Classify always selects the same group as first choice

When using ‘See Also & Classify’, the Classify panel always selects the same group as the top choice, no matter what document I ask DT to Classify. The rest of the options are usually relevant, so I still use the feature, but can anyone tell me why this might be happening?

[size=85]Edit: comment removed - wrong advice[/size]

It’s not the See Also bit that I’m interested in; it’s the Groups listed in the top panel, where you choose a group to move the selected document to. If, for example, I select a document in my database’s inbox and ask DT to ‘See Also & Classify’, it always gives one particular group as the top choice.

Also, when I browse through the See Also list of suggested documents, the starting point sometimes remains at the top, and sometimes it changes to whatever I’ve clicked on in the See Also list (and then the Classify groups also change). It seems inconsistent.

One more thing: can anyone tell me if browsing through the See Also list changes documents read/unread status? I hope it doesn’t.

The larger your database, in terms of number of documents, words, and number of groups, the better the AI gets at proposing a classify group. Also, the AI will only suggest groups contained in the database, so all indexed groups are excluded from the classify pane. I too have 1-2 groups that almost always appear at the top, but when they are proposed, they would be a valid group to classify to.

I never see this, unless I open a document from the See Also list in a new tab or did a Command-R to reveal the document in the database.

The document will be marked as read if you actually select and view it from the See Also list. The status does not change just because it appears in the list-it must be ‘read’ by viewing it to be marked as such.

Thanks Greg, I still think something odd is going on as the AI choices for the subsequent groups (after the first one) seem to work well (c.200 groups, none indexed, and c.1000 documents), and the inconsistent See Also behaviour is also strange.

I wish the read/unread status didn’t change every time I glanced at something, e.g. for the purposes of checking its relevance (especially in something like See Also, where the status is not displayed). I’ll post on the requests & suggestions board.

Personally, I think this is the enhancement that I would like to see. I’m OK with a document being marked as read when I read it, but it would be nice to have a visual cue that it was originally unread, and as you have observed the See Also pane does not provide that feedback.

I get this a lot. Very irritating. The “see also list” will totally just change even though I’ve not used command+r or opened a document in a tab.

I’m perversely glad not to be the only one…

I used to get this (ie, a single group appearing when I select “See Auto & Classify” with almost any document) in one of my libraries. It was a library of around 2000 PDFs, most of which were 3-4 pages (text) but which also contained about 30 much larger review articles, up to 200-300 pages. One of these large PDFs was responsible: the group containing it was the one always at the top.

My solution: selected “exclude from Classify” from the info panel for the large PDFs. Perhaps you could try that.

I tried it, and it works. Thanks, acl.

It’s likely that one or more of your large PDF documents contained a great many words and dealt with a number of topics – so that it/they “dominated” the decisions by Classify.

For example, if I include the large PDF owner’s manual for my Acura RDX vehicle in a database, it can “confuse” Classify because it deals with such a wide range of topics and terms, from cold weather to flat tires to GPS navigation. :slight_smile:

That suggestion to use the option to exclude a particular document from classification was a good one. If you wish, however, to give particular weight to some of the information in the excluded document for classification purposes, that can be done by creating an Annotation note for it, then briefly excerpting or summarizing the “interesting” information into that note and filing it in the same group as the PDF document. In this case, the content of the Annotation note might be just “Acura RDX owner’s manual.” (Remember that Classify only looks at the content of a document, not the Name or other metadata.)

Now, if I use Classify to suggest the appropriate group for a new document about the maintenance schedule for my vehicle, Classify will probably suggest the same group that contains that owner’s manual. But it will probably suggest a different (and more appropriate) group location for a new document about cold winters in Siberia!

Good stuff, thanks for the tips Bill.