Rasa spacy enity extractor

am trying to build rasa spacy for entity extraction using regular expression:

where i have .json file containg this regex for entiy extraction the things i want to extract is: input text --> from = john and to = rishi i want the entities as [(‘from’,'john),(‘to’,‘rishi’)] in short the word before and after the ‘=’ sign

{“label”:“PROGLANG”,“pattern”:[{“LOWER”:{“REGEX”:"((\S+)\s?=\s?(\S+))"}}]}

and the pipeline am using is : language: en

pipeline:

  • name: SpacyNLP model: “en_proglang”
  • name: SpacyTokenizer
  • name: SpacyEntityExtractor dimensions: [“PROGLANG”]
  • name: SpacyFeaturizer pooling: mean
  • name: CountVectorsFeaturizer analyzer: char_wb min_ngram: 1 max_ngram: 4
  • name: DIETClassifier epochs: 1

policies:

  • name: MemoizationPolicy
  • name: KerasPolicy
  • name: MappingPolicy

not able to extract the entity

Hi Pemitra,

would I be correct to assume that you’re following this guide? Did you export a custom spaCy model?

I think there’s nothing wrong with your config.yml but it’s hard for me to parse because you’re not using the markdown syntax for code blocks. You can get a pretty render going by using three ticks (```) before and after the code.

like this

It also deserves mentioning that in Rasa 2.0 we’ll have tools that will allow you to grab entities using Regex without spaCy. You can already play with these features if you install the new alpha version.