June 23, 2010 –
Company founder Li is flanked by interns Vijay Rajakumar (left) and Inci Cetindil on move-in day.
Last week, UCI Associate Professor Chen Li and his summer interns made it official: they carried boxes and computers into a pristine lab on the second floor of the CALIT2 Building, becoming the inaugural tenants of TechPortal, UCI’s newly minted technology business incubator.
It was a move that Li, founder of BiMaple Technology, Inc., had been anticipating since first hearing of an incubator on the horizon. “I jumped at the opportunity,” he said. “It’s on campus, it’s next to my ICS building, and CALIT2 provides excellent services. The rate is very reasonable and they provide all the support that startups need.”
BiMaple, based on Li’s academic efforts to improve online search, is developing powerful interactive-search software to make Web sites faster and more responsive.
In contrast to large search engines that analyze the history of previous search queries over the entire Web before making recommendations, BiMaple’s software indexes a specific site’s database, and uses an innovative (patent pending) search algorithm that promotes interactive search and provides instant “search-as-you-type,” error-tolerant results.
The software focuses on “vertical” search, meaning deeper search on a particular domain, such as books, people and electronics. All searches share the same core technology; domain-specific adaptations can be integrated.
The effort started when Li developed a research site called PSearch, a directory that locates UCI faculty, staff and students. The software’s algorithms enable instant search of the site’s entire database as users begin the process of typing in keywords.
Rajakumar and Li set up their equipment.
Li hosts a pizza party to thank CALIT2 staffers for their help.
The software also recognizes certain synonymous words and “reasons through” possible errors. Type Bill Cohen into the search box and you will see results that include William Cohen and William Chen.
Matches can be made with bare-bones information. Someone searching for a professor or student might have only a first name or a nickname and department. The directory can usually return correct results since it searches the entire database for matches that are based on more than merely the literal connection.
Li was stunned at the response to the site. “We were getting overwhelmingly positive feedback from users,” he said. “So I thought, ‘why don’t we try to commercialize this?’”
The company’s name derived from two imposing trees outside its former headquarters.
BiMaple was incorporated in late 2008.
The overarching goal of the company’s search technology is real interactive search that is progressively adjustable in real time. “Suddenly, search becomes a friendly, meaningful and convergent conversation between the user and the machine,” Li explained.
To display the power of its new technology, BiMaple built an online demo designed around a book database. Suppose you would like to purchase a book but can remember only one word of its title. As you type each letter of that word, the search engine returns selections of books whose titles (or authors) contain those letters. The process is instantaneous; results are returned on-the-fly, before the keyword is completely typed.
For instance, a user is looking for a title that contains the word breakfast. As he types, the search engine keeps pace, first returning results containing br: selections include bread, brain, and brand. By the time he has typed break, the results have changed, and include titles with breakup, outbreak, daybreak, etc. Before the word ‘breakfast’ is completed, the search engine has returned a list of books whose titles contain that word.
Even if the searcher misspells the keyword – berkfast for instance – the site instantly provides the correct list. It displays the books’ cover images as well, increasing the odds that the user will find what he is looking for.
For searches containing multiple keywords, users can better define the search by refining one or more keywords based on the visible search results.
To create the demo, Li’s team used software that crawled a popular Web site to collect book information. The finished product will be marketed to general Web sites, where it will pull information from their existing databases.
Many search engines utilize what Li refers to as “auto completion,” making suggestions as users type. That approach, however, is prefix-based and treats user-entered keywords as a part of a precompiled sample search phrase.
“We do a full-text search, not a simple prefix search,” he explained. “Existing techniques cannot interactively find books with a keyword in the title and a keyword in the author’s name because the two are unlikely to be in close proximity. Our software can overcome this limitation.”
Li’s team used the techniques to build another prototype called iPubMed to support interactive search on the fast-growing MEDLINE database containing more than 19 million medical publications.
The technology, with its high rate of interactive speed is “a big jump from what was available before,” Li said.
“It is especially for Web sites to make sure their end-users have a pleasant experience; customers need to be able to find products or information hidden in the site.”
Noting that most website search engines are still “pretty old-style,” Li believes there is a burgeoning market for BiMaple’s product. “My goal is to make their search interfaces much more powerful.”
–Anna Lynn Spitzer
