This page only works with Google Chrome browser on a desktop computer.



Speaking Phrases

If you are having trouble with the recognition, try adjusting the volume level on your computer's built-in microphone.

For this to work, you have to grant the browser permission to access your microphone.

Please note: speaking grammatically incorrect phrases may be recognized and displayed as correct. For example, saying "He like milk", is a mistake as it should be "He likes milk"; however, spoken phrases such as this are likely to be adjusted and displayed as correct. This is due to the nature of Google Chrome's speech recognition system which automatically detects and accounts for minor errors. To give another example: if you say "Snake doesn't have leg", it is possible that Chrome will adjust this to the grammatically correct "Snakes don't have legs". In doing so, 3 speaking mistakes are smoothed over and the speaker is not be made aware of them.

To be clear, sentences spoken incorrectly are liable to be adjusted and displayed in the correct form without any notification to the user. Because of this, it is important that students pay attention to the text that is actually displayed and listen to the audio.

This page will only work on Google's Chrome browser. It uses voice recognition to process and display phrases that you say. It does this by using the voice recognition service that is built into Google Chrome browser. In order for the browser to do this, you have to grant permission to access your microphone. On the first screen, click on 'okay' and then when you see a pop-up click on 'allow'. If your browser is blocking microphone access, you should see a small icon in the top right of your browser. If you click on that icon, then you can change your browser settings.

Once you allow microphone access, you should see a title screen with a text pane and an audio input monitor. You should be able to speak into your computer and see the words appear in the text pane. You may need to adjust your volume levels for better recognition - if the volume is too low or too high then the browser may have problems recognizing what you say. This screen provides a good demonstration of what the browser recognition service can do and I think it is useful to play around with and get a feel for the technology. Once the browser starts recognizing your speech, a start button should appear at the bottom. The button only appears after you have interacted with the speech recognition.

The primary aim of this page is to be a demonstration of voice recognition and how it can be used in ESL learning. It is not intended to be pronunciation practice; rather, it is intended to practice spoken production of English phrases. The other pages on this site typically practice receptive listening of vocabulary, but this page is meant to practice auditory production.

At the title screen click on 'start' to begin. The aim is to let the player learn and practice saying specific English phrases - images are shown and then sentences must be spoken. If you do not know the phrase, then you can click on the 'H' button to see it written out. Or you can click on the audio button to hear the phrase spoken.

It is hoped that by using this game the player can slowly become able to say all the phrases correctly. The learning mechanism here is quite clear - you review the phrases you already know and then gradually learn and practice the ones you do not know. You learn the ones you do not know by listening and repeating what you hear. This is quite similar to the process of using flash cards in a real classroom. Flash cards are typically used to communicate meaning and elicit the spoken forms of vocab items. Teachers will show the cards and then monitor and provide feedback. The game is pretty much trying to emulate this process in an online format.

Just to be clear, the aim of this game is not to practice pronunciation; no grading of pronunciation is measured or evaluated. Rather, it just a case of whether the user can remember the phrase and then say it well enough for the server to recognize it. The way that the speech recognition works is that it returns the words that it most closely matches what it thinks the person is saying.

Whenever the user speaks a phrase, feedback is given. If the phrase matches the target sentence, then the gold star will appear. Words that are incorrect appear in red, while words that are correct appear in black. If a word is omitted, then an ellipsis '...' will be shown. After a few seconds, the feedback will disappear and the correct phrase is displayed in black and grey.

In order to make the game more interesting, the items are divided into groups of 6 and you are also given score for each correct word spoken. Clicking on the audio or hint buttons will decrease the score.

Please leave a comment to tell me what you think and how this is working. Alternatively, you can email me using the contact form linked at the bottom of this page.