For my bachelor thesis i am probing various text similarity metrics for how much they incorporate different aspects of sentences.

To get those aspects I use datasets that are annotated for semantic similarity, use a syntax tree parser to calculate a syntactic similarity measure and calculate BLEU to get lexical overlap.

I would now also like to include some measure of a morphological similarity between sentence pairs. I have looked at the word embeddings intended to incode morphological data that are described here: https://www.aclweb.org/anthology/N15-1140.pdf (I would then either pool them for the sentence or find alignments by some strategy).

I could also directly compare morphological (sub-)tags in some way.

However, for these and other approaches I would need at least some data annotated with morphological tags, which the data annotated for semantic similarity that i use is not.

Is it reasonable to trust an automated morphological parser for that?

And are there any other ideas to calculate a morphological similarity between senteces?

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.