Transcript semantic parsing

[tonytamsf]

fix #4935 Parsing of podcast:transcript with a simple unit test. After this is merged, my next PR will be able to display 2 lines of transcript in the CoverFragment

@ByteHamster let me know if you wanted more unit tests

[ByteHamster]

Thanks! I added some comments below

[ByteHamster]

Sorry for the late reply. I'm currently on a business trip that is quite packed with things to do. Finally found a bit of time. Comments below :)

[tonytamsf]

Sorry for the late reply. I'm currently on a business trip that is quite packed with things to do. Finally found a bit of time. Comments below :)

No worries, enjoy your trip.

[ByteHamster]

Looking at the code again, I think we should make the properties of TranscriptSegment final. That prevents later "misuse" by other developers. As it currently is, developers could bring the Transcript object into an invalid state by adding a transcript and then modifying its properties. If a segment is not modifiable, this can't happen. What do you think?

[tonytamsf]

Looking at the code again, I think we should make the properties of TranscriptSegment final. That prevents later "misuse" by other developers. As it currently is, developers could bring the Transcript object into an invalid state by adding a transcript and then modifying its properties. If a segment is not modifiable, this can't happen. What do you think?

I assume you mean these properties below? In order for a future PR to dynamically display transcripts, I will need to modify the contents of TranscriptSegment

public class TranscriptSegment {
    private long startTime;
    private long endTime;
    private String words;
    private String speaker;
    private Boolean trimmed;

Here is a future PR

https://github.com/tonytamsf/AntennaPod/pull/6/files#diff-14d98525c5df8d24fa975f628614e462438e8ccb86a08dab5f5bafabb2ef09cfR411

1. Depending on the screen size, I will need to break up segments using setWords() and setStartTime()
2. When a user scrubs the audio, they might land in the middle of a segment and I will need to redo the segments.

[ByteHamster]

Instead of modifying the existing transcript specifically for the UI, I would rather create a copy with the new line breaks. That way we ensure that the broken-up transcript doesn't somehow end up in the other UI pages where it needs to be broken up differently.

[tonytamsf]

Making a copy seems very wasteful of memory and execution time.

As I dig into the details of displaying transcripts and supporting users scrubbing to random locations, I end up needing to dynamically chunk up the transcript segments.

Another way I could do this is probably with a cursor that walks the transcript segments for display.

I will give that a try

[ByteHamster]

Don't we need a copy anyway because we need to re-layout on screen size changes? There we probably want to start fresh instead of keeping to modify the same transcript on subsequent size changes/rotations, which would otherwise look different every time and probably even degenerate at some point.

How large are these objects? Is having them twice really such a big problem?

[tonytamsf]

How large are these objects? Is having them twice really such a big problem?

I'll start implementing without a copy and see if I can solve it with a cursor instead. I'm not sure why I personally cannot visualize how to solve it by copying since the transcript segments will need to be layout more than once since the user can scrub the time, fast forward, rewind and rotate the screen any time. Doesn't seem to make sense to copy and redo the layout for the entire Transcript when only the current and next N segment of text needs to be rearranged.

[tonytamsf]

Looking at the code again, I think we should make the properties of TranscriptSegment final.

This is done, the values are final in TranscriptSegment

[tonytamsf]

@ByteHamster What do you think of merging?

[ByteHamster]

I did a few changes, let me know what you think :)

About displaying, I think we need to discuss a bit again. In my opinion, most users don't need the live transcript, just the full one for search/sharing/etc. So displaying it on the main cover screen by default adds quite some complexity and "movement" for everyone while helping just a few users. Maybe we can discuss this at the community call next Saturday.

[tonytamsf]

I did a few changes, let me know what you think :)

looks good, thanks!

[ByteHamster]

Thanks!