Google Research competition: Scoring Q&A algorithms

Using Bi-Directional LSTM with Convolutional Layers

I competed in the Question-Answering Labelling Competition, and emerged top 15% out of 1571 participants.

Deep learning using Keras, PyTorch

Multi-target task: Predict 30 targets using 1 model

Text embeddings extraction using GloVE, BERT

About

Question-answering is a discipline within the field of natural language processing that is concerned with building systems that automatically answer questions posed by humans in a natural language.

This project builds a deep learning model that predicts the quality of automated answers to human-posed questions, taking only text features of the question title, question body, and answer.

Project Methodology

1. Preprocess text features
2. Create numeric representations for text features using state-of-the-art pre-trained algorithms: GloVe, BERT
3. Model architecture: LSTM with Convolutional layers
4. Model tuning: Increase model regularization
5. Model tuning: Increase model complexity

Still curious?

Check out this project on my website here :)