How Twitter real time search achieves human like relevance and context?

Twitter has explained how the real-time search works and ways of using real-time searches to get relevant advertisement for each of search query. It is explained in great detail and is bit vague to understand with various jargon floated around.

twitter real time search

Source:cache.clickonf5.org

This is my attempt at explaining the whole process in a simplified manner.

1.User makes a search in Twitter.

2.In the back ground,Twitter monitors each such keyword search and displays results from automated search result algorithm similar to how Google uses search algorithm.

3.But if the same keyword is searched for multiple times and the volume of search increases for that keyword,instead of automated search results, a different process is followed so as to get more contextual and real-time relevance.

4.Twitter sends such keywords search request  to Amazon’s ‘Mechanical Turk‘ (Amazon’s crowd-sourcing online platform for small jobs). So when Twitter sends these search queries,people would analyze and tag them with appropriate photos,articles or any relevant videos etc.

5.Once the response is received,Twitter gets more context to the search query,it has a human context and relevancy to search keyword.Finally Twitter returns the search result with the usual search result algorithm and in addition to that all the relevant pictures , links , videos tagged to search queries making the search results more relevant through the  combination of both human and machine based search algorithm.

Also Twitter will also add any relevant advertisement to these search query results there by making money from search results similar to Google. Only difference here is that the search results are real-time.

Sounds amazing right?

Advertisements