dcsimg
RealTime IT News

Microsoft Tools Separate Friends and Flamers

Joining an online message board or newsgroup discussion can be risky considering it's sometimes hard to tell who's your friend and who's your foe. Microsoft says it is working on a way to tell the difference.

Microsoft Research sociologist Marc Smith Tuesday said the company is building graphical tools that could give users new insight into what's going on in newsgroups. By looking at the pattern of one author's posts, for example, you could tell whether she was a helper or a flamer. You could discover new groups that matched your interests or quickly figure out which ones to avoid like the plague.

Smith, head of Microsoft Research's Community Technologies Group, presented his group's work Tuesday at Microsoft's Silicon Valley campus.

"We're attempting to build tools that visualize and improve the interfaces to conversational databases," Smith told InternetNews.com, such as newsgroups and online bulletin boards. "It's always been my goal to be able to back away and see Usenet in its entirety."

While not a specific application yet, Smith points to Microsoft's NetScan as an early evolution of his goal. The platform creates graphs and charts showing posters, replies and unreplied messages for a particular newsgroup. It can create a report card for the news group showing such things as the average number of message, the number of returnees, the number of people who replied to messages, the largest threads and the most consistently present authors. It also makes a graphical map of how newsgroups are linked by cross-posted messages.

"Our tools kick in when there's too much data to evaluate. Machines have gotten us into this problem, they're going to have to get us out," Smith said.

Data mining of patterns of participation can tell you plenty, at least if you're a sociologist like Smith. He says, for example, that he can tell which threads are flames just by how long the thread is. And he can make inferences about an author based on what size threads the person writes to.

"Are you the kind of person who touches a thread, says, 'The answer is on page 93, goodbye,'?"

From that kind of information alone, he said, NetScan can help users figure out what authors they want to pay attention to.

"The tools that are available today have been very spotty as far as what feedback is given to the content programmer," said Nancy Rhine, executive producer for SociAlchemy. In her work consulting with businesses on how to get the most out of online interactions, she often goes through reports manually to create data charts and visualizations.

"We have a good intuitive sense, and we can observe, but having the science of the actual data tools is so helpful. They always show up some little points that you missed. You can't do that when you're running huge communities."

Rhine said that corporations have found that they can save money and get better customer insight with online communities than with focus groups; the tools Smith is developing could be crucial for businesses. As a consultant or project manager, she said, "You can put in all your observations, all the rich anecdotal evidence, but you need to have that data."

Smith's ultimate goal would be to turn the research project into a broadly-available platform.

"It's often said that Microsoft should have a community," Smith told InternetNews.com. "Perhaps one of the contributions I've made is to show that we have one."

He said that in 2002, 1.5 million participants generated almost 10 million messages about Redmond, Calif.-based Microsoft and its products.

"That's really great content -- our customers talking about our products, finding new uses for them, working around issues."