# Feedback on ConveRT Model + Rasa NLU

**URL:** <https://forum.rasa.com/t/feedback-on-convert-model-rasa-nlu/21312>\
**Category:** Rasa Open Source\
**Created:** [November 15, 2019, 4:55pm UTC](https://forum.rasa.com/t/feedback-on-convert-model-rasa-nlu/21312 "2019-11-15T16:55:34Z")\
**Posts on this page:** 1\
**Showing post:** 57

<div class="post-metadata">

**Author:** ![rslater](https://avatars.discourse-cdn.com/v4/letter/r/4491bb/32.png) [@rslater](https://forum.rasa.com/u/rslater)\
**Post date:** [December 12, 2019, 3:30pm UTC](https://forum.rasa.com/t/feedback-on-convert-model-rasa-nlu/21312/57 "2019-12-12T15:30:12Z")

</div>

The data model is English Words Only–that may be a source of issue.

I also was going through the paper last night on ConveRT (I can never spell it right) and I noticed it uses sub word tokens but not character level embeddings so i bet spelling has a huge impact on performance. I was going to experiment with that myself this am.

---

_[View the full topic](https://forum.rasa.com/t/feedback-on-convert-model-rasa-nlu/21312)._