Google: Speech, natural language recognition are major challenges

As Google aims to expand search's ability to answer increasingly difficult questions based on contextual information, several key obstacles still stand in its way.

Knowledge Graph, a feature designed to put users' searches in context and deliver tailored results, as well as understanding speech, natural language and conversation are the four biggest technological challenges the company faces today, said Amit Singhal, Google's senior vice president of search.

"None of these are solved problems," he said Sunday during a Q&A session at South by Southwest Interactive in Austin, Texas last week.

It's not surprising that Singhal would identify those four areas of search as the most in need of improvement -- they comprise the backbone of the company's ambitious efforts to turn Google search into a "Star Trek computer."

That goal, Singhal explained, is to make Google's search a multi-modality platform capable of delivering search results regardless of how people query it, or even when they don't query it at all.

For example, while people may be growing accustomed to searching in different ways -- via voice, touch or text -- often times the situation the person is in determines how they are going to search, Singhal said, and Google needs to be able to adapt accordingly.

Mobile phones, for instance, provide a more natural interface for voice- as opposed to text-based searches, but if a person is in, say, a lecture hall, searching via voice in that context is not the way to go, he said.

It's a concept that is already playing out with Google Now, a mobile Android app designed to give people personalised information on the fly without their asking for it. The company is also developing Google Glass, a head-mounted unit to give people real-time information throughout the day.

"We're designing search for everywhere," Singhal said. But in the end, "the technology should fade into the background so it gives you what you need as you're doing what you're already doing," he added.

Meanwhile, Singhal kept mostly mum in response to questions over whether Facebook's early-stage Graph Search tool, which is designed to let users ask more questions on the site tied to their network of friends, had implications for any of Google's search goals.

"Time will tell if people need that kind of search," he said.