Scope

The goal of our machine learning model is going to be to translate a query string into a vector of latent factors. This is a classic regression problem, but with the twist of unstructured text data. Therefore, it will require the use of an NLP (Natural Language Processing) model.

Challenge

The issue with using NLP models for regression tasks on unstructured data is that they are primarily designed for classification and language understanding tasks. It's not inherently equipped for regression, which involves predicting continuous numerical values instead of discrete categories or labels. To adapt an NLP for regression tasks, additional layers and modifications are required to transform its output into a format suitable for predicting continuous values.

Furthermore, it’s going to be important to make sure that the NLP model we use is low power and free to use. As a group of students, computing power and monetary resources are limited.