Inspiration
*Autocomplete * a predictive search feature recommends the next probable word user will type as a suggestion. This feature helps to quickly select the word and reduce the typing duration which in turn adds credit to user experience.
What it does
Predicts the next probable word user will type to query.
How we built it
Built using Python,Tensorflow,colab,LSTM
Challenges we ran into'
Tried to train with custom data - Code mantra Policy document.Due to limited data, 1750 unique words, model started to overfit. Hence training with generic dataset containing 581887 words.
Accomplishments that we're proud of
The model can predict the fifth word based on sequence of 4 words seen. The model is trained in a Supervised way containing previous words as features and the next word as Label.
What we learned
This idea can be extended to custom datasets for accurate results pertaining to business domain.
What's next for Generic Autocomplete Feature
Model is trained with 5 epochs and 3704 steps accounting for an hour of model training. To increase the efficiency, further training is required.
Built With
- deeplearning
- natural-language-processing
- python
Log in or sign up for Devpost to join the conversation.