Assistant Intent Classification

The classifer repo is split into multiple sub modules:

• fuzzy_similarity: Tests fuzzy checking of words with text
• infersent_cosine_similarity: Test trigger similarity via HuggingFace and Infersent
• location_recognition: Location recognition for weather
• many_intent_classifier: Trying to classify lots of intents at a time
• qa_context_evaluation: Tests using qa models for context
• shopping_classifier: Finds shopping items mentioned

Running the notebooks

Download Glove for word embeddings

mkdir GloVe
curl -Lo GloVe/glove.840B.300d.zip http://nlp.stanford.edu/data/glove.840B.300d.zip
unzip GloVe/glove.840B.300d.zip -d GloVe/

and Download infersent for sentence embeddings

mkdir encoder
curl -Lo encoder/infersent1.pkl https://dl.fbaipublicfiles.com/infersent/infersent1.pkl

Run the classifier notebook classifier.ipynb to load and view the data.

Running the Server

Current Accuracy:

Download deepspeech-0.7.3-models from deepspeech directory. The files that should be included are

├── deepspeech-0.7.3-models.pbmm
├── lm.binary
├── output_graph.pb
├── output_graph.pbmm
├── output_graph.tflite
└── trie

Download + Accessing Colabs

Copy the model files for the relevant classification system https://drive.google.com/drive/folders/1gzlJk7iL88Fpe2OzW7NQI_9iSyIqx39B?usp=sharing Download ffc_keras_model and infraset_model.torch from the notebook