Transcription doesn't respect language tag set in RSS

[johan456789]

Description

Take these two podcasts for example: https://pocketcasts.com/podcasts/d2012340-a822-0137-fae6-0acc26574db2 and https://pocketcasts.com/podcasts/2d1c2ae0-1136-0137-f265-1d245fc5f9cf

The first's RSS feed says:

<language><![CDATA[zh-Hant]]></language>

The seconds's RSS feed says:

<language>
<![CDATA[ zh-tw ]]>
</language>

but the transcripts seem to all default to Simplified Chinese (zh-Hans or zh-cn) for Chinese podcasts instead of respecting the language set in the tag.

My experience is limited to Chinese podcasts. I don't know if this happens in other languages.

Step-by-step reproduction instructions

1. Go to any Taiwan-based podcasts (this uses zh-Hant, this and this use zh-tw)
2. Open any new episode's transcript
3. It uses Simplified Chinese (zh-Hans or zh-cn) instead of Traditional Chinese (zh-Hant or zh-tw)

Screenshots or screen recording

Screenshots

1st Podcast	2nd Podcast

Snippets of current and expected results

好,3,2,1 大家好,我是游上杰 大家好,我是强尼 欢迎收看《解锁地球》 我刚差一点要出研究证你说 是收听不是收看 但是其实现在也可以收看 而且对啊 我现在讲收听还是收看 这是好问题  [...]

should be (translated with chatgpt)

好，3，2，1 大家好，我是游上杰 大家好，我是強尼 歡迎收看《解鎖地球》 我剛差一點要出研究證你說 是收聽不是收看 但是其實現在也可以收看 而且對啊 我現在講收聽還是收看 這是好問題  [...]

and

本期节目由大人学赞助提供 根据我们的观察 大人学的女性学员 很多在职场属于专业工作者 比方说专案经理 公司主管 企业经营者等等 那这些女生的常见困扰是 在工作的场合 [...]

should be (translated with chatgpt)

本期節目由大人學贊助提供 根據我們的觀察 大人學的女性學員 很多在職場屬於專業工作者 比方說專案經理 公司主管 企業經營者等等 那這些女生的常見困擾是 在工作的場合 [...]

Did you search for existing bug reports?

• I have searched for existing bug reports.

Device, Operating system, and Pocket Casts app version

No response

[damien-rivet]

Would it be possible for me to work on the issue?

[danielebogo]

@damien-rivet absolutely! We really value and appreciate the contributions from external collaborators 😃

[damien-rivet]

Thanks! I'll start working on it ASAP, could you assign it to me?

[danielebogo]

@damien-rivet done 👍

[damien-rivet]

@danielebogo I'm investigating the issue on my free time but I'll need some pointers if you can provide.

• Does pocket cast API for episode supports a language parameter?
  • So we can ask the API to point toward a localized generated transcript instead of a generic one
• Where are the protobuff models for the generation?
  • I noticed that the search results network responses does include a language field for the podcasts (but it's not parsed by the JSONDecoder since it's not a property of the objects)

I suspect that there might be some changes that need to happen on the API to make sure we can specify the language for all the calls 🤔

Thanks in advance for your answers.

[danielebogo]

Does pocket cast API for episode supports a language parameter?

@damien-rivet This is something we are working on (see #3553)

I noticed that the search results network responses

The search API doesn't use Protobuf models. If you look at podcasts/search you can see where it's used