Moodgrapher — Some background

Weblogs, or blogs, are a relatively new form of web content that is growing rapidly in recent years. These personal, highly opinionated journals publicly available on the internet have grown from a handful in the late 1990s to more than 10 million in 2005, as reported by major blog tracking services; this growth is steady and exponential. Not only the amount of blogs is on the rise, but so is their influence in terms of number of readers.

What Moodgrapher is, and how it works

LiveJournal is one of several simple-to-use personal publishing ("blogging") tool. Bloggers using LiveJournal can report their mood at the time of writing a post; about 80% of the posts indeed have a mood attached to them. Moodgrapher tracks these moods by periodically gathering recent posts from Livejournal, and checking for their "mood report". It currently tracks about 100,000 posts every day.

Two numbers are reported by Moodgrapher: the percentage of posts reporting a certain mood (the dashed, black line), and the "rate of change" of a mood — the difference between the usual amount of posts with this mood and the amount in a given hour (this is the continuous red line).

What is the motivation for doing this?

Our long-term research aim is to develop novel methods for searching, discovering and retrieving blogs. We believe that the non-factual aspect of blog entries is an important part of what makes people read and browse around blogs rather than, say, online news papers.

One direction we are interested in concerns automatic methods for predicting the "mood" of blog entries in cases where there is no explicit mood indicator (such as with various blogging tools other than LiveJournal). Originally, then, we implemented Moodgrapher for a study we are conducting on estimating the mood of a people according to the language they use.

Soon after watching the first results returned by Moodgrapher, we realized that it is a great tool for analyzing mass-behavior over time. Among the interesting phenomena revealed by Moodgrapher are:

the cyclic nature of many states-of-mind: e.g., people feel energetic in the mornings and relaxed in the evening;
day-of-the-week related behavior: people tend to get drunk on weekends, feel busy on Monday morning, etc.;
the effect of global events on the global mood: mass increase in the level of distress and sadness after terror attacks (details).

What are our plans with Moodgrapher?

We plan to keep Moodgrapher up and running for some time to come. Right now, the data gathered by Moodgrapher is being used for the validation of the automatic mood recognition work that we mentioned above; this, in turn, will at some point be used in innovative web search scenarios that we're working on. In addition, we think that Moodgrapher's output offers a broad spectrum of research opportunities, ranging from marketing to cognitive science.

Comments and questions are welcome here.

Back to Moodgrapher.