CRFEntityExtractor or DIETClassifier is splitting a entity into multiple if contains punctuation

[ridhimagarg]

Rasa 1.10.14:

Python 3.7.5:

Ubuntu 19.10 :

Issue:

CRFEntityextractor or DIETClassiifer is splitting an entity into multiple words.
I was using RASA 1.10.2 initially and it was working fine but the ConveRTFeaturizer issue reported in RASA 1.10.2 for the POLY-AI model URL changed. Hence I have updated to RASA 1.x latest version i.e, RASA 1.10.14

Now these CRFEntityextractor or DIETClassiifer are not identifying the entities propely.

Error:

5306 Walnut Ave., Building A, Sacramento, CA 95841
{
  "intent": {
    "name": "address",
    "confidence": 0.9996100664138794
  },
  "entities": [
    {
      "entity": "insuredaddress",
      "start": 0,
      "end": 15,
      "confidence_entity": 0.8947956045473241,
      "value": "5306 Walnut Ave",
      "extractor": "CRFEntityExtractor"
    },
    {
      "entity": "insuredaddress",
      "start": 18,
      "end": 28,
      "confidence_entity": 0.9364809066380787,
      "value": "Building A",
      "extractor": "CRFEntityExtractor"
    },
    {
      "entity": "insuredaddress",
      "start": 30,
      "end": 40,
      "confidence_entity": 0.9457472912523117,
      "value": "Sacramento",
      "extractor": "CRFEntityExtractor"
    },
    {
      "entity": "insuredaddress",
      "start": 42,
      "end": 50,
      "confidence_entity": 0.9197204154102323,
      "value": "CA 95841",
      "extractor": "CRFEntityExtractor"
    },
    {
      "entity": "insuredaddress",
      "start": 0,
      "end": 15,
      "value": "5306 Walnut Ave",
      "extractor": "DIETClassifier"
    },
    {
      "entity": "insuredaddress",
      "start": 18,
      "end": 28,
      "value": "Building A",
      "extractor": "DIETClassifier"
    },
    {
      "entity": "insuredaddress",
      "start": 30,
      "end": 40,
      "value": "Sacramento",
      "extractor": "DIETClassifier"
    },
    {
      "entity": "insuredaddress",
      "start": 42,
      "end": 50,
      "value": "CA 95841",
      "extractor": "DIETClassifier"
    },
    {
      "entity": "GPE",
      "value": "Walnut Ave",
      "start": 5,
      "confidence": null,
      "end": 15,
      "extractor": "SpacyEntityExtractor"
    },
    {
      "entity": "ORG",
      "value": "Building A",
      "start": 18,
      "confidence": null,
      "end": 28,
      "extractor": "SpacyEntityExtractor"
    },
    {
      "entity": "GPE",
      "value": "Sacramento",
      "start": 30,
      "confidence": null,
      "end": 40,
      "extractor": "SpacyEntityExtractor"
    },
    {
      "start": 0,
      "end": 4,
      "text": "5306",
      "value": 5306,
      "confidence": 1.0,
      "additional_info": {
        "value": 5306,
        "type": "value"
      },
      "entity": "number",
      "extractor": "DucklingHTTPExtractor"
    },
    {
      "start": 45,
      "end": 50,
      "text": "95841",
      "value": 95841,
      "confidence": 1.0,
      "additional_info": {
        "value": 95841,
        "type": "value"
      },
      "entity": "number",
      "extractor": "DucklingHTTPExtractor"
    }
  ],
  "intent_ranking": [
    {
      "name": "address",
      "confidence": 0.9996100664138794
    },
    {
      "name": "affirm",
      "confidence": 7.791238749632612e-05
    },
  ],
  "response_selector": {
    "default": {
      "response": {
        "name": null,
        "confidence": 0.0
      },
      "ranking": [],
      "full_retrieval_intent": null
    }
  },
  "text": "5306 Walnut Ave., Building A, Sacramento, CA 95841"
}

