Voice-to-text: The future of search?

You’re driving, so you can’t use your phone. But you really need to know when the next Marvel movie is coming out. What do you (and 52% of people) do? You use voice search! Voice search, or voice-to-text search is supposedly the next big thing. Let’s find out if that’s true, and if it’s good or bad news.

What is voice search, or voice-to-text search?

It’s in the name: instead of using a keyboard to search online, you use your voice. But this isn’t possible without a program or handy search assistant like Google or Alexa. They convert your voice into text, then use that prompt to fulfill your request. A voice search could sound like this: “Hey Google, how many people live in Wijchen?” 

(The answer is 40,951 as of 2019).

Are people using more voice-to-text search?

Let’s look at some numbers. A study by Uberall found that 21% of people were using voice search on a weekly basis. A HubSpot survey found that 74% of people had used voice search within the last month. And last but not least: almost 50% of people are now researching products before they buy using voice search. 

Looking at these numbers, you can’t deny that voice search is becoming increasingly popular. But why?

Why are people using more voice search?

Obviously, a lot of this growth is due to smart devices like Apple’s Siri, Amazon’s Alexa, and Google. Because they’re everywhere. You can talk to your phone, laptop, tablet, car, speaker, thermostat, and probably many more. Sure, you could do that before, but now these devices actually understand what you’re saying. With voice search being all around us, it isn’t hard to grasp why people start to use it. 

Another reason is that it’s easy. As we’ve just established, voice search is almost always within reach. And it’s much faster to ask a question, than to open a web browser or app, type out your query, scroll through the results, then pick one that sounds good. 

Better and smarter

Yep, we’re looking at AI again. Now that machine learning capabilities are better, our voice assistants will get smarter over time. Which means that the more you use voice search, the better it will understand your preferences and provide you with the answer you’re looking for. 

An accessible option

Better voice-to-text search devices also means improved independence for people with disabilities who can’t use a keyboard or touch screen. Tasks that might’ve been difficult or impossible are now suddenly possible with the help of Google, Siri, or Alexa. Which is awesome news. 

And it doesn’t stop there. Voice-to-text devices can also help people who have a visual disability, since a voice assistant can read a web page, blog post, or email aloud. And it’s useful for people with hearing impairments too. When they’re attending a meeting, a voice assistant can transcribe what’s being said into text.

The problems with voice search

All good things, right? Well, kind of. The problems arise when we start looking at language. Most voice assistants are trained exceptionally well in English, which shouldn’t be a problem because everyone speaks English, right? Wrong. 75% of people in the world don’t speak English at all. So how useful are these voice assistants anyway?* 

Still, of those 25% that do speak English, only 6% are native speakers. Which means 19% of people will probably have an English accent, or might not be familiar with certain English phrases, sayings or metaphors. And since voice assistants rely heavily on natural language processing and machine learning, this means those 19% of people will have a harder time using voice-to-text devices. 

*While Google is stepping up their game with more than 40 languages available for their voice search, Siri only speaks 21 languages, and Alexa only eight. 

Independent or dependent?

While voice assistants allow people with disabilities to become more independent, it might have the opposite effect on other people. Because only using voice search has its limitations. Voice search gives you one answer. Usually the one at the top of the search results. Now, we know that Google’s Helpful Content update makes sure that the top results are helpful, but it isn’t 100% foolproof. And looking at the top results nowadays, we see… ads.

If people only and unquestioningly use voice search (because it’s so easy), they’ll only learn about one answer and one narrative. And this scenario isn’t that far-fetched. Already, people are using ChatGPT to answer their questions, without checking if the answers are actually correct. 

Hey Google, are you biased?

Remember those 6% of people who are native English speakers? Turns out, you have to be the right English speaker in order for certain smart devices to understand you. If you’re from the UK or Ireland, chances are your device won’t understand you.  

In addition, smart devices have trouble understanding input from people who are not white men. As is the case with facial recognition, speech recognition also performs worse for women and BIPOC people. And while this might not be intentional, it’s still problematic.

What do the voice assistants have in common?

To look further into the biases of voice assistants, we can’t ignore that the majority of them are still, by default, women. And this, as UNESCO so aptly says in their report: “sends the signal that women are obliging, docile, and eager-to-please helpers, available at the touch of a button or with a blunt voice command like ‘hey’ or ‘OK’. The assistant holds no power of agency beyond what the commander asks of it… In many communities, this reinforces commonly held gender biases that women are subservient and tolerant of poor treatment”.

Lots of potential, and lots of work

In conclusion: voice-to-text search sounds awesome in theory, but isn’t quite there yet in practice. Devices that support voice search need to add and improve other languages, and let go of their biases. Because when I search online (if it works for my voice at all), I want to know the truth; not just what is true for one group of people. 

Read more: Why accessibility is important (in the time of AI) »