# Access to the input generated for the DIET Classifier
**URL:** https://forum.rasa.com/t/access-to-the-input-generated-for-the-diet-classifier/31484
**Category:** Rasa Open Source
**Created:** [July 17, 2020, 5:50pm UTC](https://forum.rasa.com/t/access-to-the-input-generated-for-the-diet-classifier/31484 "2020-07-17T17:50:44Z")
**Posts on this page:** 3
**Page:** 1
### Author:  [@lfabing](https://forum.rasa.com/u/lfabing)
#### Post date: [July 17, 2020, 5:50pm UTC](https://forum.rasa.com/t/access-to-the-input-generated-for-the-diet-classifier/31484/1 "2020-07-17T17:50:44Z")
Hi there,
I want to measure intent similarity and was wondering if there was a way to access the input given to the DIET Classifier. In particular, I’d like to retrieve the sample embeddings generated by conveRT.
Has anyone here tried to do this?
---
### Author:  [@koaning](https://forum.rasa.com/u/koaning)
#### Post date: [July 23, 2020, 11:47am UTC](https://forum.rasa.com/t/access-to-the-input-generated-for-the-diet-classifier/31484/2 "2020-07-23T11:47:04Z")
There’s a ticket on the [whatlies library](https://github.com/RasaHQ/whatlies/issues/36) for exactly this feature. It should make it easy to use jupyter to explore what clusters appear in these embeddings.
Out of curiosity, what’s the usecase?
---
### Author:  [@koaning](https://forum.rasa.com/u/koaning)
#### Post date: [July 24, 2020, 8:20am UTC](https://forum.rasa.com/t/access-to-the-input-generated-for-the-diet-classifier/31484/3 "2020-07-24T08:20:56Z")
There’s a [PR](https://github.com/RasaHQ/whatlies/pull/95) for this feature now.