Resampling Summarizer for Social Media


Simply resample the social stream, using the auto-summarization algorithm as resampling function.


Today, we have pretty good summarization algorithms for text (for example, sumy). The social network messages are simply rows of our database. Take equidistant time intervals, and summarize the text of many posts of the chatty people, as one post (in chosen intervals of time, e.g., daily), and leave those rarer posts unsummarised.

Apparently, someone's working on something related:: "Facebook will soon allow you to combine multiple posts into one story", or maybe they are following Infinity? Just apply LDA, such a simple thing to do.

Many ideas on improving social media time-line sharing:

  1. Time-sticky posts in other people's timelines, in proportion of how long the poster hasn't posted... (i.e., filling up attention time credit)
  2. Auto-combining posts in timeframes, if there are too many posts from a single poster too close in time.
  3. Using text summarizers for combined posts (like with gensim / LDA).
  4. Limiting number of posts per time-frame per user (i.e., slow tweeting).
  5. Filter posts from friends by topic, for example, read only math posts, not cooking posts.

(suppress notifications) (Optional) Please, log in.