Topics

The Kred Topic scoring engine parses posts into keywords and keyword clusters (named entities) by using the Python NLTK platform and the Wordnet lexical database. The named entities are structured into a hierarchy of topics using a combination of the associations from Wordnet and relative scores.

Topic points earned for an individual post are calculated and applied to the users score in each topic in the hierarchy. A leaderboard of the highest scoring users is then constructed for each topic.

For example:

I’ve been a Miami Dolphins fan since 1985. I have NEVER seen the fan base so excited after a draft.
Now, let’s see these young men on the field because that’s when the party will really be kicking!!!!
#FinsUp!!!!!!!!!!!!!!!

From this we derive the following Topics:

Miami Dolphins, NFL, Sport
Miami, Florida, USA

Another example:

Male bottlenose dolphins call each other by name and touch frequently to reinforce friendships, which can last a lifetime.

From this we derive:

Dolphin, Mammal, Sea Life

Topics are currently only supported for English language content, with support for other languages in development.