On Fri, Aug 9, 2024 at 5:42 PM Charles David Lindsey < ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In papers/blaine_mooers/main.md <#934 (comment)> : > +These text areas include those of web-based email software and online sites that support distraction-free writing like [Write Honey](https://app.writehoney.com). +These text areas also include the Markdown and code cells of Jupyter notebooks and other web-based computational notebooks. +Voice In also works in plain text documents opened in Jupyter Lab for online writing. +It also works in the web-based version of [VS Code](https://vscode.dev/). +Voice In will not work in desktop applications that support the editing of Jupyter notebooks, such as the *JupyterLab* application, the *nteract* appication, and external text editors, such as *VS Code*, that support the editing of Jupyter notebooks. +*Voice In Plus* is limited to web browsers, whereas other automated speech recognition software can also operate in the terminal and at the command prompt in GUI-driven applications. + +*Voice In Plus* is very accurate with a word error rate that is well below 10\%. +Like all other dictation software, the word error rate depends on the quality of the microphone using used. +*Voice-In Plus* can pick out words from among background ambient noise such as load ventilation systems, traffic, and outdoor bird songs. + +The language model used by *Voice-In Plus* is quite robust in that dictation can be performed without an external microphone. +We found no reduction in word error rate when using a high-quality Yeti external microphone. +Our experience might be a reflection of our high-end hardware and may not transfer to low-end computers. + +Because of the way*Voice-In Plus* is set up to utilize the Speech-to-Text feature of the Google API, there is not much of a latency issue. space between way and Voice in Plus — Reply to this email directly, view it on GitHub <#934 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADTZEC5SMIC3JEW4PIXUWVTZQVAUTAVCNFSM6AAAAABIT6HN2CVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDEMZRGE3DOOJUG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>

On Fri, Aug 9, 2024 at 5:44 PM Charles David Lindsey < ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In papers/blaine_mooers/main.md <#934 (comment)> : > +You have to pay attention to the progress of the transcription if you want all of your spoken words captured. + +If the transcription halts, it is best to deactivate the plugin, activate it, and resume the dictation. +Great care must be taken to pronounce the first word of the sentence loudly so that it will be recorded; otherwise, this first word will likely not be recorded. +This problem of omitted words is most acute when there has been a pause in the dictation. + +The software does not automatically insert punctuation marks. +You have to vocalize the name of the punctuation mark where it is required. +You also have to utilize the built-in new-line command to start new lines. +We have combined the period command with the new line command to create a new command with the voice trigger of `new sentence`. + +You have to develop the habit of using this command if you like to write one sentence per line. +This latter form of writing is very useful for first drafts because it greatly eases the shuffling of sentences in a text editor during rewriting. +This form of writing is also very compatible with version control systems like git because the changes can be tracked by line number. + +The practical limit of on the number of commands it set by the trouble you are willing to tolerate in scrolling up and down the list of commands. you mean "is set by" rather than "it set by" — Reply to this email directly, view it on GitHub <#934 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADTZEC7TBXUJBJCLOT7EYMDZQVA5TAVCNFSM6AAAAABIT6HN2CVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDEMZRGE3TCOBWHE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

On Fri, Aug 9, 2024 at 5:45 PM Charles David Lindsey < ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In papers/blaine_mooers/main.md <#934 (comment)> : > +The second extension, [jupyter-voice-comments](https://github.com/Banpan-Jupyter-Extensions/jupyter-voice-comments), relies on the DaVinci large language model to make comments in Markdown cells and request code fragments. +This program requires clicking on a microphone icon frequently, which makes the user vulnerable to repetitive stress injuries. +The third extension is [jupyter-voicepilot](https://github.com/JovanVeljanoski/jupyter-voicepilot). +Although the extension's name suggests it uses GitHub's Copilot, it uses whisper-1 and ChatGPT3. +This extension requires an API key for ChatGPT3. +The robustness of our approach is that the *Voice-In Plus* should work in all browser-based versions of Jupyter Lab and Jupyter Notebook. + +### Coping with the imperfections of the language model + +One aspect of speech-to-text that it is important to bring up is persistent errors in transcription. +These persistent errors may be due to the language model having difficulties interpreting your speech. +For example, the language model often misinterprets the word *write* as *right*'. +Likewise, the letter *R* is frequently returned as *are* or *our*'. +The remedy for these situations is to map the misinterpreted phrase to the intended phrase. + +This remedy might be the best that can be done in for those users who are from a country that is not represented by the selection of English dialects available in Voice In Plus. "This remedy might be the best can be done in for those" should be "This remedy might be the best can be done for those" — Reply to this email directly, view it on GitHub <#934 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADTZEC5KUT6FEGEUARQ6MBLZQVBCLAVCNFSM6AAAAABIT6HN2CVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDEMZRGE3TIMBXGM> . You are receiving this because you authored the thread.Message ID: ***@***.***>

On Tue, Sep 17, 2024 at 12:10 PM Amey Ambade ***@***.***> wrote: Approving for publication. — Reply to this email directly, view it on GitHub <#934 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADTZEC7L5RHONRPB62QKDP3ZXBO7FAVCNFSM6AAAAABIT6HN2CVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGNJWGQ3TCMZSGM> . You are receiving this because you authored the thread.Message ID: ***@***.***>