I’ve been performing some more testing for my submission for the Twitter Innovation Challenge #Promote 2017.
In this short post, I further validate the SocialOpinion AI service is correctly identifying users on Twitter that are considering buying a new product or trying a new service.
For context, a Campaign was setup to track all models of “iPhone”
I then ran the SocialOpinion service which processes incoming tweets. The NLP and POS tagging features process each token (words), determining sentiment, noise words and POS tags continuously.
And now for the value prop.
What follows are code extracts where the service has identified users on Twitter expressing the desire to purchase a new iPhone.
You’ll see there is a value 0.70, this is a probability.
In plain English:
If the AI determines there is a 70% chance the user is expressing the desire to purchase a new iPhone (or any product) then we want to capture the users data.
Text highlighted in red such as”iPhone 5c but mines so broke I need a new phone” clearly belong to this category.
The probability threshold is hard coded for the time being but naturally this could be configurable.
What happens next?
The user id and accompanying meta data from each tweet is encrypted and added to a Tailored Audience associated with the Campaign.
- These leads are being generated in real-time and can be served with Creatives
- Authoring Creatives (Promoted Tweets and Cards etc) is another feature I’ve built.
With about 4 weeks left until the final submission date, next steps are to collect datasets, generate metrics and complete submission documentation!