https://github.com/armandgiraud/evaluatesearch

Repository dedicated at Evaluating ranking model vs ground truth ranks
https://github.com/armandgiraud/evaluatesearch

learning-to-rank nlp

Last synced: 6 days ago
JSON representation

Repository dedicated at Evaluating ranking model vs ground truth ranks

• Host: GitHub
• URL: https://github.com/armandgiraud/evaluatesearch
• Owner: ArmandGiraud
• Created: 2020-02-07T18:11:07.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-02-08T09:48:01.000Z (almost 5 years ago)
• Last Synced: 2024-01-17T20:34:13.963Z (10 months ago)
• Topics: learning-to-rank, nlp
• Language: Python
• Size: 1.34 MB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Evaluate Semantic Search

These scripts helps evaluate the quality of a ranking model for the [Code du travail Numérique](code.travail.gouv.fr) Search Engine against manually labelled data aka the [Datafiller](https://datafiller.num.social.gouv.fr/).

### Install:

`pip install -r requirements.txt`

### Run on [FlauBert](https://github.com/getalp/Flaubert):

`python eval.py`

### Run on [CamemBert](https://camembert-model.fr/):

`to do`

### Run on [USE](https://tfhub.dev/google/universal-sentence-encoder-multilingual-qa/2)

 `to do`

## Evaluate your own model:

 create a new file `yourModel.py` which defines a Predictor class having a predict method

 - given a string query should return k slugs from the slugs parameters

 and a name attribute and param attributes.

 see [Flaubertmodel.py](https://github.com/ArmandGiraud/EvaluateSearch/blob/40e1fb956d93059adfe7f58a54309a516cd9fb71/Flaubertmodel.py#L25) for example