I’m trying to use DT beta demo. I’ve read a lot of lit on the app, and how it compares to others of its kind.
I’d really like to make DT AI work for me, but the more I think about it (and use it) the more I’m not confident about the linchpin: setting up an intelligent, appropriate folder hierarchy. DT screencasts say the more you folder and subfolder the more AI works on down the line; and you don’t need to drag new content into folders, as the AI will do it for you. Sounds good, but what about files that include desperate content of paragraphs and quotes that can (and should) be filed in various subfolders? So much for AI doing the work for you.
Comment: Your remarks about the relationship between good management of organizational structure and improved performance of the AI really concerns the Classify AI. Classify examines the contents of the documents you have placed into groups, and the more unique and coherent your groups are, concerning the words and patterns of word usage in the contents of those groups, the better will be the performance of Classify in suggesting one or more filing locations for new content.
See Also and See Selected Text compare words and the contextual patterns of words in the document you are viewing to all the other documents in the database. Suggestions of similar documents are made, regardless of their filing locations in the database.
You raised the issue of a large document, such as a book-length PDF, that may deal with a number of topics.
I often let the two AI features work together for me if I want to extract useful portions of the text in a large file, and then file the extracts appropriately.
Step 1: Clip an interesting passage from the large document into a new rich text note. In the case of a PDF, add the Page Link to the note as documentation of the source of the clipped text.
Step 2: Invoke Classify on the new rich text note. Examine the suggested group locations, and if one or more seem appropriate, move the note into the group(s).
But if the Classify suggestions don’t seem especially appropriate, the See Also list may turn out to be helpful. Command-click on one or more of the ‘similar’ documents in the See Also suggestions so as to add them as new tabs. Open a tabbed document and press Command-R (Reveal). The group location of that similar document will be displayed, and might turn out to be a good filing location for your new note. (The Tags bar in public beta 8 also shows the group organization of the tabbed document, so if one hasn’t excluded groups from tagging, it’s not necessary to invoke the Reveal command.)
I confess that I’m often dilatory in filing new content, so at any time I’ll have a lot (gulp, thousands) of documents that haven’t yet been organized. I make a lot of use of See Also and See Related Text. Fortunately for me, See Also forgives my organizational sloppiness and still provides useful results. Other tools including Search and Option-Click also forgive organizational sloppiness.
There’s still another AI feature, Group, that I sometimes use to seed a classification system for content that differs topically from the existing organized content in a database. Group will examine a collection of selected documents and attempt to define new groups of related documents. It won’t do the whole job and it doesn’t do hierarchies, but can be useful to get a new organizational structure started. The new groups created will each be assigned the name of one of the documents it contains (I usually change this to a shorter group name, but the original name helps me understand what kind of content the group holds).
Billy, Kate,
First off, sorry for the delay. Many irons in the fire issue.
Thanx for your input. It is helpful.
Bill, I understand what you’re saying. It would be great if Devon would put out a screen cast of what you’re talking about. A picture’s worth a thousand words.