Your problem is two-fold here.
A single bibliography database
Although you can change the built-on DB for anyone you like, Writer references a single DB in any document. Consequently, all entries go into this single DB. And entries are "unified", i.e. they all have the same "value". You can't tag them to tell Writer you want to make a distinction between the document subjects/topics.
A simplistic extraction algorithm
The bibliography may be much richer than what you strictly need for the present document (the BD may be shared between several documents). When you build the table, the document is scanned for bibliography "citations" and marks the corresponding records in the DB. These records are extracted to be processed for the table.
As I mentioned above, you can’t tag entries and you can’t either add a filter in the parameters for
TOC & Index>
TOC, Index or Bibliography. All entries are collected in the table.
Possible workaround (not experimented)
Since the Bibliography feature extend over a document, you could try to split your book as several files provided your overall structure allows it. My understanding of your description is as follows:
- Cover page and introductory material (dedication, thanks, legalese, …
- Chapters made of:
- Chapter Bibliography
If your structure matches this model, you can organise your book as a master document containing everything except chapters and one sub-document per chapter. The subdocument will contain its own bibliography table. Since the algorithm is local to a file, the table will contain only entries found in this one.
Of course, to guarantee formatting consistency, I highly recommend to base all files on the same template and to proscribe absolutely any direct formatting.
My suggestion assumes that your partial bibliography tables are chapter-based. Should they be topics-based (e.g. all related to chemistry, then all for medicine, …), there is unfortunately no solution because of the already mentioned reasons.