TD Waterhouse Group Inc. says that customers using its Web site to check an account balance or review their investment options will find accomplishing such tasks easier thanks to a new natural-language search engine.
The brokerage firm rolled out iPhrase Technologies Inc.’s One Step software in just six weeks and made the new functions available in March. One Step attempts to live up to its name by eliminating some of the steps associated with traditional Web searches. Typically, if a visitor were to type “What is your fee schedule?” into a search box, the site would retrieve a list of documents containing the phrase “fee schedule,” including a link to the brokerage’s fee schedule somewhere among several other linked documents.
“Now it skips the middle step and takes you right to the fee schedule page,” says Joe Kubikowski, a TD Waterhouse vice-president. For queries that are not so clear-cut, the software delivers a list of related products and services. “How do I plan for my child’s education?” for example, yields a range of TD Waterhouse education planning resources.
Natural-language search technology sets out to improve navigation by accommodating a range of user queries phrased in everyday language, and automatically adjusting for misspellings and abbreviations, for example. This advanced technology also can be used to customize search results based on a visitor’s access privileges, such as by responding to the question “What’s my account balance?” by bringing the end user directly to his account specifics.
Such search features are part of a wider umbrella of customer self-service products that aim to ease the burden on call centres and divert users to less-expensive online channels, where they can answer their own questions.
It’s a crowded market – vendors competing for self-service dollars include Ask Jeeves, Banter, iPhrase, Kana Communications, Kanisa, noHold, RightNow Technologies and dozens more.
It’s also a volatile market. Of the roughly 30 vendors that offer some sort of self-service tools, only 20 per cent will survive, predicts Esteban Kolsky, senior analyst at Gartner. Most will be acquired or merge with other companies, he says. But the payoff could be big for those that stick around, as Gartner estimates 70 per cent of customer service interactions for information and remote transactions will be automated by 2005, whereas 25 per cent to 30 per cent are today.
Among natural-language search technology vendors, the companies that stand out combine natural language with self-learning, Kolsky says. Natural-language technology requires an underlying taxonomy, or classification system, that establishes links between available content and questions asked. Taxonomies consist of three layers: general vocabulary and grammar; terminology specific to a vertical industry; and company-specific content, such as proper product names.
Maintaining that taxonomy is critical as Web content changes. If it has to be done manually, that’s a huge problem, Kolsky says. Self-learning automates the process of discovering new content and establishing contextual links to questions asked.
Meanwhile, development of vertical-specific taxonomies is driving financial services firms, in particular, to become early adopters of natural language technologies. Banter, for example, counts ABN AMRO, Royal Bank of Canada and Wells Fargo among its customers, and Ask Jeeves has attracted Datek Online, Wachovia (First Union) and PNC Financial Service Group.
