Custom NER for identifying products

Question

I am trying to buld a custom named entity extractor for product names and their model numbers.

My use case contains sentences like: "Microsoft used product ABC-300 and also integrated it with ASQ". Product mentioned in the above sentence are: ABC-300 and ASQ

I have already tried using Stanford and Spacy NER, accuracy of both is less than desired.

Are there any datasets that contain product names in paragraphs or sentences I can use for training custom NER model? The sentences for training can be simple or complex. Any kind of data will be useful.

Any leads on how to approach this problem with less training data will also be appreciated.

Answer 1

one possible solution is to use the TokensRegexNERAnnotator ( https://stanfordnlp.github.io/CoreNLP/regexner.html )

This is assuming you can 'regex' the product names