I'm currently starting a research project focused on NLP.

One of the steps involved in this project will be the development of a text simplification system, probably using a neural encoder-decoder architecture.

For most Text Simplification research available, the most commonly used dataset is one derived from pairing Wikipedia entries in both English and Simplified English. My problem arises from the fact that the focus of my research is not on the English Language, but rather in Portuguese, specifically Portugal Portuguese.

There exists no Simple Portuguese Wikipedia page and it seems that there exists no publicly available text simplification dataset in Portugal Portuguese at all. Due to this fact I'm curious if there would be any way of tackling this problem. Maybe having a dataset simply of complex Portuguese and simple portuguese, but with no pairings, although I'm not quite sure how that could be formulated to train a NN with.

So my question is if there are any text simplification datasets in Portugal (or maybe Bazil, as a last resource) Portuguese and, if not, what would be the optimal way to build that dataset.

Thank you.

  • $\begingroup$ What is your specific question? Please, edit your post to clarify this (and put the specific question in the title). It seems that your problem is about training a model with a very small dataset (that you're planning to build). Have you heard of 1. one-shot learning or 2. transfer learning? If not, you may want to have a look at these concepts (then, as I said, please, edit the post to ask a specific question that also takes into account the info that I just gave you.) $\endgroup$ – nbro Dec 7 '20 at 14:40
  • $\begingroup$ Reformulated the question $\endgroup$ – joão correia Dec 7 '20 at 15:45
  • $\begingroup$ If you're asking for datasets, then this is off-topic. See ai.stackexchange.com/help/on-topic. I'm not sure you're asking for datasets though. Can you clarify this? If you're looking for datasets, maybe opendata.stackexchange.com (or eventually Data Science SE) is a better place to ask this question. Here we focus on theoretical aspects of AI. $\endgroup$ – nbro Dec 7 '20 at 16:20

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.