Quick Q: Are item names part of AI?


when the AI is doing the magic of the AI, does it take the title of a document/rtf/txt/entry into consideration ?

Reason for asking is because a large amount of my items in DevonThink Pro Office is empty documents (the title is the information itself). When using “see also” there are sometimes very obvious connections between subjects that are not being made, and it seems to me as “entry title” is not part of the AI system?

Thanks in advance for any insight!

Right. Only the content of documents is used by the AI routines.

That seems to me a good idea. Ever notice how little the headlines have to do with the actual story content in some newspapers and tabloids? Creative headline writers can sell more papers, but at the expense of the trustworthiness of the publication.

Some users have asked that the AI routines give special weight to the Names of documents, so as to change them from their present functions as analysts of the contextual use of terms in documents, to instruments for classification of documents.

I can see potential utility in that, but at the expense of actually decreased functionality of those AI routines in many databases. Some users, for example, use structured but very cryptic naming routines. Quite a few DTPO users don’t bother to rename scanned documents but let them retain the sequence title provided by the scanning driver – in such cases, it’s the content that’s important and the Name has no relationship whatever to the content.

So I would suggest that if one wishes to use Name as an important element for classifying or tagging documents, that can easily be done by using the Search tool to assist in moving search results into an appropriate classification, or identifying relevant items by searching by Name. That sort of approach can be even more powerful in DEVONthink 2, because of the enhanced query operators and syntax.

Thanks Bill for explaining the reasoning behind the names not being used for AI. However, I have lots of images of historical documents that cannot be OCRed. Auto-Classify is useless for these.

Oh well, can’t fix one thing without breaking another!