[Recording Feedback (beta)] on iOS!

Simply do a voice recording while in a session!

Tess Yang avatar
Written by Tess Yang
Updated over a week ago

This article will discuss:


About Recording Feedback

Glossika is able to give you feedback on your pronunciation. Simply record yourself speaking the sentence you’re drilling while in a session, and then we'll quickly do an analysis and let you know how you did.

As of date, this is an early-stage beta release of the featre. For now, this means:

  • This is not the final version of the feature.

  • This feature is currently available for six target languages:

    • Italian

    • German

    • Russian

    • English (US) and English (UK)

    • Spanish (Spain) and Spanish (Mexico)

    • Chinese (Taiwan) and Chinese (Beijing)

  • The feedback provided is quite simple. We’ll let you know which parts of your speech we were able to successfully detect and map to sounds in your target language.

Some FAQs are available at the bottom of the document.


How to use the Recording Feedback feature

First, check that the feature has been enabled. While in a new items session, click the gear icon in the top-right corner of your screen, scroll down to recording, and then confirm that Recording Feedback has been enabled.

From there, it's easy!

Upon finishing a sentence, simply click the "Tap to Speak" button. This button will appear after the "Base Language" and "Target Language" audio has finished playing.

Upon recording your voice, you'll get visual feedback, as shown below:

  • Green text means that your pronunciation was accurate enough to be mapped to the displayed characters.

  • Grey text means that we were unable to map your pronunciation to the correct characters, either because (a) you pronounced an incorrect sound or (b) there was audio interference with your recording / you spoke too quietly for Glossika to recognize what you said.

Some commonly asked questions:

Q1. What do the colors green and grey represent?

Green indicates that we were able to recognize a sound you made and map it to a sound in your target language (TL). If we could not detect your voice, or if we detected an incorrect sound, the text remains the original grey color.

Q2. Why are punctuation marks or special symbols all grey?

Pronunciation marks are not audible, so they are not something that our AI can detect in your speech. As such, while pronunciation marks always appear as gray/incorrect, they do not cause you to lose points.

Q3. Why are some words/sounds not recognized, no matter how I say them?

Pronunciation is difficult. Just as you do not start out with perfect grammar or knowing 30,000 vocabulary words, it takes time and effort to improve your accent. To give a few examples:

  • Learn — Your TL likely contains some new sounds that don't exist in your NL (native language), and you need to learn to make these sounds

  • Avoid — Your NL likely contains some sounds that don't exist in your TL, and you need to learn to avoid these sounds

  • Adjust — Your TL and NL likely contain many sounds that are similar, but not quite the same, and you'll need to experiment with and tweak these sounds

  • Explore — Pitch, intonation, rhythm, and how one word connects to another often differ from language to language — what feels natural to you as an NL speaker is likely not natural in your TL

On the other hand — if you are very certain that your pronunciation is correct, there is a chance it may be (a) due to an issue with our AI or (b) because you're speaking too softly. This is a beta feature, and it's still being improved!

If in doubt, we encourage you to explore the phonology of your target language. It's interesting, and there's a lot to be learned!

Q4. Why can't some names be recognized?

We have observed issues with the recognition of some names in Italian and Chinese. We're investigating the source of these problems, and in the mean time have made our scoring criteria more lenient. Incorrect names will not significantly affect your final score.

This optimization/debugging requires engineering assistance, so it's something that will be released with a future update.

Q5. Why can't Chinese numerals, repeated Chinese characters, and gender-specific words (she/it/he) be recognized?

We're investigating the source of these problems, and in the mean time have made our scoring criteria more lenient. These issues will not significantly affect your final score. As an additional note, AI voice recognition is not 100% accurate, and we are working on improving our accent detection capabilities.

This optimization/debugging requires engineering assistance, so it's something that will be released with a future update.

Did this answer your question?