(abstract_txt:ihre in 4912) [ClassicSimilarity], result of:
        0.02930596 = score(doc=4912,freq=2.0), product of:
          0.07812121 = queryWeight, product of:
            1.2894394 = boost
            4.8504715 = idf(docFreq=919, maxDocs=43254)
            0.012490621 = queryNorm
          0.3751345 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            4.8504715 = idf(docFreq=919, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.11494818 = weight(abstract_txt:maschinellen in 4912) [ClassicSimilarity], result of:
        0.11494818 = score(doc=4912,freq=2.0), product of:
          0.19429712 = queryWeight, product of:
            2.0335257 = boost
            7.649493 = idf(docFreq=55, maxDocs=43254)
            0.012490621 = queryNorm
          0.5916103 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            7.649493 = idf(docFreq=55, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.12404602 = weight(abstract_txt:lernens in 4912) [ClassicSimilarity], result of:
        0.12404602 = score(doc=4912,freq=2.0), product of:
          0.2044185 = queryWeight, product of:
            2.0858188 = boost
            7.846204 = idf(docFreq=45, maxDocs=43254)
            0.012490621 = queryNorm
          0.60682386 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            7.846204 = idf(docFreq=45, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.14322154 = weight(abstract_txt:intelligenz in 4912) [ClassicSimilarity], result of:
        0.14322154 = score(doc=4912,freq=2.0), product of:
          0.28345266 = queryWeight, product of:
            3.4735355 = boost
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.012490621 = queryNorm
          0.505275 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
      0.19719748 = weight(abstract_txt:künstlichen in 4912) [ClassicSimilarity], result of:
        0.19719748 = score(doc=4912,freq=2.0), product of:
          0.35081315 = queryWeight, product of:
            3.8642874 = boost
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.012490621 = queryNorm
          0.56211543 = fieldWeight in 4912, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.0546875 = fieldNorm(doc=4912)
    0.36 = coord(9/25)
  2. Krempl, S.: Missing Link : wie KI das menschliche Handlungsvermögen untergräbt (2020) 0.17 
    0.1707612 = sum of:
  0.1707612 = product of:
    0.711505 = sum of:
      0.016569953 = weight(abstract_txt:auch in 63) [ClassicSimilarity], result of:
        0.016569953 = score(doc=63,freq=1.0), product of:
          0.046985842 = queryWeight, product of:
            3.76169 = idf(docFreq=2732, maxDocs=43254)
            0.012490621 = queryNorm
          0.35265842 = fieldWeight in 63, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            3.76169 = idf(docFreq=2732, maxDocs=43254)
            0.09375 = fieldNorm(doc=63)
      0.019638967 = weight(abstract_txt:über in 63) [ClassicSimilarity], result of:
        0.019638967 = score(doc=63,freq=1.0), product of:
          0.052621745 = queryWeight, product of:
            1.0582764 = boost
            3.9809077 = idf(docFreq=2194, maxDocs=43254)
            0.012490621 = queryNorm
          0.3732101 = fieldWeight in 63, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            3.9809077 = idf(docFreq=2194, maxDocs=43254)
            0.09375 = fieldNorm(doc=63)
      0.12330767 = weight(abstract_txt:algorithmen in 63) [ClassicSimilarity], result of:
        0.12330767 = score(doc=63,freq=1.0), product of:
          0.17909339 = queryWeight, product of:
            1.9523437 = boost
            7.3441114 = idf(docFreq=75, maxDocs=43254)
            0.012490621 = queryNorm
          0.6885104 = fieldWeight in 63, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.3441114 = idf(docFreq=75, maxDocs=43254)
            0.09375 = fieldNorm(doc=63)
      0.13933823 = weight(abstract_txt:maschinellen in 63) [ClassicSimilarity], result of:
        0.13933823 = score(doc=63,freq=1.0), product of:
          0.19429712 = queryWeight, product of:
            2.0335257 = boost
            7.649493 = idf(docFreq=55, maxDocs=43254)
            0.012490621 = queryNorm
          0.71713996 = fieldWeight in 63, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.649493 = idf(docFreq=55, maxDocs=43254)
            0.09375 = fieldNorm(doc=63)
      0.17361072 = weight(abstract_txt:intelligenz in 63) [ClassicSimilarity], result of:
        0.17361072 = score(doc=63,freq=1.0), product of:
          0.28345266 = queryWeight, product of:
            3.4735355 = boost
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.012490621 = queryNorm
          0.61248577 = fieldWeight in 63, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.09375 = fieldNorm(doc=63)
      0.23903944 = weight(abstract_txt:künstlichen in 63) [ClassicSimilarity], result of:
        0.23903944 = score(doc=63,freq=1.0), product of:
          0.35081315 = queryWeight, product of:
            3.8642874 = boost
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.012490621 = queryNorm
          0.68138677 = fieldWeight in 63, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.09375 = fieldNorm(doc=63)
