Word of mouth is one of those impossible things people like to track. The best people seem to be able to do is to ask the question on a registration form, “Where did you hear about us?”
Is it possible to track word of mouth any better? This is the premise of my latest project, Conversation Clouds.
Imagine a conference on multiple floors, with thousands of people. You stand at a screen visualizing all the conversations going on simultaneously. And there on the screen is the general location of a topic you’re interested in, so you venture down a floor and join in.
Now, privacy concerns aside, the system works as such:
- Microphones are placed around a physical space. Each is marked with an id which relates to its geographic location.
- The audio is transmitted to a server which performs voice to text translations in small batches.
- As each sample is translated it pushes to the visualization script which then displays on a central monitor.
In this early proof of concept I’m using recorded conversations left through answering machines which are transcribed into textual form. Once a new message arrives the script parses it through a custom UIComponent in Flex. Words are added to a hashmap and as repetition occurs the visualization increases the scale and opacity. With future versions I hope to add in a decay parameter as well, so as the conversation flows from one topic to the next so does the visualization. (In the demo above I’m using the content from a previous post.)

One Trackback/Pingback
[…] after posting the previous entry I went to sleep, only to find myself thinking about improving the visual aspect of the Conversation […]
Post a Comment