How does this work?
Glad you asked.
The system works in three separate steps:
1. Every day we generate and expand the social graph of the Indieverse, adding connections between people depending on who they interact with (mentions, retweets, replies…)
2. After that, we monitor the activity on the social graph and index what is being said (mostly tweets and replies). Every tweet is then assigned a score based on its level of actionability (e.g. is the tweet a question? Is it a request for help?) and the stats of the tweet itself (number of retweets, quotes, preferred…)
3. Once the scoring algorithm has run, we let the indexed tweets be full-text searchable by the user. Results are then presented combining the level of matching of the query with our internal tweet score.
You can think of IndieTweets as a self-updating tracker of the activity of a community (in this case, Indie Makers!). It is self-updating as it adds new people to follow based on the interactions of the pre-existing network of people.
The current list of people in our ever expanding social graph was built seeding the algorithm with around a hundred hand-picked Twitter accounts belonging to well-connected Indie Hackers. Starting from these accounts, we let the algorithm monitor their activity for about a month and we expanded our internal social graph by adding accounts which consistently interacted with our core users.
Rinsing and repeating you get a critical-mass social graph (fancy way of saying that you have enough people to catch the geist of a community) which allows the final user to engage in highly relevant conversations and discover new interesting people to follow.