Project

General

Profile

Actions

dev #1713

open

POC on Identifying Attributes from Natural Language (NL) Query

Added by Zahid Hassan about 1 year ago.

Status:
Complete
Priority:
High
Assignee:
Target version:
-
Start date:
01/16/2025
Due date:
01/16/2025 (about 15 months late)
% Done:

100%

Estimated time:
4:00 h
Spent time:

Description

Findings

  • We need to use a Named Entity Recognition (NER) Model to extract attributes from natural query.
  • Fine tune and Train the NER model with our custom dataset of NL queries.
  • Initially selected the large NER model from Spacy .
  • We need to annotate our custom dataset manually because Spacy only accepts a pre defined structured dataset.

Completed Tasks

  • Downloaded and used the pre trained large NER model.
  • Generated 500 records of NL queries related to AI tools using GPT.

Next Tasks

  • Annotate data following spacy's train/test data structure.
  • Train the pre trained model with this data.
Actions

Also available in: Atom PDF