Social Media
Sentiment Analysis Data

How We Do It

We ingest real time social media data from a large number of sources.

Our proprietary deep learning models assess the relevance of each topic, filter out the noise, compute and display sentiment data and other metrics.

To achieve the final class predictions our AI models use a many-to-one architecture for text modeling. This means that while the data used is a sequence of words, the output is a class label, positive or negative, more specifically the probability that the text is positive, on a scale from 0 to 1.

As an architecture we use ensembles of LSTM networks to analyse each text and output the prediction, and this is done at the end of the day for each company listed in the Available Stocks section below.