Limitations of cross-lingual learning from image search

    Abstract

    Cross-lingual representation learning is animportant step in making NLP scale to allthe world’s languages. Previous work onbilingual lexicon induction suggests that itis possible to learn cross-lingual representationsof words based on similarities betweenimages associated with these words.However, that work focused (almost exclusively)on the translation of nouns only.Here, we investigate whether the meaningof other parts-of-speech (POS), in particularadjectives and verbs, can be learned inthe same way. Our experiments across fivelanguage pairs indicate that previous workdoes not scale to the problem of learningcross-lingual representations beyond simplenouns.
    Original languageEnglish
    Title of host publicationProceedings of the 3rd Workshop on Representation Learning for NLP
    Publication date2018
    Pages59–163
    Publication statusPublished - 2018
    Event3rd Workshop on Representation Learning for NLP - Melbourne, Australia
    Duration: 20 Jul 201820 Jul 2018

    Workshop

    Workshop3rd Workshop on Representation Learning for NLP
    Country/TerritoryAustralia
    CityMelbourne
    Period20/07/201820/07/2018

    Cite this