    0.24 = coord(6/25)
  3. Rötzer, F.: Computer ergooglen die Bedeutung von Worten (2005) 0.15 
    0.15345857 = sum of:
  0.15345857 = product of:
    0.9591161 = sum of:
      0.036788434 = weight(abstract_txt:einer in 5386) [ClassicSimilarity], result of:
        0.036788434 = score(doc=5386,freq=1.0), product of:
          0.050374072 = queryWeight, product of:
            1.0354283 = boost
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.012490621 = queryNorm
          0.73030496 = fieldWeight in 5386, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.1875 = fieldNorm(doc=5386)
      0.0970274 = weight(abstract_txt:google in 5386) [ClassicSimilarity], result of:
        0.0970274 = score(doc=5386,freq=1.0), product of:
          0.096160285 = queryWeight, product of:
            1.4305873 = boost
            5.381426 = idf(docFreq=540, maxDocs=43254)
            0.012490621 = queryNorm
          1.0090173 = fieldWeight in 5386, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            5.381426 = idf(docFreq=540, maxDocs=43254)
            0.1875 = fieldNorm(doc=5386)
      0.34722143 = weight(abstract_txt:intelligenz in 5386) [ClassicSimilarity], result of:
        0.34722143 = score(doc=5386,freq=1.0), product of:
          0.28345266 = queryWeight, product of:
            3.4735355 = boost
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.012490621 = queryNorm
          1.2249715 = fieldWeight in 5386, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.1875 = fieldNorm(doc=5386)
      0.47807887 = weight(abstract_txt:künstlichen in 5386) [ClassicSimilarity], result of:
        0.47807887 = score(doc=5386,freq=1.0), product of:
          0.35081315 = queryWeight, product of:
            3.8642874 = boost
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.012490621 = queryNorm
          1.3627735 = fieldWeight in 5386, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.1875 = fieldNorm(doc=5386)
    0.16 = coord(4/25)
  4. Kasprzik, A.: Automatisierte und semiautomatisierte Klassifizierung : eine Analyse aktueller Projekte (2014) 0.15 
    0.15197356 = sum of:
  0.15197356 = product of:
    0.6332232 = sum of:
      0.026549766 = weight(abstract_txt:einer in 3935) [ClassicSimilarity], result of:
        0.026549766 = score(doc=3935,freq=3.0), product of:
          0.050374072 = queryWeight, product of:
            1.0354283 = boost
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.012490621 = queryNorm
          0.5270522 = fieldWeight in 3935, product of:
            1.7320508 = tf(freq=3.0), with freq of:
              3.0 = termFreq=3.0
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.078125 = fieldNorm(doc=3935)
      0.02137766 = weight(abstract_txt:software in 3935) [ClassicSimilarity], result of:
        0.02137766 = score(doc=3935,freq=1.0), product of:
          0.06288018 = queryWeight, product of:
            1.1568402 = boost
            4.351674 = idf(docFreq=1514, maxDocs=43254)
            0.012490621 = queryNorm
          0.33997452 = fieldWeight in 3935, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            4.351674 = idf(docFreq=1514, maxDocs=43254)
            0.078125 = fieldNorm(doc=3935)
      0.1161152 = weight(abstract_txt:maschinellen in 3935) [ClassicSimilarity], result of:
        0.1161152 = score(doc=3935,freq=1.0), product of:
          0.19429712 = queryWeight, product of:
            2.0335257 = boost
            7.649493 = idf(docFreq=55, maxDocs=43254)
            0.012490621 = queryNorm
          0.5976167 = fieldWeight in 3935, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.649493 = idf(docFreq=55, maxDocs=43254)
            0.078125 = fieldNorm(doc=3935)
      0.1253054 = weight(abstract_txt:lernens in 3935) [ClassicSimilarity], result of:
        0.1253054 = score(doc=3935,freq=1.0), product of:
          0.2044185 = queryWeight, product of:
            2.0858188 = boost
            7.846204 = idf(docFreq=45, maxDocs=43254)
            0.012490621 = queryNorm
          0.61298466 = fieldWeight in 3935, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.846204 = idf(docFreq=45, maxDocs=43254)
            0.078125 = fieldNorm(doc=3935)
      0.14467561 = weight(abstract_txt:intelligenz in 3935) [ClassicSimilarity], result of:
        0.14467561 = score(doc=3935,freq=1.0), product of:
          0.28345266 = queryWeight, product of:
            3.4735355 = boost
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.012490621 = queryNorm
          0.5104048 = fieldWeight in 3935, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            6.5331817 = idf(docFreq=170, maxDocs=43254)
            0.078125 = fieldNorm(doc=3935)
      0.19919954 = weight(abstract_txt:künstlichen in 3935) [ClassicSimilarity], result of:
        0.19919954 = score(doc=3935,freq=1.0), product of:
          0.35081315 = queryWeight, product of:
            3.8642874 = boost
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.012490621 = queryNorm
          0.56782234 = fieldWeight in 3935, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.2681255 = idf(docFreq=81, maxDocs=43254)
            0.078125 = fieldNorm(doc=3935)
    0.24 = coord(6/25)
