Built to help automate and monitor online communities mentioning certain keywords in realtime, a bespoke system called Radar, uses the latest technologies in machine learning, natural language processing, laravel combined with VueJS.
The system communicates with a series of Radar scanners (raspberry pi) which load selected websites in a browser, which is controlled via a bespoke plugin that periodically checks for the given keywords.
If a match is found, an alert is sent back to the system. From here, the system will perform some cleaning tasks; such as encoding, filtering out rude words and discarding posts containing duplicates or retweets. The next task is to broadcast the new alert to users logged in. If e-mail alerts have been selected then an email will be sent, notifying the user of the new alert. Similarly, the alert will be sent to the chosen API if the option is selected.
A number of scheduled tasks also run periodically; one of the priority tasks involves performing predictions on whether the alert is interesting based on whether previous alerts of a similar ilk have been marked as interesting. Another task performs sentiment analysis which rates the emotion of the alert to detect whether it is negative or positive. One of the last tasks checks the system performs, is to define whether any of the new alerts are similar to any of the last 100 alerts – this is useful for detecting trends.
* There are a number of tasks which haven’t been mentioned to protect the intelligence of the system
If you’re looking for bespoke systems to automate your business processes, then please feel free to use my contact page here.