# NLU detects random input with wrong intent that is none with zero confidence

**URL:** <https://forum.rasa.com/t/nlu-detects-random-input-with-wrong-intent-that-is-none-with-zero-confidence/5853>\
**Category:** Rasa Open Source\
**Created:** [March 8, 2019, 10:34am UTC](https://forum.rasa.com/t/nlu-detects-random-input-with-wrong-intent-that-is-none-with-zero-confidence/5853 "2019-03-08T10:34:05Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![noman](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/noman/32/10071_2.png) [@noman](https://forum.rasa.com/u/noman)\
**Post date:** [March 8, 2019, 10:34am UTC](https://forum.rasa.com/t/nlu-detects-random-input-with-wrong-intent-that-is-none-with-zero-confidence/5853/1 "2019-03-08T10:34:05Z")

</div>

If I provide random input text like "aoa, klh, umh” , Rasa NLU returns zero confidence with ‘none’ as an intent. I want my bot to call the fallback action on an out of scope words. Exact error is `intent": None` to `"intent": {"name": None, "confidence": 0.0}` . I am using tensorflow\_embedding as a pipeline. Help would me much appreciated. Thanks.

---

<div class="post-metadata">

**Author:** ![NikhilBansal21](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/nikhilbansal21/32/418_2.png) [@NikhilBansal21](https://forum.rasa.com/u/NikhilBansal21)\
**Post date:** [March 8, 2019, 11:45am UTC](https://forum.rasa.com/t/nlu-detects-random-input-with-wrong-intent-that-is-none-with-zero-confidence/5853/2 "2019-03-08T11:45:29Z")

</div>

The words you have mentioned above like "aoa, klh, umh” are termed as gibberish content , After 0.12 rasa returns intent as None for these cases as these are termed as out of scope messages. I have also seen these with tensorflow\_embedding , you can either try different intent classifier like mitie or spacy or you can provide some phrases like these to your fallback itself (but that works up to some extent only). Or you can handle None also in your intent handler code for your client

---

<div class="post-metadata">

**Author:** ![noman](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/noman/32/10071_2.png) [@noman](https://forum.rasa.com/u/noman)\
**Post date:** [March 8, 2019, 12:36pm UTC](https://forum.rasa.com/t/nlu-detects-random-input-with-wrong-intent-that-is-none-with-zero-confidence/5853/3 "2019-03-08T12:36:54Z")

</div>

Don’t you think there should be some other good solution rather than using mitie or spacy as intent classifiers.

---

<div class="post-metadata">

**Author:** ![NikhilBansal21](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/nikhilbansal21/32/418_2.png) [@NikhilBansal21](https://forum.rasa.com/u/NikhilBansal21)\
**Post date:** [March 8, 2019, 12:40pm UTC](https://forum.rasa.com/t/nlu-detects-random-input-with-wrong-intent-that-is-none-with-zero-confidence/5853/4 "2019-03-08T12:40:02Z")

</div>

if you find some then do tell me too 🙂