Does webkitSpeechRecognition send recorded audio to a remote web service by default?
Reported by
guest271...@gmail.com,
Feb 24 2018
|
|||||||
Issue descriptionUserAgent: Mozilla/5.0 (X11; Linux i686) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/64.0.3282.167 Chrome/64.0.3282.167 Safari/537.36 Steps to reproduce the problem: 1. Try to use webkitSpeechRecognition() when offline 2. Try to locate documentation of why webkitSpeechRecognition() does not provide the same results offline 3. What is the expected behavior? webkitSpeechRecognition() provides the same results when offline as are provided when when online. Locate documentation which clearly provides the processing model of webkitSpeechRecognition. Specifically, if using webkitSpeechRecognition What went wrong? webkitSpeechRecognition() does not provides the same results when the user is offline as when the user is online Have not been able to locate clear documentation clearly stating that webkitSpeechRecognition is either a) processes the recorded user audio locally without making a request to a remote server to provide results of the speech to text processing b) sends recorded user audio to a remote web service which provides the results of the speech to text processing Did this work before? N/A Does this work in other browsers? Yes Chrome version: 64.0.3282.167 Channel: n/a OS Version: Flash Version: See https://stackoverflow.com/questions/39850585/webkitspeechrecognition-architecture https://cs.chromium.org/chromium/src/content/browser/speech/speech_recognition_engine.cc
,
Feb 25 2018
,
Mar 5 2018
guest271314@ Thanks for the issue. Request you to provide a Extension or URL where this issue can be reproduced, which will help in further triaging. Thanks..
,
Mar 7 2018
#3 Not sure what you mean by "Extension or URL"? The OP includes an HTML file https://bugs.chromium.org/p/chromium/issues/attachmentText?aid=326366 which reproduces the issue.
,
Mar 7 2018
Thank you for providing more feedback. Adding the requester to the cc list. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Mar 7 2018
This is probably a WontFix, because this is simply how it is (on non-Android). And if there is no documentation, it is because this can change at any time.
,
Mar 7 2018
#7 > This is probably a WontFix, because this is simply how it is (on non-Android). Why? That is, why cannot the speech recognition code be included in Chromium build? > And if there is no documentation, it is because this can change at any time. Documentation for any component can change at any time, correct?
,
Mar 7 2018
#8 - it is also not included on Chrome for Android (it is provided by the operating system/externally installed engine). The speech engine is proprietary, not open source, so Chromium cannot include it. Yes, but implementation details are generally not documented, as a common practice.
,
Mar 7 2018
By the way - > Does this work in other browsers? Yes Which non-Android browser supports offline speech recognition? (Which browser supports speech recognition, other than Chrome?) Also, if you have a speech engine, you can use it via chrome.tts. Oh, wait, tts is text-to-speech, sorry, so I guess you cannot use it for this. But you can implement it by yourself using getUserMedia, Web Audio, Web Assembly and so on. Not an easy job, but that is all I can think of right now.
,
Mar 7 2018
Are implementation details concerning users audio (potentially their distinct voice) being sent to an external web service, and possibly being stored (forever) by said proprietary service considered general or specific to that user? From perspective here, such a detail should be clearly stated to avoid confusion https://bugs.chromium.org/p/chromium/issues/detail?id=815388. > not documented could also imply that the retention of user audio could in fact be forever; at least it is not clear if the audio is retained at all - for further proprietary purposes. See https://bugzilla.mozilla.org/show_bug.cgi?id=1248897. There was interest expressed by users in using pocketsphinx at a Firefox bug report, will try to find the link. Yes, the requirement should be able to be implemented with existing components without using an external service.
,
Mar 7 2018
,
Mar 7 2018
> the requirement should be able to be implemented with existing components without using an external service. I have not heard of any requirement of such. It is your need, but not a requirement. I guess the terms of use (about:terms of Chrome) should mention this somehow, if it does not already. Non-Android Chromium cannot access the web service without an API key (and even with a user generated one, its quota is severely limited, because it is meant for development). So the terms of use of Chromium (if here are any, since it is not supposed to be used - it is not a product) would probably not mention that. But the terms of Chrome should, I guess (maybe not explicitly, since other browser features use web services, like Chrome Sync, translate, push messaging... So it is probably a general chrome-may-use-web-services clause).
