feature request: historic news sentiment

[M00NSH0T]

I love that you added news sentiment, but I can't use it without historic data to use for training a model. Ideally, I'd like to be able to get historic timestamped articles each with a sentiment score, as opposed to aggregated weekly scores, and then subscribe via a websocket connection to get live sentiment scores for newly published articles. This is what would be needed to successfully use news sentiment in a machine learning driven trading algorithm.

You might want to consider reaching out to FinSentS.com. For months, I tried building a news sentiment trading algorithm with their historic data available via Quandl, but the historic data was always a day old and the live feed wasn't aggregated in the same way, so there was no way to train a model and deploy to production.

It would be great if either you or them (or both of you partnering), could bridge the gap and create an easily accessible historic database of news sentiment scores that can be used to train a model which could run in real-time using data in the same format.

[finnhubio]

We will look into it. Thanks for the detailed feedback

[RobertoFriasN]

Hi, are there any updates regarding this topic? I've run into the same issue and also think it would be quite useful to access historic news sentiment data, in a similar fashion that company-news allows.