# Rasa extracting entity with special characters

**URL:** <https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554>\
**Category:** Rasa Open Source\
**Created:** [August 20, 2019, 11:27am UTC](https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554 "2019-08-20T11:27:42Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![ice2tricky](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/ice2tricky/32/4361_2.png) [@ice2tricky](https://forum.rasa.com/u/ice2tricky)\
**Post date:** [August 20, 2019, 11:27am UTC](https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554/1 "2019-08-20T11:27:42Z")

</div>

Hi all,

I have a problem with extracting special characters, for example word1 \> word2, then I get word1 word2. Anyone has a solution for this?

using this config:

language: en pipeline: supervised\_embeddings policies:

- name: MemoizationPolicy
- name: KerasPolicy
- name: MappingPolicy
- name: FormPolicy

---

<div class="post-metadata">

**Author:** ![tmbo](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/tmbo/32/19_2.png) [@tmbo](https://forum.rasa.com/u/tmbo)\
**Post date:** [August 23, 2019, 8:17am UTC](https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554/2 "2019-08-23T08:17:46Z")

</div>

So the `>` character is omitted from the result?

Can you please add a bit more detail on what you get from Rasa as a result and how your training data look like?

---

<div class="post-metadata">

**Author:** ![ice2tricky](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/ice2tricky/32/4361_2.png) [@ice2tricky](https://forum.rasa.com/u/ice2tricky)\
**Post date:** [August 23, 2019, 8:56am UTC](https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554/3 "2019-08-23T08:56:10Z")

</div>

## intent:get\_relations

- can you show me the related business terms
- what are the related business terms for Surprise
- what are the related business terms for kiva\_loans.csv \> funded\_amount in Kiva Loans?
- Can you give me the definition from
- Show me the definition from Caster Sleeve
- Give me the definition from Veneer
- get relations for Surprise
- can you check the related business terms for Surprise
- can you show me the relations from mpi\_subnational\_old.csv \> Sub-national region
- can you show me the relations from mpi\_subnational\_old.csv \> Sub-national region
- Get the relations from mpi\_subnational\_old.csv \> Sub-national region
- I would like to have only relations related to the Kiva Loans data set

He doesn’t pick up the \> sign from the entities, but the rest of the words it does. So result of the entity extraction is for example: mpi\_subnational\_old.csv Sub-national region

The same with:

- Is Massaud Conference Chair (mid-back) level 2 an asset?

gives: Massaud Conference Chair mid-back level 2

As a temporary solution I hacked a search function together, but it would be nice if it would extract the special signs.

And one I can’t solve, because he picks up nothing is, when the hook is at the back

- is [GSA Contract (General Services Administration)](asset\_name:GSA Contract (General Services Administration)) an asset?

- is GSA Contract (General Services Administration) an asset?

gives no entities

---

<div class="post-metadata">

**Author:** ![ice2tricky](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/ice2tricky/32/4361_2.png) [@ice2tricky](https://forum.rasa.com/u/ice2tricky)\
**Post date:** [September 10, 2019, 8:07am UTC](https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554/4 "2019-09-10T08:07:23Z")

</div>

Any update on this or should i report this as an issue?

---

<div class="post-metadata">

**Author:** ![1412392](https://dub1.discourse-cdn.com/flex013/user_avatar/forum.rasa.com/1412392/32/3008_2.png) [@1412392](https://forum.rasa.com/u/1412392)\
**Post date:** [October 23, 2019, 4:40am UTC](https://forum.rasa.com/t/rasa-extracting-entity-with-special-characters/14554/5 "2019-10-23T04:40:17Z")

</div>

find any solution for this problem?