Ask.com's newly appointed U.S president, Doug Leeds, says he intends to reinvent the "question and answer" engine, and has embarked on a project to take the company in a new direction. Part of the project, supported by technology built on search, aims to solve a completely different problem: not where to find relevant information on Web pages using link structure, but where to find relevant information based on someone's question.
The technology relies on query signals that Leeds claims have not previously been built into search. To extract and rank existing answers, rather than ranking Web pages that contain information, Ask has developed a unique set of algorithms and technologies based on relevant signals geared toward questions and answers.
Aside from traditional Web search, the engine will tap into a community of Ask loyalists -- people who know the answers but have not published the information on the Web -- to provide a real-time component. These experts can answer a question the moment it's asked.
The revamped engine required new sets of algorithms, such as question pattern-matching that can identify the type of question by its structure. The algorithm identifies a "how" question as having a different format than a "when" question. The "when" looks similar to data, while the how" question looks similar to an exclamation, Leeds explains. He says the algorithms' ability to identify these format differences can help extract answers to questions that would never be found by traditional search engines.
As an example, Leeds tells a story about his daughter looking for a personalized autograph from an astronaut for a school project. She types the keywords "astronaut autograph" into a search box on Google and gets back hundreds of pages on where to buy a Neil Armstrong autograph with a certificate of authenticity. The query results come from autograph brokers relying on SEO to monetize their Web sites.
Meanwhile, the exact answer to his daughter's question does exist on a page at NASA's Web site. That page doesn't rank in a search query on Google, for example, because there are no inbound links to the text.
But that's only one innovation, says Leeds. Finding answers to questions also means that Ask.com will rely on accessing information not previously published. This algorithm relies on signals that index people, not pages, depending on their expertise in a particular topic. The engine infers the validity of answers from the way these experts use search or other sites, how they answer questions, and their social Web and blog updates. Leeds says from this information Ask will create an index of people to route questions to that they answer in near-real time.
Ask engineers have been working on the technology for nearly a year. "Since then, we have seen the light," he says. "This is where the future is in search, so we are putting more effort into it."
People searching for answers on Ask will begin to see changes in the quality and presentation of the data in search queries by the end of November. In the beginning of 2010, the engine will tap into the community of "90 million people who use the Ask network" to create a source for answering questions, Leeds says.
Discounting rumors that Ask is up for sale, Leeds says, "I completely reject the sentiment that we don't have a niche" and won't survive without one. In fact, he rejects the word "niche," suggesting that Ask's question-and-answer format is "much larger than search, but we have to build a technology that demonstrates that."
Ask.com's core search market share remained flat at 3.9% from September to October, according to comScore. Ask's core search volume rose 4.2% year-to-year, with 4.5% growth in the total third quarter, cites J.P. Morgan analyst Imran Khan in a research report.
There are plans to launch a mobile application too. If it all sounds a bit like the mobile question and answer service ChaCha, think again. Leeds says, "This is much more sophisticated."