I used a stopwords based approach.
- First step is to “tokenize”.
- NLTK gives us stopwords for several languages (here I used Arabic,French and English stopwords).
- I compute language probability depending on which stopwords are used.
- Finally, I chose the “key” with the biggest “value”.