license: apache-2.0
Model Overview
This is a text classification model to classify documents into one of 26 domain classes:
'Adult', 'Arts_and_Entertainment', 'Autos_and_Vehicles', 'Beauty_and_Fitness', 'Books_and_Literature', 'Business_and_Industrial', 'Computers_and_Electronics', 'Finance', 'Food_and_Drink', 'Games', 'Health', 'Hobbies_and_Leisure', 'Home_and_Garden', 'Internet_and_Telecom', 'Jobs_and_Education', 'Law_and_Government', 'News', 'Online_Communities', 'People_and_Society', 'Pets_and_Animals', 'Real_Estate', 'Science', 'Sensitive_Subjects', 'Shopping', 'Sports', 'Travel_and_Transportation'
Model Architecture
The model architecture is Deberta V3 Base
Context length is 512 tokens
Training (details)
Training data:
- 1 million Common Crawl samples, labeled using Google Cloud’s Natural Language API: https://cloud.google.com/natural-language/docs/classifying-text
- 500k Wikepedia articles, curated using Wikipedia-API: https://pypi.org/project/Wikipedia-API/
Training steps:
Model was trained in multiple rounds using Wikipedia and Common Crawl data, labeled by a combination of pseudo labels and Google Cloud API.
How To Use This Model
Input
The model takes one or several paragraphs of text as input.
Example input:
q Directions
1. Mix 2 flours and baking powder together
2. Mix water and egg in a separate bowl. Add dry to wet little by little
3. Heat frying pan on medium
4. Pour batter into pan and then put blueberries on top before flipping
5. Top with desired toppings!
Output
The model outputs one of the 26 domain classes as the predicted domain for each input sample.
Example output:
Food_and_Drink
Evaluation Benchmarks
Accuracy on 500 human annotated samples
- Google API 77.5%
- Our model 77.9%
PR-AUC score on evaluation set with 105k samples
- 0.9873
References
https://arxiv.org/abs/2111.09543 https://github.com/microsoft/DeBERTa
License
License to use this model is covered by the Apache 2.0. By downloading the public and release version of the model, you accept the terms and conditions of the Apache License 2.0.