How Facebook automatically analyzes posts

NetEase technology News July 1, according to foreign media reports, Facebook has exposed its researchers have manipulated Facebook's "information Feed", artificially to some users to show more positive or negative content. So how do Facebook researchers determine whether a post is positive or negative?

They use an automated text analysis software program called the language Acquisition and vocabulary count (linguistic Inquiry and Word COUNT,LIWC). The software, which collects more than 4,500 words and roots, can help researchers dig and analyze the emotions contained in the text.

According to Facebook's research report, if a post contains at least one word that is LIWC categorized as positive, the post is identified as positive, whereas if the post contains at least one word that is LIWC classified as negative, the post is identified as negative.

For example, LIWC the "weeping (cried)" As a sad emotion, while sadness is one of several negative emotions, such as "Anger (annoyed)" classified as angry emotions. Words such as "perhaps (cames), possibly (perhaps) or guessing (guess)" indicate that the content of the post is tentative.

The software is able to evaluate the contextual context in which words are used by analyzing the part of speech. Developers of such software say they are working hard to get the software to solve the ambiguity problem on the basis of evaluating hundreds of thousands of sentences.

More and more companies and researchers are using software such as LIWC to analyze the sheer volume of content in social media. Marketers want to know how users evaluate their brands, and academics want to study people's behavior and social trends by tapping into social media content.

LIWC's website shows that the software was developed in the early 90. Based on researchers ' decades of research, the software found that people's physical and mental state can be predicted by the vocabulary they use. The researchers also found that writing can affect people's emotional state. (Chun)