This topic came up in this forum about a decade ago, and the answer was then that English was the only language in which DT’s AI could work effectively. The world has become increasingly broader since and more people are working with multilanguage databases. I wonder if there is a change in DT’s interests that would lead us to hope that there would be a multilanguage component in the future.
AI works fine for me in other languages. Have you encountered a specific problem with any language?
FROBGOBLIN is correct. The AI routines for analyzing contextual relationships work in any language.
Some users have asked whether its possible to do true semantic analysis for comparisons in databases that contain documents in multiple languages. It would be wonderful if DEVONthink could recognize the similarities of documents about a topic written in English, French, German, Romanian, etc. That would require comparisons of the meanings of terms and context across multiple languages, despite differences in vocabulary and syntax. No, DEVONthink can’t do that. That would be a really tall order. It would go beyond having very large dictionaries (including recognition of the jargon used in different disciplins) to do machine translations, and would require true semantic analysis to work well. Companies like Google and agencies like NSA are throwing large resources at that kind of problem. One of these days, it will become possible. Maybe.
In our Support work we often receive queries in languages other than English. I confess that I often resort to computer translation to help me understand the user’s problem. I’ve yet to see a translation program that doesn’t often require guessing the intended meaning. In many cases, only my familiarity with common problems in use of DEVONthink allows me to tackle the issue. Even the differences in localized menu terms and error codes for English, French and German users sometimes have me scratching my head. That’s because localization tries to use conventions for computer menu terms that are common in each language, and those terms often translate differently when put to computer translation.
Sometimes a computer translation of a German Support ticket might indicate that the user is talking about a nose. No, that’s not what he meant.
Dear FROBGOBLIN and Bill_DeVille
Thanks for your responses. My AI is not yet working at its best, perhaps because I still need to classify a large number of PDFs (I am moving slowly), so I can’t judge well if DT is doing well with a multilanguage database. These files are mostly in English, but about a third are in Spanish, Portuguese, French and German.
The problem I am finding is that a simple word search in a specific language is not crosslisting entries from sources in other languages. For example, if I am looking for documents on the topic of immigration, a simple word search will only yield documents in English. If I search for the topic with a Spanish word (inmigración) it will only yield Spanish PDFs. In other words, it will not read across languages (which is part of Bill_DeVille’s post subject).
For now, it means that I will have to perform individual searches for each language.
DEVONthink’s analysis of contextual relationships of terms in documents does not result in recognition of similarities of expression of concepts across different languages. That would be true semantic analysis. Computers and information science haven’t achieved that level, which approaches understanding text content.
DEVONthink has no training in the scientific disciplines of papers included in my document collections, and no understanding of their meaning. But DEVONthink partners with me to add capabilities that I don’t have, such as a See Also list of documents that may be contextually similar to the one I’m viewing, based on comparison of the content of that document to all the others in a database – produced in a second or so. It’s up to me to review those suggestions and decide whether they are useful.
Not always, but often enough, a See Also suggestion helps me extend my understanding of a topic. That makes it a valuable research tool.
I do draft writing within DEVONthink. I can select a block of text I’ve written, Control-click on the selection and choose See Related Text. DEVONthink will produce a list of suggestions of similar documents in my database and that list may be useful to compare my treatment of a topic with others.
Thanks, Bill_DeVille. It makes sense.