smart_intent

Classify query to various intents as below

1. information_intent: intention to learn more about something. Eg. what is programming?, How does docking station work ?
2. yelp_intent: intention to look for services/local businesses. Eg. italian food near me, floor repair cost
3. navigation_intent: intention to navigate from search to the website. Eg. email login, bank branch routing number
4. travel_intent: intention to travel. Eg. hotels in Paris, cruise trip to carribean islands
5. purchase_intent: intention to purchase products. Eg. price of us open tennis tickets, buy iphone 10
6. weather_intent: to know the weather or temperature of a city. Eg. weather in Miami, temperature in Seattle
7. translation_intent: to translate from and to various languages. Eg. what is hello in spanish, translate hi to japanese
8. unknown: when the intent is inconclusive

Steps to install the App

1. clone the repo
2. python -m venv venv
3. source /venv/bin/activate
4. python -m pip install -r requirements
5. streamlit run src/intent_app.py

steps to run the sample intent inference

steps 1 - 4 from above 5) python src/infer_intent.py

steps to train intent classifier model

data preparation:

python src/data_prep_intent.py

intent model training:

python src/train_intent.py

(optional) upload the model to hub:

python src/upload_intent.py

steps to train NER model

data preparation:

python src/data_prep_ner.py

NER model training:

python src/train_ner.py

(optional) upload the model to hub:

python src/upload_ner.py

Steps to quantize:

venv/bin/python convert.py --model_id Mozilla/mobilebert-uncased-finetuned-LoRA-intent-classifier --quantize --modes q4 q8 fp16 --task text-classification  --output_parent_dir output_models

venv/bin/python convert.py --model_id Mozilla/distilbert-uncased-NER-LoRA --quantize --modes q4 q8 fp16 --task token-classification  --output_parent_dir output_models