Silent Movie by Richard Gatarski

Psychographic text analysis is about finding linguistic signal patterns to try to determine the psychological profile of the author of a text. Today (Update: this english version cerated 13th sep) version at Sweden Social Web Camp I presented a small demo of my integral model for determining sentiment (mood), values and perspective (type).

During a little more than a month I´ve collected tweets from the participants of the unconference and ran them through the tool. The data set consisted of about 30 000 tweets from 269 individuals.

The methodology I´ve developed looks at signals from three different psychological levels: perspective/type (which rarely change), values (that changes in different periods in life) and sentiment/mood (which frequently change). The purpose of the research is to aggregate large amounts of signals from large amounts of data to find macro-patterns, which is a different use case than trying to determine the meaning of single sentences, as in e.g. media sentiment-analys.

I´ve been help by volonteers to build a crawler, analysis engine and database making use of my linguistic model. The diagram shows that the categories are very stable over time, which either means that the models are wrong and mainly show normal statistical frequencies in language or that there are actually surprisingly stable use of certain language patterns in a group culture such as social media geeks. (update: I´ve found some errors in my models that might adjust these phenomena).

On the following pages you can view the data from the demo session (Note: you might have to adjust the settingsto make them fully comparable):
Type and sentiment analysis


Values (Spiral Dynamics)

A big thank you to the audience at the session for feedback and interesting discussions about applications!