If I entered without punctuation then it's working -:

5306 Walnut Ave Building A Sacramento CA 95841                
{
  "intent": {
    "name": "address",
    "confidence": 0.9991991519927979
  },
  "entities": [
    {
      "entity": "insuredaddress",
      "start": 0,
      "end": 46,
      "confidence_entity": 0.9889847823938246,
      "value": "5306 Walnut Ave Building A Sacramento CA 95841",
      "extractor": "CRFEntityExtractor"
    },
    {
      "entity": "insuredaddress",
      "start": 0,
      "end": 46,
      "value": "5306 Walnut Ave Building A Sacramento CA 95841",
      "extractor": "DIETClassifier"
    },
    {
      "entity": "FAC",
      "value": "Walnut Ave Building",
      "start": 5,
      "confidence": null,
      "end": 24,
      "extractor": "SpacyEntityExtractor"
    },
    {
      "entity": "GPE",
      "value": "Sacramento",
      "start": 27,
      "confidence": null,
      "end": 37,
      "extractor": "SpacyEntityExtractor"
    },
    {
      "entity": "CARDINAL",
      "value": "95841",
      "start": 41,
      "confidence": null,
      "end": 46,
      "extractor": "SpacyEntityExtractor"
    },
    {
      "start": 0,
      "end": 4,
      "text": "5306",
      "value": 5306,
      "confidence": 1.0,
      "additional_info": {
        "value": 5306,
        "type": "value"
      },
      "entity": "number",
      "extractor": "DucklingHTTPExtractor"
    },
    {
      "start": 41,
      "end": 46,
      "text": "95841",
      "value": 95841,
      "confidence": 1.0,
      "additional_info": {
        "value": 95841,
        "type": "value"
      },
      "entity": "number",
      "extractor": "DucklingHTTPExtractor"
    }
  ],
  "intent_ranking": [
    {
      "name": "address",
      "confidence": 0.9991991519927979
    },
  ],
  "response_selector": {
    "default": {
      "response": {
        "name": null,
        "confidence": 0.0
      },
      "ranking": [],
      "full_retrieval_intent": null
    }
  },
  "text": "5306 Walnut Ave Building A Sacramento CA 95841"
}

Content of configuration file (config.yml) (if relevant):

NLU pipeline

language: en
pipeline:
  - name: SpacyNLP
    model: en_core_web_md
  - name: ConveRTTokenizer
    "model_url": "https://github.com/PolyAI-LDN/polyai-models/releases/download/v1.0/model.tar.gz"
  - name: ConveRTFeaturizer
    "model_url": "https://github.com/PolyAI-LDN/polyai-models/releases/download/v1.0/model.tar.gz"
  - name: RegexFeaturizer
  - name: LexicalSyntacticFeaturizer
  - name: CRFEntityExtractor
  - name: DIETClassifier
    epochs: 100
  - name: EntitySynonymMapper
  - name: ResponseSelector
    epochs: 100
  - name: SpacyEntityExtractor
  - name: "DucklingHTTPExtractor"
    url: "http://0.0.0.0:8000"
    locale: "en_GB"
    timezone: "US/Pacific"
    timeout : 3

[sara-tagger]

Thanks for the issue, @koaning will get back to you about it soon!

You may find help in the docs and the forum, too 🤗

[ridhimagarg]

Thanks, Waiting for the solution.

Required an urgent action on the same. I am using RASA in production and it stuck!

Thanks in advance!

[akelad]

Hi @ridhimagarg, this behaviour is intentional as of PR #6191. We recognise that there might be different use cases like yours though, so I've opened another issue #6852 to address this.

For now, the solution would be to combine the various different entities and fill a slot with them in a custom action. Or you could use a list slot, which will store all the values in a list automatically.

[ridhimagarg]

Thanks @akelad

It would be great if this enhancement is been done.