# What does Rasa use internally for text classification?

**URL:** <https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747>\
**Category:** Rasa Open Source\
**Created:** [July 10, 2019, 11:32am UTC](https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747 "2019-07-10T11:32:41Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![qeia](https://avatars.discourse-cdn.com/v4/letter/q/439d5e/32.png) [@qeia](https://forum.rasa.com/u/qeia)\
**Post date:** [July 10, 2019, 11:32am UTC](https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747/1 "2019-07-10T11:32:41Z")

</div>

Hi people, I’m new here. But I could not find a comprehensive article/documentation on what exactly goes on under the hood for text (intent) classification. I found a blog post that detailed the use of StarSpace but I didn’t find it superficial. I would like a more in depth description. Could any one point me to the right direction?

---

<div class="post-metadata">

**Author:** ![JiteshGaikwad](https://avatars.discourse-cdn.com/v4/letter/j/77aa72/32.png) [@JiteshGaikwad](https://forum.rasa.com/u/JiteshGaikwad)\
**Post date:** [July 10, 2019, 12:08pm UTC](https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747/2 "2019-07-10T12:08:56Z")

</div>

hey @qeia, you can read this article:

[http://www.thegeekette.me/?p=503&preview=1&\_ppp=000fd6c5e2](http://www.thegeekette.me/?p=503&preview=1&_ppp=000fd6c5e2)

---

<div class="post-metadata">

**Author:** ![qeia](https://avatars.discourse-cdn.com/v4/letter/q/439d5e/32.png) [@qeia](https://forum.rasa.com/u/qeia)\
**Post date:** [July 10, 2019, 2:13pm UTC](https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747/3 "2019-07-10T14:13:41Z")

</div>

Doesn’t open. Is there an official source?

---

<div class="post-metadata">

**Author:** ![JiteshGaikwad](https://avatars.discourse-cdn.com/v4/letter/j/77aa72/32.png) [@JiteshGaikwad](https://forum.rasa.com/u/JiteshGaikwad)\
**Post date:** [July 10, 2019, 3:47pm UTC](https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747/4 "2019-07-10T15:47:51Z")

</div>

hey @qeia, you can check it now:

[NLP Behind Chatbots - Demystifying RasaNLU - #1 - Training](https://medium.com/bhavaniravi/demystifying-rasa-nlu-1-training-91a08429c9fb)

---

<div class="post-metadata">

**Author:** ![qeia](https://avatars.discourse-cdn.com/v4/letter/q/439d5e/32.png) [@qeia](https://forum.rasa.com/u/qeia)\
**Post date:** [July 14, 2019, 4:46am UTC](https://forum.rasa.com/t/what-does-rasa-use-internally-for-text-classification/12747/5 "2019-07-14T04:46:47Z")

</div>

I have read this, but as I said, I’m looking for a more official source. Moreover, the link that you had provided does not explain the intent classification process clearly. I understand that the tokens are converted to their respective word vectors. Then what? Do we average them? Do we add them? Do we do a doc2vec classification? Nothing is clear.

If any one could point me to the right directions on the internals of intent classification of RASA NLU, that would be great!