Forum - View topicEncyclopedia Suggestion List.
|
Goto page Previous Next |
| Author | Message | |||
|---|---|---|---|---|
|
Shiroi Hane
Encyclopedia Editor
 Posts: 7580 Location: Wales |
|
|||
|
It wouldn't work with the existing system since the data and the source can only be attributed to one person, but perhaps Dan could extend the new precision system (which does allow multiple submissions for the same role as long as the precision is different) to allow for a second "blank" precision when the source field is blank?
The problem I can see with this, is it would result in two "i"s in the admin/error/sources view with no clear distinction between them. |
||||
|
||||
|
Dan42
Chief Encyclopedist
 Posts: 3786 Location: Montreal |
|
|||
|
For this case please submit an endorsement with the additional source info.
|
||||
|
||||
Devil Doll
 Posts: 656 Location: Germany |
|
|||
|
That's what I already did (mostly here for staff), and what I stumbled over while replacing "precision endorsement" with overriding contributions (thus I wondered whether the same overriding mechanism ought to be used for this).
- - - - - - - - - - - - A completely different topic: Entries at ANN, be that persons or companies, tend to have more than one kanji name throughout their existence. Seiyuu change their stage name, writers use pseudonyms, companies are given with a more or less complete specification within kanji source pages (such as "St." vs "スタジオ"). The existing mechanism of replacing kanji input by romanized values does a good job but is limited to the one kanji name any entry can have at most. (In the kanji field, that is - there are kanji names in the birth name field of a person but they're not used for auto-translation, and most likely not even usable given their special syntax.) The translation feature doesn't do (bidirectional) substring matches for company names; even if the difference between the known kanji name of a company at its ANN entry and the given input is no more than a "・" character (a word separator for katakana names) the match will fail. I wonder whether this matching could produce even better results than it does today (while the "training" of the translation tables is still in process). Doing substring matches for company names or ignoring word separators might be a first step; making the birth name field accessible for person name input conversion (only when the first matching attempt for this person's kanji name failed, i. e. with lower priority than current kanji names) might be another one. Putting links on translated company names (like for translated person names in the case of a non-unique kanji name) or on the message "this company exists in the database" would be another nice idea: When I enter a company name in Japanese notation and don't get a match I have to guess its meaning, and when the re-verify confirms my guess I would like to surf to this company and check which kanji name it already has (and whether I can contribute this one in the process). Currently not even the confirmation message after contributing a company credit has a link to the company (unlike the confirmation message for person credits) so I have to return to the anime page and check all contributed companies for potentially missing kanji names afterwards as the context knowledge during the (mass input) contribution process has been lost. I don't know exactly how this whole matching is implemented in ANN's database, therefore I can't know whether allowing more than one kanji name for companies (such as a comma-separated list or whatever) would improve things (as to "emulate" the substring match by the contributor) or whether including the birth name field for the person name translation would even be possible. I don't have a concise wish in this paragraph, I'm just "thinking aloud" as I get the impression that the current implementation still has some space for improvement. Questioning the current concept of one kanji name for each entry or splitting the birth name into four fields will most likely be no option, so I guess there's not much to expect in this area anyway. - - - - - - - - - - - - Another completely different topic: There are many persons without a kanji name but with sources that are kanji only in the ANN database. How did the contributor enter this credit? Apparently by translating the kanji to roumaji themselves. I wonder whether the ANN input process could learn to perceive such a situation. For example: A person without kanji name gets a credit from a source with a ".jp" domain such as D2_STATION or livedoor wiki (both of which I see frequently in use as decent sources). This kind of check would not be fool-proof but I'd consider an 75% success rate worth the effort (as it would be easy to implement). Would it make sense to check for this situation and then display a message to the contributor asking them to contribute the missing kanji name for the person (as they apparently believe to be able to do this)? There are other cases when a contributor decides a person to be the correct one even if they have diverging kanji names (e. g. when they have the same roumaji name and the same type of staff credits). In another thread we discussed that in this case the contributor might prepend the source URL with the kanji name they considered a match in round brackets, like
Would it make sense to add this recommendation to the text that explains how the source URL should be selected (i. e. append it to the "If you found the information online, provide the URL of the exact page where you found it (not just the home page of the website, please)" paragraph)? - - - - - - - - - - - - My personal favorite wish for an improvement of the contribution process would still be an automatic check whether contributing a kanji name creates a kanji name duplicate (for companies: check always; for persons: check only when both family name and given name are filled as it could be the first of two steps to separately set these fields; I'd prefer being able to set both in one transaction and with one source URL for both but that's certainly a mere nice-to-have). Becoming aware of the collision in this very moment would allow the contributor to see the potentially duplicate person as early as possible (i. e. before even creating the kanji name duplicate). The main problem of any feature request in this direction (while rather easy to implement) is that it would significantly increase the number of duplicate person reports, and that's of no use given the existing backlog in this department. So this idea will probably rest until ANN has more encyclopedists. |
||||
|
||||
|
littletykes
Posts: 33 |
|
|||
|
I have 2 suggestions regarding error reports.
