A computer scientist from Mexico City has developed Mexica, a software that understands emotion and tension well enough to compose stories.

In an Internet survey where the Mexica's stories were pitted against human and other computer-generate stories, the program of Rafael Perez y Perez was ranked highest in flow, coherence, structure, content, suspense, and overall ... quality.

Rafael, who has a really cool name, works for the Autonomous Metropolitan University in Mexico. He explains that his program begins with a very simple story, a few sentences outlining a beginning, a middle, and an end. Then it treats the characters as variables and assigns numerical values to emotional links between them. After that, it conducts an engagement-reflection cycle where it searches for atoms, i.e. story actions and occurrences stored in a database. The chosen atom will be the one that best fits for the context of the character at that moment. This process is repeated until no more atom can be found. Finally, the software conducts an analysis of the coherence and interestingness. A story is considered interesting when tensions levels fluctuate throughout.

Perez y Perez explains that Mexica does not want to replace human writers, but rather to help them understand the way they write, and thus allow them to see how they can improve their writing.
Unfortunately, impossible to find this software nor a example of generated story...

Source: Discovery News, last friday.