Rasa: Unable to extract entities in arabic chatbot

[balinda-1995]

Issue: I am trying to create an arabic chatbot but my model can't extract the entities. I get this error when I test my model on an utterance that contains entities to extract:

UnicodeEncodeError: 'ascii' codec can't encode characters in position 2-4: ordinal not inrange(128)`

This is my configuration file:

pipeline:
- name: "nlp_spacy"
- name: "tokenizer_spacy"
- name: "intent_entity_featurizer_regex"
- name: "intent_featurizer_spacy"
- name: "ner_crf"
- name: "ner_synonyms"
- name: "intent_featurizer_count_vectors"
- name: "intent_classifier_tensorflow_embedding"
  intent_tokenization_flag: true
  intent_split_symbol: "+"

[akelad]

Could you post what NLU version you're using and the full error trace please?

[balinda-1995]

I actually no longer get that error (I have no idea what happened), but I am getting a different one now.
I still have the issue with the entity extractions, my NLU model works fine, now the problem is with my dialgue management model. I am trying to train my model by providing the domain file.

My NLU version is 0.12.3

My domain file is:

slots:
  horo:
    type:text

intents:
 - سلام
 - طلب
 - برج
 - وداع+شكر

entities:
 - horo
actions:
 - actions.Selem
 - actions.Goodbye
 - actions.Horoscope
 - actions.AskSign
`````

and the error I get is:

Traceback (most recent call last):
File "C:/Users/Asus/Desktop/eng_intern/arabicBot/train_init.py", line 21, in
agent = Agent('domain.yml', policies=[MemoizationPolicy(max_history=2), KerasPolicy()])
File "C:\Users\Asus\AppData\Local\Programs\Python\Python36\lib\site-packages\rasa_core\agent.py", line 51, in __init__
self.domain = self._create_domain(domain)
File "C:\Users\Asus\AppData\Local\Programs\Python\Python36\lib\site-packages\rasa_core\agent.py", line 390, in _create_domain
return TemplateDomain.load(domain)
File "C:\Users\Asus\AppData\Local\Programs\Python\Python36\lib\site-packages\rasa_core\domain.py", line 402, in load
return cls.load_from_yaml(read_file(filename), action_factory=action_factory)
File "C:\Users\Asus\AppData\Local\Programs\Python\Python36\lib\site-packages\rasa_core\domain.py", line 411, in load_from_yaml
slots = cls.collect_slots(data.get("slots", {}))
File "C:\Users\Asus\AppData\Local\Programs\Python\Python36\lib\site-packages\rasa_core\domain.py", line 452, in collect_slots
slot_class = Slot.resolve_by_type(slot_dict[slot_name].get("type"))
AttributeError: 'str' object has no attribute 'get'
```
I am thinking maybe it's a problem with the type of the slot I have specified in my domain file, but I can't think of what else can I put.

[akelad]

i'd suggest also naming your intents in english, but i can't really see anything wrong with your domain file otherwise. What version of rasa_core are you using?

[balinda-1995]

My Rasa core version is 0.10.1
I had no problems with the intents, I have actually tried removing the entity and working only with intents, the chatbot worked perfectly.
I am using ner-crf to extract the entities, maybe it doesn't work with the arabic language?

[balinda-1995]

After long long researches, I finally found something. I tried to add - name: "ner_duckling" to my configuration file but I get this error when training my NLU model:

jpype._jvmfinder.JVMNotFoundException: No JVM shared library file (jvm.dll) found. Try setting up the JAVA_HOME environment variable properly.

What does Java have to do here?!

[balinda-1995]

SOLVED, it was a very silly mistake (missing space in my domain file costed me 1 week of search! )

[akelad]

what was the mistake?

[balinda-1995]

I should have put a space before text in my domain file, so it should have been:
Type: text
as silly as that ;)
It works perfectly now, no errors

[akelad]

ah cool!