1. "View all your entries" type of function that lets users see a sort of index that lists all the entries where the user has entered an error report. I've made several error reports and when I review them (whether to add more information or just to see if the error report is still valid/unresolved), sometimes i forget other entries I've submitted reports to (and thus could not get back to them). This can also be applicable to all the information that the user has submitted beforehand, and help the original contributor get to it faster, kind of like the "view all your posts" option in the forums, but for all the entries you've made to the encyclopedia.
I too would be willing to hear advice on my submissions, and hope others will be as well. In this light, 2. Receiving notifications about error reports made on your contributions. Sometimes, people don't really know that the entries they've made have received error reports. This would likely hasten the editing of the encyclopedia errors if the OC's would learn about the report. |
||||
|
||||
|
doc-watson42
Encyclopedia Editor
 Posts: 1709 |
|
|||
This has recently been implemented: Go to the "Encyclopedia" tab—> "and more..."—> the "Contribute" section—> "Lookup all the contributions and error reports you made". Or were you thinking of something a bit different? |
||||
|
||||
|
Devil Doll
Posts: 656 Location: Germany |
|
|||
|
The contribution table is a good thing, only that it displays wrong data at times so that you can't always find the entries in question. Getting a list of links to entries would be more user-friendly than having to search for (sometimes even non-existent) icons. But fixing the bug here would be sufficient for me, additional GUI comfort would merely be a nice-to-have.
Getting a message when someone flags one of my contributions as error would be something I'd like to have as well as the contributor can fix certain things faster than an encyclopedist. |
||||
|
||||
|
littletykes
Posts: 33 |
|
|||
Didn't know about this, thanks!  It certainly helps, even if it's a bit cumbersome to use.
|
||||
|
||||
|
littletykes
Posts: 33 |
|
|||
(I do hope this will be fixed soon. I just checked my contributions and found that a couple of them didn't belong to me, haha.) EDIT: my mistake! i found the reports i made. sorry. As i said it's a bit cumbersome to use, but the current functionality is okay. And yeah, a more user friendly GUI would be great, something similar to that blue "i" icon that was introduced a while back. |
||||
|
||||
|
Devil Doll
Posts: 656 Location: Germany |
|
|||
|
Another "nice to have" but could support encyclopedists during the process of checking potential split request cases in the future:
When a credit is being entered using a kanji name for the person, this kanji name is translated to a person entry first, and the input form is then re-displayed to the contributor so that they can check whether the person founding by this translation would be a plausible candidate. This form then has to be submitted again in order to complete the contribution. This procedure is reasonable but has one slight disadvantage: The information that this person was identified by a kanji name is lost in the process. My suggestion would be that the contribution process would keep track of this fact (say, in a hidden HTML element of the contribution form section for each credit "unit") and then prepend this kanji name in round brackets to the source URL given by the contributor. Example: Let's say I credit "仕上檢査:辻美紀子" and give the source URL http://www1.vecceed.ne.jp/~m-satomi/HAIBANERENMEI.sub.html. The submit form translates 辻美紀子 to Mikiko Tsuji and asks for confirmation. What I'd like to have is that the form keeps the information about the original kanji name input "辻美紀子" and automatically concatenates this information to a source string "(辻美紀子) http://www1.vecceed.ne.jp/~m-satomi/HAIBANERENMEI.sub.html" which would show three things: 1. The original name for this submission has been entered in kanji, 2. these are the kanji to be used to find and double-check the credit at the given source URL, 3. this credit entry confirms the kanji given for the person. Note that this URL would have to be formed separately for each contribution during a mass-input form processing, prepending the individual kanji names to the given source URL. This mass-input procedure is the reason for this request: Mass-input is extremely efficient but has the side effect of giving all submitted credits the same source string which then can't contain the kanji name used during contribution. Therefore the contribution process would have to automatically insert this kanji name into the individual credit, thus creating individual source strings for credits that were submitted during one mass-input step. Without mass-input there wouldn't be a problem, I'd then be able to form these "extended source strings" manually for each contribution. |
||||
|
||||
|
doc-watson42
Encyclopedia Editor
Posts: 1709 |
|
|||
You're welcome. 
|
||||
|
||||
|
Devil Doll
Posts: 656 Location: Germany |
|
|||
|
I'd like the encyclopedists to have two scripts that list all persons that have more than 0 credits but all of their credits are flagged as errors, sorted by number of credits ascending, and the same for companies.
