Big-name companies increasingly recognize the importance of discussions about their products on social media sites such as Facebook and Twitter. But keeping track of so many conversations in real-time is a daunting challenge.

A startup called Viralheat, based in San Jose, CA, has launched an analytics package designed to allow in-depth analysis of the tenor of this back and forth. The company hopes its software will help its clients sort through the often overwhelming amount of available data from the social Web.

"The huge issue that everybody is having with social media right now is that it's just incredibly fragmented," says Viralheat's founder and CEO, Raj Kadam. There are dozens of sites where users may be discussing a brand, Kadam says. "Big brands only have so many hours in their day to understand this," he says. "They need to listen, measure what's happening with their audience, analyze it, and then figure out a way to engage their customers."

Viralheat monitors tweets, blog posts, video uploads, discussion on Facebook pages, and other social data sources. A user can be quite specific about what she wants to monitor--for example, focusing on discussions about the movie Avatar that take place outside the United States. The tool collects data that fits the bill and lets customers view them individually or as graphs that show broader trends.

The software also tries to identify the mood of a conversation, and to figure out who the main influencers are. To do this, Viralheat uses natural language processing techniques and a set of proprietary algorithms that tag posts as either "positive" or "negative" in tone. Kadam stresses that this is not the same as looking for simple phrases such as "this is cool" or "this stinks." Instead, he says, the company trains its system regularly using sets of data taken from real social media sites, to ensure it recognizes complex statements and stays accurate and up-to-date.

The data collected by Viralheat can be sliced in several ways to uncover the most influential parties in a conversation. A user can see how many people have tweeted about a particular topic, and a rough estimate of the impact of each tweet. The estimate is based on how many followers a Twitter user has, whether the message was original or simply a retweet, and whether any new information was added. Viralheat identifies two types of top influencers--those with lots of followers, and those who are most active in discussing a topic. The software offers similar tools for other social platforms as well as an application programming interface (API) so that customers can repurpose the data.