The Movie Database Support
Support → Website Support
Translations: Replacing Non English to English
posted by Joe RoseSTAFFMOD on June 23, 2015 at 10:59 AM
Travis;
As you can see by the Leatherboard I am editing a lot of records at the moment. Specifically (at this moment Original Languages). However I do come across the validation requirement to change add the English Title and/or Overview when they are in another language other that English as you know. So with title there is no problem because it has a separate entry from the Original (language) Title. In the case of the overview which is in the {native} language, I can translate it, but then appear to have to destroy the overview of the native language with the English version. I was expecting your application the put the native overview in it's translation "language" container, if not already there, and if it is not, create the translation container as required thus retaining the original overview. Comments on how I should proceed?
A Thought:
Should not the Orig Language, Status, Adult Movie, and other non-language dependent items in the form(s) be excluded from translation validation. These are what they are (though they may appear in drop-downs (where applicable) in the browsing language) must still use the same id code behind. Comments?
Could you be so kind as to ask your moderators to avoid altering/locking my edits for the time being as I am running some complex SQL procedures at my end to analyse for missing or possible invalid entries. Some of these require that I change/add one data element before being able to evaluate another and/or doing in dept research on movie records later, if required. With the 12 hour WEB to API delay their changing my edits are going the make further analysis/changes/additions useless. I log all my transactions locally so I can monitor my own work thus, if I make a bad data call, I can fix it subsequent to or after any rescrapes.
Joe
Reply by Joe Rose
on June 23, 2015 at 1:49 PM
Travis;
I see I was part of the problem! OOPS The movies in question all seem to be in the Spanish/Mexican set. The user(s) entered Spanish In the English Overview. Might be a legacy thing. Not so many I cannot fix them.
Might this be possible at your end....
If the overview is entered into the wrong language container can it be auto translated (i.e. English) (and flagged as an auto translation to be reviewed by others more familiar with the language) . If the translation container does not exist it is auto created (it appears you do this in some cases) and the auto translation added. If the translation container does exit and/or there is no overview it is translated and then inserted. ( User message something to the effect: "Your overview was being entered under the wrong language translation. It has been entered under the "Spanish" translation."
So if I am browsing/entering in the English container and enter Spanish in or it already exists in the English Overview the Spanish container is created if it does not exist. If it does exist but the Spanish Overview is missing then the Spanish Overview is added it that container it is auto translated and the English translation replaces the Spanish in the English container. Though the auto translators are not prefect it's a staring point, hence the flag to denote the auto translation.
Joe
Reply by Travis Bell
on June 24, 2015 at 12:40 PM
Yup, related to some legacy data. I mentioned it here.
Trying to machine process some of this is definitely possible but there's some external variables at play here (like our 3rd party language detection service). I have some demo tech that I've been playing with locally to bring the language checks in house but given everything I have going on, it's one of those things that's relatively low priority given the current setup works. It would indeed be nice to run through all of the translations to clean some of this up, and hopefully I'll find some time to do it.
Reply by Joe Rose
on June 25, 2015 at 11:02 PM
Thx Travis;
I've been thinking of adding the Yandex API to my app. Have been doing some testing on their sampler but not enough testing yet to commit to them. Who are you using and do they just detect or will they translate too?
I have found even Google can be iffy as I a got "Maori" on a "Spanish" title once. It seems that there is better accuracy if there is at least ~ twenty words in the sample. I'm doing testing on this with some of the overviews. Yandex only detects/translates 40 languages currently. A problem with all the Indian (Hindi, Tamil, Telugu) languages (all of which Yandex doesn't have yet) and number of movies the Indians produced.
BTW, I am finished the original language changes to all the East Indian movies, Mexico and Central American Countries and all the South American Countries -working on the last (Brazil) and should be finished it by tomorrow. I have had to set my re-scrape timer to 16 hours because 12 ain't enough. Probably my doing??
Joe