# Test a Model's knowledge

**URL:** <https://discuss.huggingface.co/t/test-a-models-knowledge/17471>\
**Category:** Beginners\
**Created:** [May 3, 2022, 1:24pm UTC](https://discuss.huggingface.co/t/test-a-models-knowledge/17471 "2022-05-03T13:24:09Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![giklo](https://avatars.discourse-cdn.com/v4/letter/g/8dc957/32.png) [@giklo](https://discuss.huggingface.co/u/giklo)\
**Post date:** [May 3, 2022, 1:24pm UTC](https://discuss.huggingface.co/t/test-a-models-knowledge/17471/1 "2022-05-03T13:24:09Z")

</div>

I’m currently training Semantic Textual Similarity using the SentenceTransformers Library.  
Models are [BERT](https://huggingface.co/bert-base-german-cased) and [Paraphrase MPNet](https://huggingface.co/sentence-transformers/paraphrase-multilingual-mpnet-base-v2).

Is there any way to check or even proof if a model understands a specific word or sequence?

I tried masking or calculating the similarity to specific categories related to the word but the results were hardly distinguishable.

Thanks in advance