arxiv.org/abs/2004.09813

Preview meta tags from the arxiv.org website.

Linked Hostnames

26

Thumbnail

Search Engine Appearance

Google

https://arxiv.org/abs/2004.09813

Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation

We present an easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models. The training is based on the idea that a translated sentence should be mapped to the same location in the vector space as the original sentence. We use the original (monolingual) model to generate sentence embeddings for the source language and then train a new system on translated sentences to mimic the original model. Compared to other methods for training multilingual sentence embeddings, this approach has several advantages: It is easy to extend existing models with relatively few samples to new languages, it is easier to ensure desired properties for the vector space, and the hardware requirements for training is lower. We demonstrate the effectiveness of our approach for 50+ languages from various language families. Code to extend sentence embeddings models to more than 400 languages is publicly available.



Bing

Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation

https://arxiv.org/abs/2004.09813

We present an easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models. The training is based on the idea that a translated sentence should be mapped to the same location in the vector space as the original sentence. We use the original (monolingual) model to generate sentence embeddings for the source language and then train a new system on translated sentences to mimic the original model. Compared to other methods for training multilingual sentence embeddings, this approach has several advantages: It is easy to extend existing models with relatively few samples to new languages, it is easier to ensure desired properties for the vector space, and the hardware requirements for training is lower. We demonstrate the effectiveness of our approach for 50+ languages from various language families. Code to extend sentence embeddings models to more than 400 languages is publicly available.



DuckDuckGo

https://arxiv.org/abs/2004.09813

Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation

We present an easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models. The training is based on the idea that a translated sentence should be mapped to the same location in the vector space as the original sentence. We use the original (monolingual) model to generate sentence embeddings for the source language and then train a new system on translated sentences to mimic the original model. Compared to other methods for training multilingual sentence embeddings, this approach has several advantages: It is easy to extend existing models with relatively few samples to new languages, it is easier to ensure desired properties for the vector space, and the hardware requirements for training is lower. We demonstrate the effectiveness of our approach for 50+ languages from various language families. Code to extend sentence embeddings models to more than 400 languages is publicly available.

  • General Meta Tags

    16
    • title
      [2004.09813] Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation
    • title
      open search
    • title
      open navigation menu
    • title
      contact arXiv
    • title
      subscribe to arXiv mailings
  • Open Graph Meta Tags

    10
    • og:type
      website
    • og:site_name
      arXiv.org
    • og:title
      Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation
    • og:url
      https://arxiv.org/abs/2004.09813v2
    • og:image
      /static/browse/0.3.4/images/arxiv-logo-fb.png
  • Twitter Meta Tags

    6
    • twitter:site
      @arxiv
    • twitter:card
      summary
    • twitter:title
      Making Monolingual Sentence Embeddings Multilingual using...
    • twitter:description
      We present an easy and efficient method to extend existing sentence embedding models to new languages. This allows to create multilingual versions from previously monolingual models. The training...
    • twitter:image
      https://static.arxiv.org/icons/twitter/arxiv-logo-twitter-square.png
  • Link Tags

    12
    • apple-touch-icon
      /static/browse/0.3.4/images/icons/apple-touch-icon.png
    • canonical
      /abs/2004.09813
    • icon
      /static/browse/0.3.4/images/icons/favicon-32x32.png
    • icon
      /static/browse/0.3.4/images/icons/favicon-16x16.png
    • manifest
      /static/browse/0.3.4/images/icons/site.webmanifest

Links

64