The marketing department of a hedge fund, with a large customer mailing list.
The marketing department was looking to improve the content of its email newsletter. In particular, it wanted to include trending topics in the financial world, and asked Faculty to develop a tool that could detect the topics covered by a number of financial influencers on Twitter.
Using Faculty Platform, we built a machine learning model capable of collecting and analysing tweets from a set of financial influencers’ Twitter activity. The challenge was to convert unstructured data into recognisable and usable topics to make it easy for the client to pick up. We used a technique that falls within Natural Language Processing (NLP) known as Latent Dirichlet Allocation (LDA) to build a body of tweets. We then used LDA to decompose the data set into two matrices: one containing the topics and the other containing the weights of each topic for each tweet. Using this information, the tweets were ranked to highlight the importance of different keywords and to extract and display the most relevant tweets for each topic.
An example of a topic that the model might pick up, together with meta data.
The model was successful and has been put into production as a Faculty API. The program automatically analyses each day’s fresh set of tweets in the background, which the client can query and convert into an output via the API. The output is displayed on a dashboard, which makes it easy for the client to monitor and decide which topics to include in its newsletter.