Closed gustafl closed 7 years ago
I have now evalulated MongoDB and consider it unsuitable for this application. In short, the Lexeme data model has too many relations, and I believe it's possible to deal with whatever tree modelling problems we encounter within the constraints of a SQL database.
It turns out the simple grammatical category/grammeme model probably isn't enough to describe all the grammatical features we want to include. We need a much more flexible model to make room for the many kinds of relationships that may exist between words. In particular, we need to model multi-level hierarchical data easily. We need to model grammar rules like trees where the same leaf can appear on many branches. If that sounds confusing, it's because grammar rules are this unstructured. MongoDB should be perfect for the job.