Values of each entity are specified in a separate file with the name corresponding to the entity name
e.g. first_name.csv
There is one entity value per row of the file. The value can be followed by its synonyms separated by semicolumns
Example:
amy
john;johny;johnny;ian
susanne;susan
Recommendation: Entities files are typically placed to a separate directory <app_root>/entities or directories.
For pattern entities the pattern should be prefixed by ~
Example:
~email;\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b
~fullUSphone;(\d{3})-(\d{3})-(\d{4})
Contextual entities are annotated on intent examples using XML tags. More than one entity can be annotated in one example.
Example:
I'd like <color>red</color> dress
i like <color>blue</color> and <color>green</color>
Annotations cannot be nested and only the most inner ones are taken into account. Validations is provided for matching tags. When there are invalid annotation the script exits with error. The functionality is covered by intents_csv2json.py script. Currently only one way conversion is provided (CSV -> JSON).
Only english letters, numbers, hyphens and underscores are allowed in entities names.
You are supposed not to use entities names starting with sys- as Conversation tool offers some system entities with this prefix.
Following are only recommendations, intent name has no impact on the behavior of the conversation.
- The entity name should be lower-cased
- There shouldn't be any domain prefix as entity is usually cross-domain (e.g. color)
Example:first_name