Wiktionary:Beer parlour/2024/November: difference between revisions
Content deleted Content added
Line 263:
:::As for bots, there was a previous discussion where there was a consensus to limit allowing bots to add audios without some kind of review process for this exact reason. I can find the discussion later. [[User:AG202|AG202]] ([[User talk:AG202|talk]]) 15:25, 12 November 2024 (UTC)
::::How do you know I sent that on Discord? [[User:Flame, not lame|Flame, not lame]] ([[User talk:Flame, not lame|Don't talk to me.]]) 17:47, 12 November 2024 (UTC)
::::@[[User:AG202|AG202]]: Let me first explain how Lingua Libre works. A user gets a list of let's say 300 words and starts recording them, spending no more than 5 seconds on each of the words. After the recording phase is finished, all these words are presented in a big list. It's possible to replay words and remove any of them from the list if the quality is undesirable. Doing this, the list of 300 words may shrink to even merely 100 due to various reasons (possible pops, clicks, breathing or slurping noises in the audio, abrupt cutoffs in the beginning or end, if it's too quiet or muffled, if the intonation doesn't feel good, or just because of any other reason). With this extra review phase and discarding bad audios, the average time spent on one audio sample may increase from 5 seconds to maybe even 30 seconds. But in my opinion it's still very fast. Producing 500-1000 audios in one weekend is perfectly doable without any quality sacrifices. So ~18K or ~12K audios created over the span of a few months isn't anything particularly unusual or suspicious. That's just a normal productivity enabled by the Lingua Libre tool.
::::Now if a Wiktionary patroller spots a bad audio (e.g. a mispronounced "chameleon"), how many seconds have to be spent to resolve the problem? My suggestion is to add the <code>|bad=</code> property and move on. This way only a few seconds are spent, and the ball is on the audio recorder side again, who can re-upload a better audio and remove the <code>|bad=</code> property. Or start a dispute if they believe that everything is already fine as-is.
::::For comparison, the @[[User:Derbeth|Derbeth]]'s suggestion and the old bot policy was to make requests for bad audio samples removal from Commons, which is non-workable, because it's too labor intensive. That's the reason, why there was that consensus to suspend the bot. In my opinion, anything longer than 30 seconds per one bad audio is too labor intensive for the patrollers. --[[User:Ssvb|Ssvb]] ([[User talk:Ssvb|talk]]) 05:11, 13 November 2024 (UTC)
: {{ping|Flame, not lame}}: The [https://www.oed.com/dictionary/hydrogeniferous_adj OED] gives the pronunciation of {{m|en|hydrogeniferous}} so there was no need to guess for that one. [[User:Ioaxxere|Ioaxxere]] ([[User talk:Ioaxxere|talk]]) 03:25, 12 November 2024 (UTC)
| |||