Summarizer - Steam Reviews

A team of neural nets is trained on millions of Steam game reviews. The nets learn which reviews are informative, which sentences are important, and so on. They are asked to extract the voice of the people, for any given product. And then write a summary in eight sentences.

It is a tall order: to extract the essence of thousands of scribbles. Concocted at different times, by people from different walks of life, with different education or prejudice, of different mood. To top it off, these chaotic ingredients have to be "cooked" into a readable essay. In milliseconds.