# Specific category in GPT output

**URL:** <https://discourse.cogstack.org/t/specific-category-in-gpt-output/218>\
**Category:** Foresight\
**Created:** [June 11, 2023, 6:04pm UTC](https://discourse.cogstack.org/t/specific-category-in-gpt-output/218 "2023-06-11T18:04:04Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![eredekop](https://avatars.discourse-cdn.com/v4/letter/e/6de8d8/32.png) [@eredekop](https://discourse.cogstack.org/u/eredekop)\
**Post date:** [June 11, 2023, 6:04pm UTC](https://discourse.cogstack.org/t/specific-category-in-gpt-output/218/1 "2023-06-11T18:04:04Z")

</div>

Hi! Could you please explain how you generate the next item in the patient timeline stratified by specified category, e.g. just Disorders?

---

<div class="post-metadata">

**Author:** ![Jthteo](https://avatars.discourse-cdn.com/v4/letter/j/898d66/32.png) [@Jthteo](https://discourse.cogstack.org/u/Jthteo)\
**Post date:** [June 13, 2023, 8:41pm UTC](https://discourse.cogstack.org/t/specific-category-in-gpt-output/218/2 "2023-06-13T20:41:27Z")

</div>

Hi are you using the web app? If so, it should be an option on upper right of the demo web app.

---

<div class="post-metadata">

**Author:** ![eredekop](https://avatars.discourse-cdn.com/v4/letter/e/6de8d8/32.png) [@eredekop](https://discourse.cogstack.org/u/eredekop)\
**Post date:** [June 13, 2023, 8:53pm UTC](https://discourse.cogstack.org/t/specific-category-in-gpt-output/218/3 "2023-06-13T20:53:16Z")

</div>

Hi! Thank you very much for your reply! I am curious how it works from the training prospective. As GPT models outputs the sequence of next probable tokens, so how do you force it to predict only diagnosis? or only procedures?