Talavara, a recently formed spinoff of Manning & Napier Information Services (MNIS), is bringing to market a product focusing on e-mail management and search-and-retrieval technology for e-commerce Web sites. "Talavara is focused on the business to business (B2B) market," said Michael Colson, president and CEO of Talavara. The company plans to leverage the experience and patents of parent company MNIS in the areas of data mining and linguistically based natural language search-and-retrieval technology in its initial product, which "makes e-commerce sites easy to use, and as a result, increases the look-to-buy ratio," said Colson.The Talavara product consists of two separate but integrated parts. The first is its natural language navigation and search interface. The search interface uses patented natural language processing (NLP) technology that retrieves and extracts meaning from text on multiple linguistic levels. "Talavara lets you ask questions of any length in everyday language," said Colson. "For instance, even if you don't know the name of a part, you can still find it in a massive catalog simply by describing its function."
The underlying technology of the Talavara search engine is drawn from MNIS' experience with patent searching. Patents try to obfuscate or confuse, or make up their own words to protect trade secrets, according to Colson. By matching concepts rather than keywords to a query, the system can return good results from a simple natural language description of a product without requiring the user to know specific part numbers, trade names, or other proprietary labels.
Web sites using the Talavara search interface include Edgar Online (http://www.edgar-online.com), for its IPO Profiler, and Small Business Depot's (http://www.smallbusinessdepot.com) TestStation, which lets business owners find recent government contracts that fit their business.
The second part of the Talavara product is called Talas, a "correspondence stream management technology" that automatically clusters and classifies incoming e-mail messages by analyzing their content. The system attempts to generate meaningful automatic responses to e-mail messages by comparing customer queries with answers in a Talas knowledge base. If the system has a high degree of confidence in understanding a message, it can automatically respond without any human intervention.
The system handles ambiguous queries either by requesting more information from the customer, or by routing a number of potentially relevant answers drawn from a pre-existing FAQ to a customer service agent. The agent can then respond using one of the proposed answers or modify a proposed answer with additional information before sending it to the customer. The revised response is added to the knowledge base, creating what Talavara calls "virtually a self-learning system."
Automated e-mail customer service systems are not new. Kana Communications, eGain Communications, and Brightware, Inc., all offer similar systems. Talavara's competitive advantage seems to lie in part with its patents in the area of understanding questions and extracting information from content. Two crucial differentiators of the Talas system are its ability to cluster and classify electronic messages on the fly, and its ability to measure emotion in those messages.
The classification system relies on Talas' ability to understand natural language. Talavara claims to have the only system of its type that fully integrates all levels of linguistic understanding in its technology. There are seven levels of linguistic analysis, proceeding from phonological, which is the most basic interpretation of speech sounds within and across words, to discourse, interpreting structure and meaning conveyed by texts larger than a sentence. Competing systems use only a few of these levels of linguistic analysis in their NLP processing systems, according to the company.
The Talas system also employs a real-time "Emotive Index" that allows a manager to see at a glance the relative balance of positive to negative customer feeling. This index can be segmented to show feedback about specific categories. This provides valuable information regarding how customers feel about core strategic competencies for an e-commerce site, such as product depth and breadth, ease of purchasing, on-time deliveries, and other key customer satisfaction criteria.
Factors such as language intensity, tone, and repetitiveness, as well as the length of key phrases and their position in an e-mail are analyzed to assess the emotional content of a message. Apart from contributing to the aggregate Emotive Index, this analysis is used to help route specific messages to the appropriate people within an organization for a response. A message threatening to sue the company would be routed to the legal department rather than to the returns department, for example.
Vast Market Potential
Talavara is targeting the huge B2B marketplace with its new product. B2B e-commerce transactions are projected to explode to $7.29 trillion by 2004, up from $145 billion in 1999, according to The Gartner Group. A large percentage of these transactions will include some e-mail component.
"With the volume of electronic mail expected to rise by as much as 600 percent by 2001, traditional systems are simply not equipped to provide e-commerce managers with the responsiveness and controls they need," said Colson. This need is amply demonstrated by the dismal performance record of e-commerce in general. At least four out of five online purchasers have experienced one failed purchase, and 28 percent of all online purchases fail, according to the Boston Consulting Group.
Talavara estimates the system will be capable of handling up to 80 percent of a customer service agent's workload. This lets agents focus on high revenue-generating accounts, and potentially reduces the need to hire additional service agents as business expands.
Talavara began operations on December 1, 1999, and the Talavara product will be introduced on April 3. The company is headquartered in Mountain View, California, with satellite offices in Rochester and Buffalo, New York, and it employs about 60 people.
Talavara was spun out of MNIS to allow it to focus solely on its own business model. "MNIS is looking at itself as more of a generator of new ideas—an incubator," said Colson. Another MNIS company recently spun off on its own is ipFrontline, an intellectual property news site. Colson, himself an ex-CEO of MNIS, expects more spinoffs in the future as MNIS finds ways to turn its research efforts into viable stand-alone businesses.
Talavara takes its name from a small spider that has found a niche in the intensely competitive jungle environment of Central America. To survive, the spider relies on its ability to move quickly and jump out of harm's way. Talavara's name seems like an apt choice. Companies of all stripes and sizes are stampeding to take advantage of the staggering potential size of the B2B market. Talavara will need all the agility and survival skills of its namesake to thrive in the fast-changing, chaotic world of e-commerce.
For more company information, visit the Talavara Web site at http://www.talavara.com.