---
base_model: bert-base-multilingual-uncased
model-index:
- name: lang-recogn-model
  results:
  - task:
      type: text-classification
    dataset:
      name: language-detection
      type: language-detection
    metrics:
    - name: accuracy
      type: accuracy
      value: 0.9836
    source:
      name: Language recognition using BERT
      url: >-
        https://www.kaggle.com/code/sergeypolivin/language-recognition-using-bert
language:
- ar
- da
- nl
- en
- fr
- de
- el
- hi
- it
- kn
- ml
- pt
- ru
- es
- sv
- ta
- tr
pipeline_tag: text-classification
widget:
- text: "Hello, world"
  example_title: "English"
- text: "Ik heb het al gezien"
  example_title: "Dutch"
- text: "Интересная идея"
  example_title: "Russian"
- text: "Que vamos a hacer?"
  example_title: "Spanish"
- text: "Hvor er der en pengeautomat?"
  example_title: "Danish"
---

# Language Detection Model

The model presented in the following repository represents a fine-tuned version of `BertForSequenceClassification` 
pretrained on [multilingual texts](https://huggingface.co/bert-base-multilingual-uncased).

## Training/fine-tuning

The model has been fine-tuned based on [Language Detection](https://www.kaggle.com/datasets/basilb2s/language-detection) 
dataset found on *Kaggle*. The entire process of the dataset analysis as well as a complete description of the training procedure
can be found in [one of my *Kaggle* notebooks](https://www.kaggle.com/code/sergeypolivin/language-recognition-using-bert) 
which has been used for the purpose of a faster model training on *GPU*.

## Supported languages

The model has been fine-tuned to detect one of the following 17 languages:

- Arabic
- Danish
- Dutch
- English
- French
- German
- Greek
- Hindi
- Italian
- Kannada
- Malayalam
- Portugeese
- Russian
- Spanish
- Sweedish
- Tamil
- Turkish

## References

1. [BERT multilingual base model (uncased)](https://huggingface.co/bert-base-multilingual-uncased)
2. [Language Detection Dataset](https://www.kaggle.com/datasets/basilb2s/language-detection)