metadata
tags:
- spacy
- token-classification
- named-entity-recognition
- text-to-sql
- schema-linking
language:
- en
license: mit
Text-to-SQL spaCy NER Model
This is a custom spaCy NER model for Text-to-SQL query understanding.
It detects:
- AGGREGATION
- METRIC
- FILTER_LOCATION
- FILTER_YEAR
- FILTER_DATE
- FILTER_CATEGORY
- FILTER_VALUE
- GROUP_BY
- ORDER_BY
- LIMIT
- COMPARISON
Example:
Input:
In 2015, how many complaints about Billing disputes were sent by clients in Portland?
Expected entities:
- 2015 -> FILTER_YEAR
- how many -> AGGREGATION
- complaints -> METRIC
- Billing disputes -> FILTER_CATEGORY
- Portland -> FILTER_LOCATION
Current baseline:
- Overall NER F1: 0.77