  5. Diedrich, H.: Wissensvernetzung : Chancen neuer Prozesse (2005) 0.15 
    0.14794339 = sum of:
  0.14794339 = product of:
    0.36985847 = sum of:
      0.013669514 = weight(abstract_txt:auch in 5487) [ClassicSimilarity], result of:
        0.013669514 = score(doc=5487,freq=2.0), product of:
          0.046985842 = queryWeight, product of:
            3.76169 = idf(docFreq=2732, maxDocs=43254)
            0.012490621 = queryNorm
          0.29092836 = fieldWeight in 5487, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            3.76169 = idf(docFreq=2732, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.01072996 = weight(abstract_txt:einer in 5487) [ClassicSimilarity], result of:
        0.01072996 = score(doc=5487,freq=1.0), product of:
          0.050374072 = queryWeight, product of:
            1.0354283 = boost
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.012490621 = queryNorm
          0.21300562 = fieldWeight in 5487, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            3.89496 = idf(docFreq=2391, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.016201321 = weight(abstract_txt:über in 5487) [ClassicSimilarity], result of:
        0.016201321 = score(doc=5487,freq=2.0), product of:
          0.052621745 = queryWeight, product of:
            1.0582764 = boost
            3.9809077 = idf(docFreq=2194, maxDocs=43254)
            0.012490621 = queryNorm
          0.30788264 = fieldWeight in 5487, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            3.9809077 = idf(docFreq=2194, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.014964363 = weight(abstract_txt:software in 5487) [ClassicSimilarity], result of:
        0.014964363 = score(doc=5487,freq=1.0), product of:
          0.06288018 = queryWeight, product of:
            1.1568402 = boost
            4.351674 = idf(docFreq=1514, maxDocs=43254)
            0.012490621 = queryNorm
          0.23798218 = fieldWeight in 5487, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            4.351674 = idf(docFreq=1514, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.017154686 = weight(abstract_txt:dass in 5487) [ClassicSimilarity], result of:
        0.017154686 = score(doc=5487,freq=1.0), product of:
          0.0688753 = queryWeight, product of:
            1.2107326 = boost
            4.5544004 = idf(docFreq=1236, maxDocs=43254)
            0.012490621 = queryNorm
          0.24906877 = fieldWeight in 5487, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            4.5544004 = idf(docFreq=1236, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.02930596 = weight(abstract_txt:ihre in 5487) [ClassicSimilarity], result of:
        0.02930596 = score(doc=5487,freq=2.0), product of:
          0.07812121 = queryWeight, product of:
            1.2894394 = boost
            4.8504715 = idf(docFreq=919, maxDocs=43254)
            0.012490621 = queryNorm
          0.3751345 = fieldWeight in 5487, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            4.8504715 = idf(docFreq=919, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.02829966 = weight(abstract_txt:google in 5487) [ClassicSimilarity], result of:
        0.02829966 = score(doc=5487,freq=1.0), product of:
          0.096160285 = queryWeight, product of:
            1.4305873 = boost
            5.381426 = idf(docFreq=540, maxDocs=43254)
            0.012490621 = queryNorm
          0.29429674 = fieldWeight in 5487, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            5.381426 = idf(docFreq=540, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.07915056 = weight(abstract_txt:technologie in 5487) [ClassicSimilarity], result of:
        0.07915056 = score(doc=5487,freq=2.0), product of:
          0.15150756 = queryWeight, product of:
            1.7956994 = boost
            6.754864 = idf(docFreq=136, maxDocs=43254)
            0.012490621 = queryNorm
          0.52241987 = fieldWeight in 5487, product of:
            1.4142135 = tf(freq=2.0), with freq of:
              2.0 = termFreq=2.0
            6.754864 = idf(docFreq=136, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.07192947 = weight(abstract_txt:algorithmen in 5487) [ClassicSimilarity], result of:
        0.07192947 = score(doc=5487,freq=1.0), product of:
          0.17909339 = queryWeight, product of:
            1.9523437 = boost
            7.3441114 = idf(docFreq=75, maxDocs=43254)
            0.012490621 = queryNorm
          0.4016311 = fieldWeight in 5487, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.3441114 = idf(docFreq=75, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
      0.088452965 = weight(abstract_txt:fallen in 5487) [ClassicSimilarity], result of:
        0.088452965 = score(doc=5487,freq=1.0), product of:
          0.20556535 = queryWeight, product of:
            2.0916617 = boost
            7.8681827 = idf(docFreq=44, maxDocs=43254)
            0.012490621 = queryNorm
          0.43029124 = fieldWeight in 5487, product of:
            1.0 = tf(freq=1.0), with freq of:
              1.0 = termFreq=1.0
            7.8681827 = idf(docFreq=44, maxDocs=43254)
            0.0546875 = fieldNorm(doc=5487)
    0.4 = coord(10/25)