ZennoLab

Automate everything

用户工具

站点工具


侧边栏

本页面的其他翻译:

zh:context-recognizer

这是本文档旧的修订版!


上下文识别

上下文识别对抗Google动物园
你可能已经知道了,Google以动物命名的算法更新影响着所有的网站。在这些算法中一个主要的(也可能是最重要和最复杂的)参数就是您网站的传入和传出链接的相关性。
为了使您免受Google动物算法的影响,我们发明了全新的功能 - 上下文识别。
上下文识别将有助于确定什么是符合您主题的的网页或文本。
改变下观点,与其疯狂地在所有的地方留下链接,不如先定义你想留下一个链接的页面的主题。如果这个页面上的文字不适合您的网站的主题,你就不应该在这里做链接。

同样,您可以检查您网站上的导出链接,并检查网页的相关性。

假设你有一批可以做链接的地址。你可以使用新的Context Recognizer功能将列表依据内容分为多个类别。然后当你想推广你的网站时,就可以不用全部发送,而是选择相关类别的网址发送了。
Let's say you have a list of URLs for posting. With the help of the new Context Recognizer feature you can split your database into multiple DBs depends on context. Then, when you will need to advertise your site, you will not post to full database, but you will use only relevant URLs list.

You will be able to post an article about the car insurance to blog with subject about cars, rather than a blog, which publishes announcements of new movies.

You can parse the site (eg, blog) and find pages that are best suited to the subject of your advertising. Leaving relevant comments and posts, you will not only get relevant links but you will have much greater chance to pass moderation, it is very important for high-quality resources.
Context Recognizer is at the stage of beta testing now, despite this, it's already shows a good percentage of recognition and we will improve it

Usage

Firstly you need to specify the text for analysis. You can use article extraction feature (getting the main article action in the toolbox) for this purpose or select the required text on the page manually. Then, you can determine the general theme of the text (~20) or a detailed theme (~150) (This feature is not available for now).

You have to set two filters after that:
Specify the maximum number of themes, which Content recognizer will detect.
Specify the minimum relevance. All values below it will be considered as inappropriate. This parameter varies from 0 to 100.
Theme of text
For example you set 3 subjects and at least 30 percent of relevancy.
As result you will get no more than 3 subjects with relevancy level not less than 30 percent with your text.

Note that final result can have less number of subjects or none at all, if parser will not find any similarities between your text and known subjects.
Found subjects will be comma separated and will be saved to variable.

Testing

ProjectMaker toolbar (at the top) has a button to test Context Recognizer.

zh/context-recognizer.1345577317.txt.gz · 最后更改: 2015/07/14 15:50 (外部编辑)