Testing software in a group of tech-crazy interpreters is so much more entertaining and insightful than just playing around on my own.
AIIC’s AI workstream (link to LinkedIn profile) recently embarked on its first team testing session, and we chose Cymo Note as our first subject of interest.
Cymo offers a whole range of support functions for interpreters, like an RSI interpreters’ back channel for colleagues working from different places, and also a fully-fledged RSI meeting platform. What we were interested in was the live transcription tool, which offers a live transcription of speech from an audio source played on your computer or fed in from an external source.
Some key facts:
Main functionalities
Complete live transcript highlighting terms plus equivalent and numbers,
Term extraction from live translation, force terms, translate selected transcript text,
consecutive mode included, can be fed with glossaries in advance, client for Windows and Mac plus Chrome-based version.
Data protection and confidentiality
Statement by Cymo: “No data is saved to the cloud and your transcript is stored locally for 24 hours only”
There are different ASR engines to select from (Microsoft Azure, Tencent Cloud STTt and iFlyTek STT)
Cost
1 credit every 10 minutes (0.14-0.18 USD/credit depending on the STT engine you chose), or monthly subscription from 58 USD
boothmates can join for free, find up to date pricing info here: https://www.cymo.io/store/index.html
Demo click for video
Our impressions
I personally find Cymo Note the most intuitive application of those I have come across, and especially appreciate the clean user interface design. Much to my surprise, others in my test team found the layout rather confusing. But in the end, we all felt the tool was very convenient to use.
Some of us tested the Chrome-based version of the tool, while others used the downloadable client. Our impression was that the client worked better than the Chrome version, where computer audio couldn’t be selected and a pen wasn’t recognised in one case. But finally, we got the system running on all our computers.
First, you chose two languages to work with from English, Arabic, Burmese, Chinese, Cantonese, Dutch, German, Spanish, French, Korean, Italian, Japanese, Polish, Portuguese, Russian, Swedish and Turkish. For those who work from more than two source languages, they will go to the settings and change languages. You also have to switch between your two pre-selected languages manually (simply clicking on the buttons) to change the source language.
We then chose our preferred STT engine and audio source.
Next we imported a bilingual glossary from a Word table, which was a simple copy and paste from a Word table. Cymo Note then showed us the terminological entries it has created and lets you edit or delete them. (Similar to other live prompting systems, the glossary of course needed to be very clean, i.e. one term and one equivalent per line, synonyms need an extra line, bearing in mind that the system will only recognise words exactly as they are written in the glossary).
Finally, we were ready to press “start captioning”. We first tested it by just speaking into our microphones and looking at the results. After that, we used a YouTube video in English on taxation and tested it with Russian, German and French as target languages. Cymo Note transcribed quite nicely, albeit not flawless, highlighting numbers in bold blue and vocab from our glossary in bold red.
Here’s what we appreciated a lot:
A detail, but very convenient: When you close your window and open it again, it will take you back to where you left the session.
Clicking on one word in the running transcript will give you the translation instantly, which you can add to your glossary by clicking on a star symbol. You can click on a word in the transcript which has been transcribed incorrectly to force replace it with the correct spelling or word.
You can save projects (with the corresponding glossaries) on your device or in the cloud. Saving it in the cloud also means that you can invite colleagues to your session by sharing a link, so that they can see the same transcript on their computer.
On the downside …
not every term from the glossary was shown,
not all the numbers were highlighted,
some numbers were spelled out as complete words in German (like siebenhundertachtundzwanzigtausend),
the quality was in general much better in English than in other languages.
Interestingly, the video transcript using computer audio from English only worked with Microsoft ASR.
And finally, we wanted to test the consec mode. We all found the idea of having a running transcript on one side of the screen while being able to take on the other very enticing. It took us a while to find our way through the settings, but we finally figured it out: Go to settings, choose flipping mode, go to the three dots, chose consec layout. The screen will then be divided in two halves, with the transcript running on the left and a notepad on the right where you can take notes or annotate the transcript, provided that you have a touchscreen and pen.
At the end we discussed a while how useful we found the tool and how likely we thought it was that we would use the tool on a real job.
The one thing we all agreed on, was that we wouldn’t use a Speech-to-Text tool that “listens in” without getting prior permission from our client.
Interestingly, opinions were split as to the usability of the tool. Some would have preferred just a list of numbers, names and terms, others liked the full transcript. Some thought that in consecutive mode they would rather read the running transcript and just annotate here and there, while others seemed to prefer to have the transcript just as a backup and rely on their handwritten notes primarily.
In a nutshell, we like the idea of having a live prompting support tools as a safety net, which is always there and easily accessible, but at the same time doesn’t distract us. But then again, it’s nothing new that interpreters are notoriously hard to please .
Further reading
Test done by Andrew Gillies EN>DE
Test done by Andrew Gillies DE>EN
————————–
About the author:
Anja Rütten has specialised in tec, information and terminology management since the mid-1990s. She holds a professorship in interpreting studies and Computer-Aided Interpreting at the Cologne University of Applied Sciences.
Contents
Leave a Reply