Saturday, 29 October 2016 01:57

We are Close To Natural-Language Voice Search That Gets Us, But We're Not There Yet


What are some of the biggest challenges in conversational search today? originally appeared on Quorathe knowledge sharing network where compelling questions are answered by people with unique insights.

Answer by Daniel Tunkelang, Data Scientist, on Quora:

What are some of the biggest challenges in conversational search today? It depends on what you mean by conversational search. Some people use the term literally to mean natural-language user interaction with the search engine, especially using voice. I prefer to use the term more broadly to describe interactive search experiences that goes beyond a single query and response. And I’m not convinced that a search experience needs to use natural language in order to be conversational.

Taking this broader view of conversational search, I see two big wins:
  • Faceted search. It’s been around for a long time, but it’s still one the most robust ways for users to interact with a search engine and iteratively refine their queries based on what they learn. We’ve also seen some evolution of faceted search interfaces: for example, the mobile-friendly integration of faceting filters into the result list on LinkedIn and Yelp.

  • Autocomplete. Autocomplete is more than a convenience to save keystrokes; it guides users to better queries. Done right, autocomplete is a character-by-character conversation, during which the search engine anticipates what you want to ask and helps you complete your thought. I particularly like autocomplete implementations that offer structured search suggestions, like those on Amazon and LinkedIn. And, despite its failure, I believe Facebook Graph Search was a worthwhile experiment.

But let’s return to the more literal definition of conversational search as being about natural language search interfaces and chatbots. There I see two main challenges:

  • Natural language understanding. Despite our advances, we’re still stuck in an uncanny valley: our natural language understanding systems work for a while and then fail unexpectedly. While we need to keep working to improve the technology, we also need to get better at managing user expectations.
  • Maintaining context across a conversation. Sometimes it just works: if you ask Google Assistant “Where is the nearest coffee shop?” and follow up with “How far away is it?”, Google Assistant figures out that the “it” refers to the answer to the first question. But when you ask “How do I block ads on an iPhone?” and follow up with “How much does it cost?”, Google Assistant thinks the “it” refers to the iPhone rather than the ad blocking app. Again, despite the impressive advances in the past few years, we’re still stuck in the uncanny valley where systems unexpectedly fail.

I’m impressed with the progress of conversational search, and I believe we’re close to having useful tools based on natural-language conversation. But I’m still typing keywords into search boxes myself, and I suspect it will be a while until we’ve climbed out of the uncanny valley.

Source : Forbes


World's leading professional association of Internet Research Specialists - We deliver Knowledge, Education, Training, and Certification in the field of Professional Online Research. The AOFIRS is considered a major contributor in improving Web Search Skills and recognizes Online Research work as a full-time occupation for those that use the Internet as their primary source of information.

Get Exclusive Research Tips in Your Inbox

Receive Great tips via email, enter your email to Subscribe.