The objective of this project is to detect intent from texts. For this, a benchmark dataset is used, which includes 7 intents (Search Creative Work, get weather, Book Restaurant, Play Music, Add to Playlist, Rate Book, Search Screening Event) and 14 thousand samples. Transfer learning has been leveraged to train a machine learning model. The model takes the raw texts, which are tokenized and vectorized to feed into the pre-trained model. Overall, 98 % accuracy is achieved in detecting the intents from the trained model.

The data is collected from the following link: https://github.com/snipsco/nlu-benchmark/tree/master/2017-06-custom-intent-engines

• Tensorflow 2.0 is used.
• BERT-base model is used (12-layer, 768-hidden, 12-heads) combining with a few dropout and dense layers.
• bert-for-tf2 package is used to train.
• 98% accuracy is achieved.

This Project’s GitHub Repository

• Classification
• Transfer Learning
• NLP