4 ms·

I note that the cancellation is in three parts: “has been cancelled”, “due to” and finally the reason, each said in slightly different ways. Why not just say “

I note that the cancellation is in three parts: “has been cancelled”, “due to” and finally the reason, each said in slightly different ways. Why not just say “…has been cancelled” and then announce the reason with a single, complete sentence in a longer piece of audio?: “This is due to livestock on the track.”, “This is due to a staffing shortage.”, etc I get that it’s impossible to foresee everything the system might have to say and to record all announcements in one sitting with the same person in exactly the same way. However, is there an important technical reason for why these systems break up sentences so that they are spoken by different people?

rippercushions 4y ago
Because it wouldn't sound like an official robot announcement without the awkward pauses. On the Jakarta metro, most stations have a sponsor's name attached to them. The sponsors change every now and then, so the new names have to switched in, but they put zero effort into equalizing the audio. So you get treated to (lovely female voice) "Stesen berikutnya, Lebak Bulus" (dude yelling) "GRAB". (lady) "Lebak Bulus" (dude again) "GRAB."
avianlyric 4y ago
That’s because the SQL powering this demo only selects the “has been cancelled” announcement section. Presumably they have other snippets they can put in there like “has been delayed”, or “has terminated early” etc. Depending on the nature of the incident, you may not have, or need, a reason. It’s probably not worth putting a reason in if the train has only bee delay 5 mins, you may not even know the reason for the delay in time for the announcement.
- gorgoiler 4y ago
  The goal is to avoid as many mid sentence voice changes as possible. That’s hard if your announcement is one sentence: The X service from A to B has been Y’d due to Z I’m wondering if, instead, the announcement could be: The X service from A to B has been Y’d. This is due to Z. I’ve said this before — there aren’t that many different train services. You could record every version of both sentences in full, and almost certainly record every version of every sentence in full by the same person in a tractable amount of time. I don’t think the problem scales as much as these engineers think it does.
  - avianlyric 4y ago
    Sounds like a lot of work, and cost (voice actors aren’t free) for a some pretty marginal gain. If people can understand the announcements easily, then they’ve done their job. Anything extra is just that, extra.
    - gorgoiler 4y ago
      The number of routes is bounded by the number of drivers you hire, all of whom draw an annual salary, and who operate rolling stock that is bought and fueled multiple times on a daily basis. For each of those you have to pay a voice artist to say the name of the service N times for the N types of notable announcement. Once a season. It’s nothing. As well as that, factor in the number of passengers who will listen to these announcements, and think about the impact getting the announcement perfect instead of just done. It really is the sort of thing worth getting right. But yes, you will probably have one naive engineer in the middle of all this saying “repetition is Bad(tm) so let’s factor it apart and build it programmatically!” oblivious to the whole system.
      - avianlyric 4y ago
        > But yes, you will probably have one naive engineer in the middle of all this saying “repetition is Bad(tm) so let’s factor it apart and build it programmatically!” oblivious to the whole system. Sure, because organisations like ScotRail are notorious for being engineer led, and clearly have plenty of extra cash sloshing around that couldn't be better spent improving the actual rail service.

rippercushions 4y ago

Because it wouldn't sound like an official robot announcement without the awkward pauses. On the Jakarta metro, most stations have a sponsor's name attached to them. The sponsors change every now and then, so the new names have to switched in, but they put zero effort into equalizing the audio. So you get treated to (lovely female voice) "Stesen berikutnya, Lebak Bulus" (dude yelling) "GRAB". (lady) "Lebak Bulus" (dude again) "GRAB."

avianlyric 4y ago

That’s because the SQL powering this demo only selects the “has been cancelled” announcement section. Presumably they have other snippets they can put in there like “has been delayed”, or “has terminated early” etc. Depending on the nature of the incident, you may not have, or need, a reason. It’s probably not worth putting a reason in if the train has only bee delay 5 mins, you may not even know the reason for the delay in time for the announcement.

gorgoiler 4y ago
The goal is to avoid as many mid sentence voice changes as possible. That’s hard if your announcement is one sentence: The X service from A to B has been Y’d due to Z I’m wondering if, instead, the announcement could be: The X service from A to B has been Y’d. This is due to Z. I’ve said this before — there aren’t that many different train services. You could record every version of both sentences in full, and almost certainly record every version of every sentence in full by the same person in a tractable amount of time. I don’t think the problem scales as much as these engineers think it does.
- avianlyric 4y ago
  Sounds like a lot of work, and cost (voice actors aren’t free) for a some pretty marginal gain. If people can understand the announcements easily, then they’ve done their job. Anything extra is just that, extra.
  - gorgoiler 4y ago
    The number of routes is bounded by the number of drivers you hire, all of whom draw an annual salary, and who operate rolling stock that is bought and fueled multiple times on a daily basis. For each of those you have to pay a voice artist to say the name of the service N times for the N types of notable announcement. Once a season. It’s nothing. As well as that, factor in the number of passengers who will listen to these announcements, and think about the impact getting the announcement perfect instead of just done. It really is the sort of thing worth getting right. But yes, you will probably have one naive engineer in the middle of all this saying “repetition is Bad(tm) so let’s factor it apart and build it programmatically!” oblivious to the whole system.
    - avianlyric 4y ago
      > But yes, you will probably have one naive engineer in the middle of all this saying “repetition is Bad(tm) so let’s factor it apart and build it programmatically!” oblivious to the whole system. Sure, because organisations like ScotRail are notorious for being engineer led, and clearly have plenty of extra cash sloshing around that couldn't be better spent improving the actual rail service.