I am trying to see whether there are differences in sentences between the topics in 2 languages ie English and Spanish. Eg. (Face masks are mandatory, la mascarilla es obligatoria) but this stretched out to n number of queries. The final goal is to find whether a document in English talks about 1 topic and the other corpus in Spanish talks about a different one.

I tried using BiLDA for generating topics in both the languages simultaneously but my data is not super clean for the model to work properly and was giving me vague results.

I then went on towards BERT and TRANSFORMERS by using this wrapper https://github.com/amaiya/ktrain for applying the pre-trained models to classify text by giving my own topics to the model only realizing that it would not work for Spanish. I also tried to look at FACEBOOK'S LASER and thought about comparing the embedding space but again I am relying on their pre-trained space and would not get my intended goal.

Tried using transformers from huggingface. I used the automodelforsequenceclassification with zero shot learning which totally works for English, But it uses Facebook's bart mnli which only supports 1 language. I need support for multi language. If the same was working for spanish my job was done.

Zero-shot learnings helps me classify text into custom topics although it only works with English language.no support for other languages.

If anybody has done anything like that or can help me direct this it would be really helpful.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.