I am trying to make a text classification for Arabic data. The problem is that there is no labeled Arabic dataset for this data. My question is then: is possible to do a classification without a training dataset? If yes, what methods can I use?

  • $\begingroup$ Hi. It's not clear to me why you have "software requirements" in the title. Can you clarify that? $\endgroup$
    – nbro
    Feb 24, 2021 at 15:34
  • $\begingroup$ the data that I want to classify is a software requirements $\endgroup$
    – Eng.sh
    Feb 24, 2021 at 18:40

1 Answer 1


Yes, it is possible to do it. This can be posed as an unsupervised text classification. You can look at TF-IDF, Neural networks (BERT) etc. for creating embedding of your text and then, use clustering techniques like K-Means, KNN etc. for classification.

  • $\begingroup$ as i know unsupervised text classification needs an unlabeled dataset which I don't have $\endgroup$
    – Eng.sh
    Feb 24, 2021 at 18:42
  • $\begingroup$ if you even don't have data, how do you expect to create a model. unsupervised is albeit a safety net when supervised is not possible. no data = no model. anyways, the BERT link is for Arabic language, maybe use that to do it directly but that would still need data. $\endgroup$ Feb 24, 2021 at 19:07
  • $\begingroup$ lionbridge.ai/datasets/… and you do have datasets. or you need to create a small dataset yourself. but dataset is a precondition for a model. $\endgroup$ Feb 24, 2021 at 19:08
  • $\begingroup$ Thanks a lot :) $\endgroup$
    – Eng.sh
    Feb 24, 2021 at 23:36

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .