The problem Dune & Lens Dashboard solves

When looking at the total number of posts, people just have a glimpse of Lens. The full picture includes the content, the sentiment around it and topics. We developed a solution that tracks the sentiment around the posts in Lens and also infers the topics around them. We that information we can also forecast future posts.

Challenges we ran into

Pagination on the Lens GraphQL endpoint was a minor issue. Time constraints also limited the complexity of analysis we could perform using Machine Learning, hence settled for NLP models based on data from movies from the spacy library. An upgrade of the model would have been using modern technologies such as transformers. For infering the topics we decided for a naive solution such as wordclouds, computing the topic modelling would come in the next interation. Finally, documentation and code clean-up would also have been done.