Custom Entity Linking

Description

This component is an Annotator that sematically links entities detected in user utterances. Entites then bound via relations.

Relation examples:

• favorite animal
• like animal
• favorite book
• like read
• favorite movie
• favorite food
• like food
• favorite drink
• like drink
• favorite sport
• like sports

I/O

Input Takes a list of user_id, entity substring, entity_tags

An input example:

"user_id": [user_id],
            "entity_substr": [entity_substr_list],
            "entity_tags": [entity_tags_list],
            "context": [context],
            "property_extraction": [property_extraction],
        }

Output:

• List of entity substrings (entity names).
• List of entity IDs.
• List of confidences.
• List of entity ID tags (entity kinds).
• List of property extraction information.
• List of preprocessed conversation

Dependencies

• annotators.ner
• annotators.entity_detection
• annotators.spacy_nounphrases