Just simple language modelling and sentiment analysis over IMdB dataset. Can be used to classify review videos t#oo.
About the Universal Language Model Fine Tuning (ULMFiT)
The model is one of the state of the art text analysis model. It is a relatively simpler model compared to giants like Google's BERT or OpenAi's GPT. Useful for transfer learning on custom data (for those who have worked with transfer learning with Image Classifiers, yes, it is the ImageNet of NLP). The feature that captivates me about this model is its rather simplicity (those familiar with DL algorithms know the power of simplicity), which makes it a possible candidate when it comes to deploying DL model in real life.
It is implemented broadly in two steps:
- Language Model Construction.
- Tuning for a particular task.
Natural Language Processing (NLP) tasks possible with this model:
- Sentiment Analysis (performed in this example)
- Question Classification
- Topic classification
I am open to suggestions/helping anyone on where and how to apply this model.