A fair bit has been written about so-called “natural language” and “meaning-based” search engines. Many believe that search engine usability – and Internet usability in general – could be improved if search engines were “smarter” – if they were able to understand our questions or perform actions based on modes of expression that make sense to us. After all, why should we train ourselves to speak Boolean if online applications can be trained to speak human? Tea, Earl Grey, hot. Mmm, good.
Although most consumers are familiar with Ask Jeeves – a relatively weak application of the concept of natural language search – much of the research and development in this field has been done out of the spotlight, in university research labs and in companies whose goal is to quietly power the back end of heavy duty applications, not to attract the end user’s attention.
iPhrase is a leading example, building a comprehensive set of navigation tools around its original technology developed at spoken language labs at MIT. According to Andre Pino, iPhrase’s Senior VP of Marketing, the technology contains a “self-learning aspect” in its site spidering technology.
The customer list is impressive: telcos, financial juggernauts like Schwab and TD Waterhouse, technology and professional services firms, government agencies, and a large office supply company. There is also a partnership with Lexis-Nexis.
The bottom line advantage of a deep technology that better understands customers’ queries, and which helps them navigate through a web-based interface more intuitively, is easy to demonstrate in most cases. According to Pino, web-based self-service applications, once they are successfully implemented, have a far lower cost per use (ten cents) than live call centers ($30 per call) or email response ($8 per response). For Schwab, iPhrase’s mission was to create a “branch-like experience” for seven million users. A tall order, but for many of us who are hooked on web-based apps, the web experience is actually more satisfying than standing around in a branch.
Meaning-based search, in the first instance, helps users find the appropriate page by not being so picky about syntax. Some corporate search engines remain so “dumb” that the phrase “401k” brings up zero results whereas the phrase “401 (k)” returns many results. When customers can’t find what they need, they pick up the phone. Better navigation cuts down on the number of phone calls.
According to Tony Frazier, iPhrase’s VP of product management, previous generations natural language software don’t do a good job of tuning the system to adjust to the types of queries typed in by real-life users. “Out of the box” solutions have tended to be brittle and require extra consulting. iPhrase wants its solution to be closer to readiness right out of the box, although for any particular industry, there is a need to train the software at the outset with a vocabulary for common industry terms.
Frazier emphasizes that there is no single raw technology that has made iPhrase so functional – it’s oriented to the whole process of providing a sound online business application. The company touts a “one step platform” which has nine patents pending and sports impressive flow charts.
The demo I saw gave just a taste of what this kind of software is capable of, but it’s impressive. A Schwab user wanting to do sophisticated stock screening can type in “all software companies with pe higher than 30” or any number of syntactical variations on the same query, and sure enough, a table listing only companies in the software industry with price-to-earnings ratios above 30 is generated with additional relevant data.
The message from iPhrase seems to be: most natural language search products are little more than fun toys, and some more sophisticated technologies can’t be applied to real-time business applications with any certainty as yet. iPhrase, on the other hand, is a proven winner making a difference to the Fortune 500’s bottom line every day. As a result, iPhrase is looking like a long-term survivor in the enterprise search space. Its experienced management team has managed to raise solid financial backing for the company ($20 million in the bank), and the growing revenues from its high profile customers should soon make the concept of “burn rate” irrelevant.