Hi everyone,<p>Not happy with the current news aggregators and thrilled about ML models for text processing, a colleague and I built an aggregator fully based on Doc2Vec: https://www.quicknews.ai/.<p>We currently poll ~120 news sources every 15 minutes, and construct a personalized news feed for every user.<p>There are plenty of ways in which Doc2Vec is used here:
1) Clustering the articles into topics
2) Deduping overly similar articles when generating the feed
3) Routing notifications about breaking news to interested users
4) Calculating similarity between domains/publishers, so as to recommend sources similar to the ones the user already clicked on
5) Dislike (only on Android, and somewhat fickle)<p>I'd be delighted if you tried it out and let me know what you think. Available on Android and iOS, with a web version planned.<p>Android app: https://play.google.com/store/apps/details?id=com.quicknews.app<p>iOS app: https://apps.apple.com/ca/app/quicknews-the-real-news-app/id1489638005
Sounds interesting! Will see how it works out. Ran into an issue: when I clicked an article I wanted to read, "swipe left" was registered, instead. Good that you have an "undo" option, but maybe the "swipe threshold" is too low?