How to extract date, time & duration using CRF Entity extractor

Rasa Open Source
1

Please advise if we can use CRF Entity Extractor to extract date, time & duration. If so how should I annotate my training data.

I am aware that duckling is more efficient in doing this. However, I would like to keep it simple and hence do not want to spin & run duckling server just for date & duration. I am heavily dependent upon crf for my custom entities.

2

hey @siriusraja, CRF Entity Extractor only works for extracting custom entities and it doesn’t support date & time extraction.

I have created an alternative for duckling for the same you can check it out here:

https://github.com/JiteshGaikwad/EntityExtractor.git

let me know if this helps you :slight_smile:

3

Hi Jitesh

Thank you for the suggestion. I was trying your solution and faced the below exception.

In the MSRTEntityExtractor.py, you are importing from rasa.nlu.constants import MESSAGE_ENTITIES_ATTRIBUTE “MESSAGE_ENTITIES_ATTRIBUTE” whereas in the constants file, i dont see this entry.

I can only see MESSAGE_ATTRIBUTES & ENTITIES_ATTRIBUTE

Can you please advise the right one.

4

hey @siriusraja, I had created it for the older versions so it was present in the older version, it might not be there in the newer version, I will update the code soon to support the newer version. Thanks for pointing it out. :slight_smile:

5

hey @siriusraja, I have updated the repository to support Rasa >=1.7.0,

https://github.com/JiteshGaikwad/EntityExtractor.git

2 Likes
6

hi jitesh i also try to fetch date by using your extractor but there is no any extraction took place while testing My registry.py: `from rasa_nlu.extractors.MSRTEntityExtractor import MSRTExtractor

 component_classes = [
# utils
SpacyNLP, MitieNLP,
# tokenizers
MitieTokenizer, SpacyTokenizer, WhitespaceTokenizer, JiebaTokenizer,
# extractors
SpacyEntityExtractor, MitieEntityExtractor, CRFEntityExtractor,
DucklingHTTPExtractor, EntitySynonymMapper,MSRTExtractor,
# featurizers
SpacyFeaturizer, MitieFeaturizer, NGramFeaturizer, RegexFeaturizer,
CountVectorsFeaturizer,
# classifiers
SklearnIntentClassifier, MitieIntentClassifier, KeywordIntentClassifier,
EmbeddingIntentClassifier

]

registered_pipeline_templates = { “pretrained_embeddings_spacy”: [

    "SpacyNLP",

    "SpacyTokenizer",

    "SpacyFeaturizer",

    "RegexFeaturizer",

    "MSRTExtractor",

    "CRFEntityExtractor",

    "EntitySynonymMapper",

    "SklearnIntentClassifier"

    

],....}`

And i mention in pipeline like this pipeline: "supervised_embeddings

But while testing no any extraction took place

sent
sent1905×211 18.4 KB

7

Hey @franklin, welcome to the community :blush:. I have just added the support for training not for testing.:sweat_smile:

8

Hey @JiteshGaikwad just read out your solution. Can you help me out with testing my rasa model? I am not able to extract entity that i have not trained. Location, Person and Date entity should be fetched even if i dont train my data since i am using Spacy pipeline. If i am wrong somewhere, help me out with it. :slight_smile:

9

@jainsaurav22 spacy doesn’t works every time for above mentioned entities better you try duckling.

10

Hi Jitesh, while training the model, i am receiving this error. Training NLU model… Traceback (most recent call last): File “h:\azurebot\bot\chatbot\venv\lib\site-packages\rasa\nlu\registry.py”, line 155, in get_component_class return class_from_module_path(component_name) File “h:\azurebot\bot\chatbot\venv\lib\site-packages\rasa\utils\common.py”, line 198, in class_from_module_path m = importlib.import_module(module_name) File “C:\Users\gt984v\AppData\Local\Programs\Python\Python36\lib\importlib_init_.py”, line 126, in import_module return _bootstrap._gcd_import(name[level:], package, level) File “”, line 994, in _gcd_import File “”, line 971, in _find_and_load File “”, line 955, in _find_and_load_unlocked File “”, line 665, in _load_unlocked File “”, line 678, in exec_module File “”, line 219, in _call_with_frames_removed File “H:\azurebot\bot\chatbot\MSRTEntityExtractor.py”, line 11, in from rasa.nlu.constants import ENTITIES ImportError: cannot import name ‘ENTITIES’

11

@guru1994 can you give me the rasa version details. by running the below command:

rasa --version

12

@JiteshGaikwad, i am using below rasa version…

rasa-sdk==1.7.0 rasa==1.7.3

13

@JiteshGaikwad , my rasa version is 1.7.3. can you please help me to implement MSRTextractor?

(venv) H:\azurebot\bot\chatbot>rasa --version Rasa 1.7.3

(venv) H:\azurebot\bot\chatbot>

14

Hi Jitest, can you tell me what are the changes to be done to it in the rasa 2.x version. I have a file in the older version but it shows an error while running rasa train