The next revolution will be powered by the most natural form of human interaction: voice.
Already, 32% of people are searching by voice daily. And there’s hardly an industry where voice search doesn’t add to the user experience. E-commerce alone expects $40 billion voice purchases in the U.S. and U.K., with other industries sure to follow.
Using only their voice, a shopper can be perusing your digital storefront to find exactly the pair of pants they’ve been looking for. Or just by speaking a few keywords, a reader can find and consume a piece of breaking news on their mobile device. It’s not hard to see how voice search can redefine the way a user interacts with your business.
Voice search replaces the traditional keyboard search field with an interactive relationship between the user and the business. How does it all work?
Search from directional to conversational
Voice search is often thought to need a true conversational, back-and-forth experience. Voice can be that, but it doesn’t have to be that. Voice can simply be another input mechanism.
The voice stack consists of an input, an output, and the fulfillment in-between. The input represents the spoken request (speech to text) and the output is the application’s response (text to speech). A good rule of thumb is ‘garbage in, garbage out’—that is to say, the better the input, the better the output. Presently, most voice search engines are at about 95% accuracy, which is actually better than human listening. Speech-to-text is good, and getting better.
The fulfillment in-between, however, is where complications with voice search may arise. The fulfillment is made up of the intent detection and business logic that takes what may be an incoherent input spoken in natural language and turns it into an actionable output.
Here are some of the parts that help make the fulfillment in-between possible.
- Removing stop words. Conversational language doesn’t always follow a phonetically clean, grammatically correct, or even a contextually coherent pattern. Often times, colloquial speech is messy and laden with stop words, like “find me”, “the” and “about”. For voice search, stop words get in the way of tidy queries that the voice stack can register. A good voice UX will ignore stop words and act like they’re not there.
- Reducing the haystack. For wordier formulations, the engine will need to pluck through many extra words to isolate the relevant ones. Think of the removal process as searching for a needle in a haystack. Instead of looking for the needle by examining each and every stack, it’s more efficient to make smaller stacks until you find the needle. By reducing the haystack, full-text searching becomes easier and more precise. Some functionalities that can reduce the haystack include:
- Filtering. Algolia’s built-in search engine has a feature called query rules, which takes the textual queries and looks for filterable values to reduce the stack.
- Personalization. The users’ own affinities can also be employed to filter results. Some results can be boosted to account for user behavior. The voice search experience is given a bespoke feel when personalization is woven into the fulfillment process.
- Context. The information surrounding the query makes up the context. Information such as a user’s recent queries, time or date of search, and number of user requests are all taken into account.
- Analytics. Using analytics, you can extrapolate insights from your voice engine to improve the conversational experience. For example, in the case of synonyms, users are often speaking the different ways to mean the same thing. With analytics at your disposal, you can monitor, iterate, and anticipate a user’s behavior to help them find exactly what they’re looking for the next time.
Voice search has already penetrated various industries and integrated into businesses that sit on the cutting edge. By connecting the user and business with the most natural form of interaction, successful voice search will inspire customers, drive brand engagement, and increase sales.
Learn more in our eBook: The Next Tech Revolution Will Be Spoken