In the final part of our series, SemanticWeb.com looks at another semantic web-based offering for news aggregation that delivers business value. Today, we look at TeamStream.com. (The series begins here and continues here.)

TeamStream

TeamStream is the most recent service to emerge from a semantically-minded startup from down under. Launched in beta last month, it aims to streamline the production of marketing and research reports, letting news experts within a business work together to curate the news and create analytics reports for broadcast throughout the organization. It came about because its founders -- Stephen Phillips and Graeme Wood -- saw potential within their other properties, which include serious business news sites wotnews.com, wotnews.com.au and wotnews.com.UK, to serve as a direct sales channel to TeamStream.

Phillips is the mind behind plugger.com.au, an Australian business news search engine that collected RSS feeds from all the major Australian news networks, blogs and companies, and allowed people to search for the news they were interested in. The site, which morphed into wotnews.com.au, used basic semantic analysis to identify the most prominent people, companies, locations and topics mentioned in the news, which enabled it to surface popularity charts for these entities, as well as allowed people to monitor the specific entities they were interested in. Phillips’ partner Wood is the founder of Australia’s successful hotel booking website, wotif.com. Together the two also have launched wearehunted.com, a site that uses semantic techniques for music discovery.

Its free web sites, Phillips says, gave his and Woods’ venture “a large captive business audience, and the challenge to develop a product that we can sell to these prospects.” TeamStream.com became that product for monetizing the million visitors they get across their sites each month, and it incorporates features for accessing live news streams from Wotnews Australia, USA, and the UK; the ability to annotate the news with individual insights and share that knowledge across the organization; the ability to create custom news channels to track specific people, companies, topics and events; and metrics on marketing effectiveness. Pre-TeamStream, “what we noticed looking at the member activity across our sites is how many users from the same organization were using our sites to track the same topics in the news. We could see a lot of duplicate activity and sensed that a lot of collaboration was happening, quite ineffectively via email, outside our system,” Phillips says.

With TeamStream, users can follow their business colleagues’ reads and recommendations and comment on them. They and their colleagues can follow thousands of publishers – the service pre-loads quality-controlled news feeds to the major news publishers here, in Australia and the UK, as well as social network and government source feeds – and work together to manage fully customized news channels to track specific topics. Customers can also add any feed including private, internal feeds to blend their own content with external news. “We could see that there was an opportunity to provide tools to allow people from the same organization to read and share news much more effectively.” Phillips says. “This would both save time and better capture the knowledge being expressed in a form that could be saved, searched and shared with others across the organization.”

News collaboration with colleagues is enabled with TeamStream.

The two primary functions of TeamStream are market research and media and brand marketing, which most businesses today currently get through traditional media monitoring firms and market research methods such as focus groups and surveys. “Both methods are slow and expensive. Both will be dead within a decade,” Phillips predicts.

Long living in their place will be the monitoring of online news, blogs and social sites like Twitter and Facebook, which he says present a whole new world of real-time market research. “If you can aggregate and measure this data, it can provide tremendous insight into what is happening in the market. TeamStream streamlines this process for marketing and research teams,” he says. For example, the company is working with recruitment companies who use the service to identify what they consider to be sales triggers in the news so that they can discover new sales leads. With the web providing access to real customers giving honest feedback about products and services, TeamStream can deliver media monitoring value by streamlining the process of producing these reports. It’s not as comprehensive as traditional media monitoring sources, Phillips notes, as it doesn’t index TV or print, but “it is faster and cheaper and delivers a broader range of sources. PR and marketing teams can work together to track their brands as they are discussed across the web.”

The semantic technologies are key to making this happen, he says. “Understanding what the data you aggregate actually means is very important,” Phillips says. “As semantic techniques improve, so does the quality of the insights captured for marketers and researchers. I see the key ideas behind semantic tools as enablers rather than platforms themselves, so I would see them being really useful in a very broad range of apps such as CRM and ERP.”

He expects more players to realize the potential semantic technologies can bring to the enterprise for these important jobs. “It is the new frontier of marketing and research and there is lots of room for new ideas,” he says. That said, he considers TeamStream’s competitive advantages to include the thousands of human-reviewed news—and thus quality-controlled—feeds it has collected, and with them, an historical archive of over 10 million articles from around the world. “The second strength is the dataset of entities (people, companies, locations, topics ) we have developed,” he says. “These are human-reviewed with tools for disambiguation/synonyms etc. While we continue to improve our named entity recognition, from our experience, nothing beats a semi-supervised approach. We use machine learning techniques to identify potential new entities, but we always have humans review the final result. Then it is a matter of high speed brute force matching to tag text for entities effectively. Our entity dataset has not only been collated by our editors, but also farmed direct from the thousands of user profiles of our members who tell us the entities they would like to track.”

TeamStream continues to invest in machine learning techniques to learn what it can from the data it sees. Its analysis includes sentiment detection to work out whether news stories are generally positive or negative about a topic; clustering and classification to group related entities together; and network analysis to discover the connections between entities and the publishers who write about them. “Sentiment analysis is tough,” he acknowledges. “What is not in doubt is the benefits of sentiment analysis and the desire for companies to pay good money for products that deliver results. We believe the nuance of language and difficulty assessing intent makes sentiment analysis at the atomic level (an individual article) fairly meaningless. But what is possible is to assess sentiment across a large set of news articles on specific topics. As long as the customer understands what they are looking at, and can make judgment decisions based on the reality of what is possible, sentiment analysis is valuable.”

TeamStream is available in a software-as-a-service model, with pricing dependent on the number of team members involved. “There seems to be a real energy inside big companies with the understanding that trying new things online will be a big source of future value,” Phillips says. “No matter what business you are in, the web delivers opportunity.” So far, some of the types of customers hoping TeamStream will help them deliver on those opportunities are banks and consultants who can profit most from the real time intelligence that can be gleaned from the web. At least one bank is using the service to detect and discover potential cases of share trader fraud “For traders and sales guys time is critical to success,” Phillips says of other interested parties. “They can see the obvious benefits of listening closely to what the web saying.
We have also had a lot of interest from telcos, ad agencies/planners, PR firms, government agencies, and more.”

Phillips expects the “TeamStream experience” to evolve aggressively over the next few months. “We plan to stay in beta until we get to where we are in solid production use with several of clients,” he says.