扩展:共享资源元数据

This page is a translated version of the page Extension:CommonsMetadata and the translation is 25% complete.
MediaWiki扩展手册
CommonsMetadata
发行状态: 稳定版
实现 API
描述 尝试从共享资源页面中提取元数据
作者 Brian Wolff (bawolff留言)
兼容性政策 快照跟随MediaWiki发布。 master分支不向后兼容。
MediaWiki 1.25+
PHP 5.4+
数据库更改
许可协议 GNU通用公眾授權條款2.0或更新版本
下載
  • $wgCommonsMetadataForceRecalculate
  • $wgCommonsMetadataPublicDomainPageUrl
  • $wgCommonsMetadataSetTrackingCategories
季度下載量 71 (Ranked 85th)
正在使用的公开wiki数 1,055 (Ranked 247th)
前往translatewiki.net翻譯CommonsMetadata扩展
問題 开启的任务 · 报告错误

共享资源元数据扩展是一个从维基共享资源中提取元数据的尝试,但也可以用于其它维基媒体项目。 它基于图片描述中的模版和分类来添加一些额外信息到imageinfo API中。 它被许多扩展和工具(例如 Extension:MultimediaViewer Extension:VisualEditor Extension:移动前端 移动内容服务(MCS))用来提供更好的边框或图像选择对话框。

目前的这个扩展是一个暂时的解决方案,它最终会被维基共享资源中的维基数据取代。

安裝

  • 下载文件,并将其放置在您extensions/文件夹中的CommonsMetadata目录内。
    开发者和代码贡献人员应从Git安装扩展,输入:cd extensions/
    git clone https://gerrit.wikimedia.org/r/mediawiki/extensions/CommonsMetadata
  • 将下列代码放置在您的LocalSettings.php 的底部:
    wfLoadExtension( 'CommonsMetadata' );
    
  •   完成 – 在您的wiki上导航至Special:Version,以验证已成功安装扩展。

动机与设计选择

参见mailarchive:wikitech-l/2013-August/071593.html

此扩展的假想如下。

  • At some point in the future, wikidata will take over handling metadata at commons. In order to avoid disruptive changes, which will soon need to be changed again, the extension should work with commons metadata as it currently is (so not introducing new parser functions). Hence screen scraping.
  • The content of many of the fields on a commons description page include rich formatting (In particular: Links, italics, bold. In some cases more complex things like embedded images)
    • As a result, extension outputs parsed html (wikitext sucks, plain text doesn't capture the data)
    • Futhermore, the data tends to be formatted for human display, rather than (for example) machine formatted dates. When the date field says something like "circa 1600s", its hard to convert that to a precise date (otoh, many examples can be).
    • To carry that forward, also apply formatting to exif metadata, which is controlled on wiki (For example, commons links the camera name to a wikipedia article)
  • If we can't extract info from the description page, but the file has the author tagged in exif/XMP/iptc metadata, we should use that as a fallback.
  • Ideally such a system would be as commons-inspecific as possible, with the commons and non-commons part separated.
  • Commons description pages have multilingual descriptions. Lots of users probably just want one language.
    • In this implementation, it applies per language conventions to dates and things. Additionally for explicitly multi-lingual fields (description), there is an option to return all, or just a single language. Even in single language mode, some things are still language specific (like the thousands seperator on numbers)

Configuration

parameter default description
$wgCommonsMetadataSetTrackingCategories false Add the following tracking categories to file pages when the corresponding information is not provided either via templates on the file page of (for some of these) EXIF metadata:
  • Files with no machine-readable license (commonsmetadata-trackingcategory-no-license)
  • Files with no machine-readable description (commonsmetadata-trackingcategory-no-description)
  • Files with no machine-readable author (commonsmetadata-trackingcategory-no-author)
  • Files with no machine-readable source (commonsmetadata-trackingcategory-no-source)
  • Files with no machine-readable patent (commonsmetadata-trackingcategory-no-patent) (for 3D files)
$wgCommonsMetadataPublicDomainPageUrl https://commons.wikimedia.org/wiki/Help:Public_domain Link used for 'license' attribute in schema.org markup for files in the public domain.
$wgCommonsMetadataForceRecalculate false Force calculation of metadata even when the image is from a foreign repository which would provide it. This is meant for local development.

测试

  警告: If you're developing or testing this extension, we do NOT suggest you copy the Commons templates for image metadata, as they take extremely long to compile and have complicated dependencies like Scribunto. Instead, get an expanded version that has only wikitext/HTML and put in the various parameter references (or don't) manually. You can find an example (to be used with Special:Import) here. 或用Vagrant,里面默认含有一些模版。

When testing with remote images (e.g. Commons images if you have enabled $wgUseInstantCommons ), you can set $wgCommonsMetadataForceRecalculate = true; to force CommonsMetadata to parse the description page of the image and extract the metadata (normally, if the remote repository had CommonsMetadata installed as well, it would just copy the API output from there).

用法

Use the imageinfo API, and include extmetadata as an image info property specified via iiprop.

示例用法:

https://commons.wikimedia.org/w/api.php?action=query&prop=imageinfo&format=json&iiprop=extmetadata&iilimit=10&titles=File%3ACommon%20Kingfisher%20Alcedo%20atthis.jpg

View this example in the API sandbox:

https://www.mediawiki.org/wiki/Special:ApiSandbox#action=query&prop=imageinfo&format=json&iiprop=extmetadata&iilimit=10&titles=File%3ACommon%20Kingfisher%20Alcedo%20atthis.jpg

返回的数据

The extension currently provides the following items in the extmetadata field of the response (the field names were chosen, where possible, to follow the IPTC-IIM format used in EXIF headers):

  • ImageDescription - image description
  • Artist - author name (might contain complex HTML, multiple authors etc)
  • Credit - source
  • DateTimeOriginal - time of creation (space-separated ISO 8601 timestamp whenever possible, but can be any other textual description of a date, possibly with HTML mixed in)
  • ObjectName - title (for a book/painting; otherwise just the file name)
  • Permission - contents of the Permission field of the template. Can be a lot of things (license template, OTRS id, details on how to attribute...)
  • AuthorCount - the number of templates with authors (e.g. Book, Photograph...). The number of actual authors might be higher if a template describes multiple authors in a single string.
  • GPSLatitude - latitude
  • GPSLongitude - longitude
  • GPSMapDatum - coordinate type (only WGS-84 supported for now)
  • LicenseShortName - short human-readable license name
  • LicenseUrl
  • UsageTerms
  • Copyrighted - True or False (for public domain images)

For multi-licensed images these values are currently unreliable.

  • Attribution - custom attribution that should replace Artist + Credit (can also originate from the Information template)
  • AttributionRequired - booleanish (phab:T86726), tells whether there is a legal requirement to attribute
  • NonFree - booleanish, true means the image is not under a free license. (Used for non-Commons images only.)

Other data:

  • CommonsMedadataExtension - contains the metadata parser version number; mostly for internal use
  • License - a best guess at the license of the image (mostly for internal use by MediaViewer, might change; LicenseShortName is probably more reliable)
  • Categories - a |-separated list of the categories of the image.

Based on parsing category names, probably won't work for images not hosted on Commons.

  • Restrictions - reuse restrictions such as trademarks or personality rights; an array of keywords (the class names from this table, without the restriction- prefix). See also the restrict-* icons in MediaViewer.
  • DeletionReason - if set, the template is being considered for deletion.

(Based on the nuke template, probably not reliable outside Commons.) It contains a deletion reason, but it is phrased to be applicable for a log entry, so it might be misleading (e.g. past tense when actually it is not yet decided whether the image will be deleted).

参见