These are entries that should be purged of all their credits (as they're most likely duplicates) as soon as all of their credits have been assigned to the correct person/company instead (which should be easy to check given the error comment). This could be checked periodically (say, once every couple of weeks) by some encyclopedist (Shiroi Hane has delete privilege and no more would be required for this  ). The list order would be to sort these entries by the estimated amount of work for checking.
An example for such a situation would be the person Lantis that has 1 credit for music production which I assigned to the company Lantis but can't delete the wrong entry. |
||||
|
||||
|
Shiroi Hane
Encyclopedia Editor
Posts: 7580 Location: Wales |
|
|||
|
I can't 'delete' the credit, only remove it from view (which I have done), but I can't remove the dummy entry altogether.
|
||||
|
||||
|
Devil Doll
Posts: 656 Location: Germany |
|
|||
|
The difference between "deleting" and "hiding" is irrelevant for both visitors and contributors; if the ANN data history requires keeping this entry, so be it.
Having the person purged from all credits indicates that it should not be given any (more) credits (as it was created by getting a credit which it "lost" later), that's why hiding the wrong credits gives a valuable message to contributors. Not being able to delete "empty" persons is understandable (they may "come into existence" later) while a bit more unfortunate. I'd like to see on their page a link to the person/company where the credits went to ("potential duplicate of", similar to the concept of "also known as" but including a pair of flag [person/company] and a number in order to form a link from these; the source entry could explain the relation further such as listing the potential kanji readings) which would both explain the existence of this entry and serve as a warning for future contributors. Once this person actually starts to exist the warning/link could be removed via a posting in a corresponding forum thread (which would be a very rare event). - - - - - - - - - - - - - - - - - - - - - Takenori TSUGUMA (津熊 健徳, 11 staff for key animation) is a spelling error of Takenori TSUKUMA (津熊 健徳, >50 staff for key animation) as "Tsukuma" is the only valid EDICT reading for 津熊. I'm posting this in the suggestion thread because the Roumaji similarity logic didn't warn me of the potential duplicate, it doesn't consider "gu" and "ku" as "similar" (which it perhaps should as e. g. 山口 could be Yamaguchi or Yamakuchi even though the latter would be very rare). I remember issues with "ka" and "ga" as well (柳川 could be Yanagawa or Yanakawa); perhaps g/k is a potential ambiguity in combination with all vowels? Kentoku TSUGUMA (津熊 健徳, 5 staff for key animation) is a case of different kanji reading (健徳 allows Takenori and Kentoku) that the Roumaji logic sure can't find; I just mention this person here for the sake of completeness. - - - - - - - - - - - - - - - - - - - - - Just wondering: Would the current data model allow to implement a new feature that allows the creator of a person/company entry to delete that very entry if and only if it has no credits, i. e. undo the creation of this entry as long as it hasn't been used by other contributors? (So that deleting your own person/company creations would be possible just like deleting your own credit entries.) I would have needed such a function a few times already... |
||||
|
||||
|
Devil Doll
Posts: 656 Location: Germany |
|
|||
|
The ANN contribution routine does a nifty check for new persons whether the family & given name are likely choices for this position in order to prevent Japanese names being entered in the wrong order.
Unfortunately this logic is pretty useless for Korean names as there's so much garbage in the database that both orders tend to be around 50%. It would improve the effect of this logic if an encyclopedist with knowledge about Korean names did a massive cleanup for names in the wrong order (and add hyphens for two-part given names in the process) so that people who don't know Korean names could get a higher probability of detecting the name order used by the source they're using for contribution. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Would it make sense to create a new subforum within the encyclopedia forum reserved for "split request" threads? (I have bookmarks of about 40 more candidates to be processed.) |
||||
|
||||
|
Devil Doll
Posts: 656 Location: Germany |
|
|||
|
Shoko NAKAMOTO (中本 尚子) doesn't have any credits, so apparently she was considered a wrong romanization.
But she still has a kanji name, making her a duplicate of Naoko NAKAMOTO (中本 尚子, >30 staff for animation), thus slowing down the contribution process. What about a report script checking for persons with kanji name but no visible credits, and an encyclopedist checking its output, say, once a month? - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - The user preference section of ANN provides a toggle flag "For Japanese cast/staff, sort titles by year instead of alphabetically". I would like this flag to have an effect not only on person pages but on company pages as well. |
||||
|
||||
 |
All times are GMT - 5 Hours |
|
|
|
Powered by phpBB © 2001, 2005 phpBB Group