I'm starting a project where I want to extract keywoards from given messages. The keywoards are for example something like: "hard disk", "watch" or other technical components. I'm working with a dataset where a technician wrote a small text if he maintenanced something on a given object.

The messages are often very different in their form. For example sometimes the messages start with the repaired object and sometimes with the current date..

I looked into some NER-Libarys and it doesn't seem like they can handle tasks like that. Especially the german language makes it hard for those libarys to detect entities.

I had the idea to use CRFsuite to train my own NER-Model. But I'm not sure how accurate the outcome will be. The process would include that i have to tag A LOT of training data and I'm not sure if the outcome will match the time I have to spend to tag those keywoards.

Does anybody has experience with such custom NER-Models? How accurate can such Model extract wanted Keywoards?

Any kind of feedback is appreaciated! Greetz


I would recommend you to use German Bert for embeddings: https://huggingface.co/bert-base-german-cased

Together with a cosine similarity, like described in this article: https://towardsdatascience.com/keyword-extraction-with-bert-724efca412ea

Bert is very powerful to extract knowledge from text efficiently, I used to reach extraordinary results thanks to that library.

  • $\begingroup$ Please do not "sign" your posts (edited). $\endgroup$
    – desertnaut
    Jun 17 '21 at 19:20

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.