Imagine a scene where the crew of the Starship Enterprise is trying to get information from the computer and you will have an idea of what Google is working on when they talk about “conversational search.” Conversational search is an effort to construct a search machine capable of responding in a flexible and intuitive manner like a person would in response to queries or commands. In a recent interview, one of Google’s Research Fellows, Jeff Dean, talked about his team’s work on machine learning and neural nets as part of an effort to make conversational search a reality.
Dean gives an example for those who might not be familiar with how conversational search works. In response to a query like, “book me a flight to DC” a human response might be along the lines of:
you’d ask me a bunch of follow-up questions, “What hotel do you want to stay at?” “Do you mind a layover?” – that sort of thing. I don’t think we have a good idea of how to break it down into a set of follow-up questions to make a manageable process for a computer to solve that problem.
At least, not yet if Dean and his team are successful. Dean is hoping to construct the code necessary to understand a requirement set forth in a query, ask intelligent questions to clarify, and then complete the query using whatever appropriate sources are available.
The dreaming does not end with building the ability to conduct a conversational search using Google’s search engine, something that already exists in a rudimentary fashion in the Chrome browser. Dean sees the capability for a computer to conduct a conversational search being coupled with Google Now and our devices to make things seamless no matter where we are. Google Now can bring situational awareness to the equation since it can assess a person’s location and things like time of day. The conversational search engine can take those additional inputs to build an even better response.
Going a bit further, Dean thinks the day may come when a device like Google Glass can be used to constantly survey your surroundings and will stand ready with augmented reality information without input from the user.
Besides identifying the multitude of tasks that humans engage in when responding to queries and breaking those down into tasks and instructions that a computer could complete, Dean’s team is also challenged by the nature of data as either labeled or unlabeled. Our reality is that the majority of data in this world is unlabeled, making it a challenge for a computer to figure out what it is and then how that can be used. Dean thinks the ability to successfully tap into unlabeled data would yield improved results by several orders of magnitude.
What are you ready to ask your computer to do?
source: Search Engine Land
via: 9to5Google
Come comment on this article: Google working on “conversational search” for devices
No comments:
Post a Comment