Can you use voice assistants to practise speaking English?

Many of us have a voice assistant in our homes or in our pockets and it is predicted that this technology will become even more commonplace over the next few years.

As a language teacher and designer of innovative tools for language learning, I started wondering — could a language learner use a voice assistant like Alexa, Siri or Google Assistant to practise the language they’re learning?

One of the biggest challenges for many language learners is finding opportunities to speak the language. Could Alexa become your new language partner? Could you use a voice assistant to train your pronunciation?

This article will investigate these questions, with a focus on learning English as a foreign language.

First let’s think about pronunciation

When it comes to learning a foreign language, one of your goals is to make sure your pronunciation is clearly understood by native speakers. This is especially true for English, which has more complex pronunciation rules (or lack of rules!) than many languages.

Could you use a voice assistant to get feedback on how clear your pronunciation is? If you say something to a voice assistant and it understands you, does that mean your pronunciation was ‘correct’ and clear? (I put ‘correct’ in quotes because there are so many varieties of English that it can be difficult to classify what is ‘correct’.)

The short answer is — no. Generally not.

Voice assistants are very forgiving in what they understand and they have been designed like this on purpose. So many different accents exist among English speakers, both native and non-native accents. The goal of speech recognition systems is to understand as many people as possible, therefore it’s not in the interests of a standard general voice assistant to start being picky about how close someone’s pronunciation is to the ‘correct’ pronunciation that an English learner is probably aspiring to.

Voice assistants were designed to be useful tools that are as helpful as possible for carrying out tasks inputted by voice commands. They were not designed to teach you pronunciation.

However, if the voice assistant’s speech recognition system can be designed to be forgiving of non-standard (some might say ‘incorrect’) pronunciation, presumably they can also be designed to be stricter. Anything that is in the system to broaden the likelihood of it understanding, can presumably be removed. I think there is great potential for using voice recognition to train pronunciation.

An app called ELSA Speak has attempted it, but with disappointing results, as seen in the video below. In this video the app is giving practice of ‘s’ and ‘sh’ sounds. I gave some typical non-native pronunciations and 3 out of 4 times ELSA marked them as ‘Excellent’, despite them definitely not being excellent.

I would of course love to develop a totally awesome pronunciation training app but I need to stay focused on other priorities for now.

Can a voice assistant help you train your grammar?

In much the same way that voice assistants are forgiving about unclear or incorrect pronunciation, they are also forgiving of incorrect grammar.

I found many examples where I asked the voice assistants questions containing horrible grammar mistakes but they still understood. A human would do this too in a conversation, so in that sense, voice assistants are managing to effectively mimic human behaviour, which is great in a way, but not great for an English learner who’s trying to improve.

What’s worse, Siri and the Google Assistant app actually repeated the grammar mistakes in the text they displayed on the screen. I feel that’s detrimental for language learning because it reinforces the mistakes.

The most fascinating discovery of all

We made one really fascinating discovery. If your Google Assistant is set to Spanish and you speak to it in English, it understands. But, it understands you much better if you speak English with a Spanish accent. If you speak English to it with a ‘perfect’, clear, neutral British accent, it doesn’t understand very well. I later discovered the reasons for this, which I’ll explain below, but first let’s think about how this impacts language learning.

What it means is, if you’re a Spanish person wanting to use Google Assistant to practise speaking English, it’s really important that you set it to English mode. Don’t just speak to it in English while it’s set in Spanish because that will encourage you to use Spanish pronunciation of English. And of course ‘Spanish’ pronunciation of English is not ‘good’ pronunciation. If you do pronounce something in a more ‘English’ way it won’t understand you and this will discourage you from doing that correct pronunciation in future.

You can see this happening in the video below.

Why does Google Assistant set to Spanish understand English but only when it’s spoken with a Spanish accent?

I’m no expert on how speech recognition systems work but, using my limited knowledge, I’m going to try and explain.

According to this blog post written by an expert on the topic, speech recognition uses “statistical representations of the most likely and meaningful combinations of sounds in the specific human language”

In other words, it makes assumptions based on which sounds commonly occur together in a particular language. If Google Assistant Spanish is listening for the sounds of Spanish, used in combinations that are common in Spanish, it’s not going to understand when you make sounds that don’t exist in Spanish or combinations of sounds that would never be found together in Spanish. A Spanish person speaking English with a Spanish accent is often substituting English sounds with the closest matching Spanish sounds, meaning Google Assistant Spanish understands that person much better.

What’s the best way to use a voice assistant to practise English or another foreign language?

As we’ve just learnt, first set the voice assistant to English mode (or whatever language you’re learning). There are then a couple of different activities you could do.

1. Pronunciation of individual words

To practise the pronunciation of individual words ask this question ‘What is a ____?, filling the gap with the word you want to practise.

In this exercise, the assistant is hearing the word completely out of context, so it has to rely only on your pronunciation of the word to understand it. There are no clues about the word from the meaning of the rest of the sentence (and systems like Google Assistant do use those clues to help understand what you’re saying).

If the assistant replies with the right answer, you know your pronunciation was correct.

Here are some examples to try with commonly mispronounced words in English.

What is a vegetable?

What is a receipt?

What is a lettuce?

What are clothes?

What is prejudice?

2. Pronunciation of longer sentences

You can also test your pronunciation in some longer sentences. The more obscure or unusual the sentence, the better. If you ask a question that is commonly asked (eg. What is the weather today? Where is my nearest doctor?) the voice assistant will use its AI to make the assumption about what you’re trying to ask and it will understand you better.

If you ask a more complicated question and one that very few are likely to have asked before, the assistant will be less forgiving and you will be required to use better pronunciation. For example, ask ‘Which flowers are good to use in cooking?’ or ‘How many ships cross the Atlantic ocean every year?’. Be creative!

3. Play a game

Perhaps the easiest and most fun way to practise is to use one of the pre-programmed word games that are available on Alexa or Google Assistant. These are not really a replacement for having a conversation but at least it’s a way to just speak some words. Here are some examples of games you might like:

Mind Your Words on Alexa

Would You Rather? on Alexa

Mystery Animal on Google Home

Speech recognition in language learning apps

If your ‘every day’ voice assistant doesn’t meet your language learning needs, try using the speech recognition feature within a language learning app.

Some language learning apps incorporate voice recognition into their learning materials. One example is Lingvist, where you are given sentences to read out as part of a pre-written conversation. This is an excellent way for beginners to practise saying words and phrases in the language they’re learning. Lingvist has many other great features but I was most interested by its speech recognition activities.

The video below shows me testing three English conversation activities in Lingvist. I tested with a ‘perfect’ English accent and a ‘non-native’ English accent (apologies for my slightly comic range of European accents). I wanted to see if it lets you know if your pronunciation is not ‘correct’. Similarly to other voice assistants, it was very forgiving of poor pronunciation. This isn’t necessarily a bad thing because a beginner in a language should focus on communicating and building confidence, rather than having ‘perfect’ pronunciation.

Other apps that use speech recognition are Busuu, Babbel, Mondly, Rosetta Stone and more. Those apps don’t have a free demo but I’d love to try out their English speech recognition if anyone is able to provide me with access for a day.

Lover of languages and language learning. Strangely fascinated with the Middle East. Owner of the Arabic language learning website

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store