I need to create special morphological vocabulary for a natural language. Each word should contain a set of characteristics. Does PostgreSQL 9.* help in such situation? I mean: should I create table from the scratch or there are some predefined means?
You are looking for things well beyond what typical solutions on PostgreSQL are designed to build. Your best bet is to build your own custom table.
Now, I am just an interested reader of linguistic topics, but for morphological analysis, I suspect you are going to run into problems with different languages having fundamentally different morphological systems. For example, the morphological concepts in Austronesian, Indo-European, and Athabascan (all of which I know just enough to be dangerous) do not strike me as very conducive to a single relational model. For example, we might have distinct morphologies for verb tenses in IE, but not the other two. Reduplication in Austronesian languages adds a bit of a curve ball (particularly when dealing with prefix/suffix/infix combinations), and Athabascan has slots that don't fit the other two language's expectations of parts of speech.
So I don't think you are likely to find many general-purpose morph database schemas around. Things are likely to be language-specific and purpose-specific. For example, building a database to look for morphological changes between Middle and Modern English is going to be quite different than something designed merely to analyse current morphology in, say, tweets.