Intact

Technologies used

For this challenge, we recognised that there would be a correlation between certain words from the description that correspond to the medical specialty. We attempted to remove generic words such as “the”, “of”, “as” and more in search of finding occurrences of other words that could highlight a more direct relation between particular words from the description and medical specialty.

For training the model through neural network, we used concepts of NLP to tokenise words and created the model with Keras framework with 80-20 Train-Test split. The data used was read through the CSV file provided and we tried testing the model with different epochs to understand the point at which highest accuracy is obtained.

Challenges faced

It was a new experience for our team to learn NLP and the associated concepts with Machine Learning. It was tough for us to directly apply the newly learned concepts with the dataset and ran into several challenges right from the start of reading the CSV files, organising data and even cleaning the data.

Overall outcomes

The final outcome from the trained model was that it could read in a certain input such as the description and provided the predicted medical specialty from the trained dataset. Along with that, our team was also able to learn basic and fundamental concepts of Machine Learning and NLP.

Built With

• keras
• natural-language-processing