Documentation – Training

MyCurator uses an AI based software learning technology we call the Relevance Engine to learn what types of articles you are looking for.  Training the Relevance engine is as simple as clicking on the Green Thumbs Up or the Red Thumbs Down that show up at the end of the article, usually after the link to the readable page or original article.    See the Options documentation, Manual Curation section, for a description of what happens to the post after it is trained.

Navigating the Training Posts Page

All of the posts found by all of the Topics go to the same Training Posts page which is a separate menu item in your WordPress dashboard, usually located below the Comments menu item.  All of the articles that meet your Keyword selection criteria for each Topic will be listed here.

As you hover the cursor over the title of each article a set of tags that show up.  The green thumbs up is used to tell MyCurator that this article is good, or one you like to see more of.  The red thumbs down is used to tell MyCurator that this article is bad and you don’t want to see more of these.

After training a few articles, MyCurator will classify all of the articles is finds in the future as good, bad or not sure.  You can filter to just good articles by selecting good from the Show All Relevance drop down.  Since training is captured by Topic, you can also filter by Topic using the Show All Topics dropdown.  Click the Filter button to perform the filtering.

Training Tips

  • After a while, you will find that MyCurator is fairly accurate at classifying articles as good, bad and not sure.  You should periodically review the bad and not sure articles though and see if any should be classified as good.  Make sure you set the Option “Keep Training Posts for how many days?” to be at least the number of days that you periodically review the bad and not sure articles or they could be deleted before you get to them.
  • You don’t need to train every article, you can focus on those that seem particularly good or bad.  Once the system is classifying your articles fairly well, you can cut back and just review them on a periodic basis as above.
  • You can ‘bias’ the Relevance Engine towards finding more good or more bad articles by training more articles as good or bad.  From the list of Topics, choose a Topic and at the bottom of the Topic page you will see MyCurator Relevance Statistics.  This gives you a count of the articles that have been trained as good or bad for this topic.
  • As what you are looking for changes, you can begin to train the system more and get it to follow your changing requirements.  Similarly, if it is producing good articles that are just right, you can almost quit training and it will keep finding similar types of articles.

Automatically posting to the Live site

As you train the system, future articles will post with a tag of good, bad or not sure by the Relevance Engine. You can train any of these however you see fit. Eventually, you may want the good articles to Auto Post to your main site.  See the Auto Post documentation for more details.

You can still train the system with articles auto-posted to the main site by clicking on the Red Thumbs Down to tell the system that the article is not what you wanted, and it will be removed from the site as well. You should also periodically browse the Training page and click the Good Thumbs Up for articles you like or the Red Thumbs Down for articles that you don’t want to see.