,
Mar 8 2018
> I have not heard of any requirement of such. It is your need, but not a requirement. The requirement is speech recognition. If the browser uses an external web service to achieve implementation of the Web Speech API, from a Chromium user persepctive that should be explicitly documented. > Non-Android Chromium cannot access the web service without an API key (and even with a user generated one, its quota is severely limited, because it is meant for development). That is not accurate. It is possible to use webkitSpeechRecognition at Chromium without an API key. Chromium about:terms states > This Space Intentionally Blank In official builds this space will show the terms of service. The other external web services mentioned are irrelevant to both Web Speech API specification and Chromium implementation of SpeechRecognition.
,
Mar 8 2018
#10 Here are a few open source speech to text approaches which could be substituted for using an remote web service to implement Web Speech API - PocketSphinx.js Speech Recognition in JavaScript and WebAssembly https://github.com/syl22-00/pocketsphinx.js by https://github.com/syl22-00 - PocketSphinx is a lightweight speech recognition engine, specifically tuned for handheld and mobile devices, though it works equally well on the desktop https://github.com/cmusphinx/pocketsphinx by https://github.com/cmusphinx - Speech-to-Text-WaveNet : End-to-end sentence level English speech recognition based on DeepMind's WaveNet and tensorflow https://github.com/buriburisuri/speech-to-text-wavenet by https://github.com/buriburisuri -
,
Mar 22 2018
Issue 815388 has been merged into this issue.
,
Apr 18 2018
#1 > On non-Android - yes, it uses a web service. Questions: Does the web service, and/or Google and any of its affiliates, contractors, sub-contractors, including but not limited to contract research organizations a) currently, or prior to 4-18-2018, store (save for any duration of time) the biometric personally identifiable information data (specifically recorded user voice data) that is posted to the external web service by Chromium/Chrome? b) if a) is "true", how does a user request for their biometric personally identifiable information (the users' recorded voice) be purged from any and all Google and, or its affiliates, contractors, sub-contractors, including but not limited to contract research organizations storage devices, facilities or other undisclosed storage medium? c) How can users of Chrome/Chromium verify that > I guess the terms of use (about:terms of Chrome) should mention this somehow, if it does not already. > So the terms of use of Chromium (if here are any, since it is not supposed to be used - it is not a product) would probably not mention that. But the terms of Chrome should, I guess (maybe not explicitly, since other browser features use web services, like Chrome Sync, translate, push messaging... So it is probably a general chrome-may-use-web-services clause). No, Chromium "terms" does not clearly disclose that Chromium implementation of Web Speech API webkitSpeechRecognition posts (sends) their biometric personally identifiable information data (specifically recorded user voice data) to an external web service. This user of Chromium certainly has not authorized nor agreed to sending biometric PII voice data to a third-party or any web service; nor for an undisclosed third-party web service to either be the recipient of, nor retain this users' biometric PII voice data, for any purposes whatsoever. Nor is it clear what the status of of users' biometric PII data once the undisclosed third-party web-service receives the users' biometric PII voice data. Can an implementer of Web Speech API webkitSpeechRecognition/SpeechRecognition at Chrome/Chromium kindly answer the above questions and provide verifiable evidence that the above described remote web service has purged/deleted/destroyed all user biometric PII voice data which users of Chrome/Chromium have, potentially unknowingly, been sending to said remote web service under the auspices of Chrome/Chromium implementation of SpeechRecognition/webkitSpeechRecognition of Web Speech API?
,
Apr 21 2018
Why is the "Status" marked "Unconfirmed"?
,
May 31 2018
phistuck@ A Gentle Ping.. Request you to please check this issue and provide an update. Thanks..
,
May 31 2018
I am not going to check this as I do not have any internal details, I am not a Googler. However, I am going to close this as this issue tracker is not a place for general (or legal) discussions. You are welcome to express your concerns at chromium-discuss - https://groups.google.com/a/chromium.org/forum/#!forum/chromium-discuss |
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by phistuck@gmail.com
, Feb 24 2018