Structure type configuring

Parameters for structure type configuring

Parameter	Possible values	Default value	Where can be used	Description
document_type	other, law, tz, diploma, article, fintoc	other	dedoc.DedocManager.parse() dedoc.structure_extractors.StructureExtractorComposition.extract()	Type of the document structure according to specific domain. If you use default manager config for DedocManager, then the following options are available: other – structure for document of any domain (Default document structure type). In this case, DefaultStructureExtractor is used. law – Russian laws (Law structure type). In this case, ClassifyingLawStructureExtractor is used. tz – Russian technical specifications (Technical specification structure type). In this case, TzStructureExtractor is used. diploma – Russian thesis (Diploma structure type). In this case, DiplomaStructureExtractor is used. article – scientific article (Article structure type (GROBID)). In this case, ArticleReader and ArticleStructureExtractor are used. fintoc – English, French and Spanish financial prospects (FinTOC structure type). In this case, FintocStructureExtractor is used. If you use your custom configuration, look to the documentation of StructureExtractorComposition
patterns	list of patterns based on AbstractPattern, or list of patterns dicts, or list of dictionaries converted to string	None	dedoc.DedocManager.parse() dedoc.structure_extractors.StructureExtractorComposition.extract() dedoc.structure_extractors.DefaultStructureExtractor.extract()	This parameter is used only by DefaultStructureExtractor (document_type="other"). Configuration of default document structure, please see Configure structure extraction using patterns for more details.
structure_type	tree, linear	tree	dedoc.DedocManager.parse() dedoc.structure_constructors.StructureConstructorComposition.construct()	The type of output document representation. If you use default manager config for DedocManager, then the following options are available: tree – the document is represented as a hierarchical structure where nodes are document lines/paragraphs and child nodes have greater hierarchy level then parents according to the level found by structure extractor. In this case, TreeConstructor is used to construct structure. linear – the document is represented as a tree where the root is empty node, and all document lines are children of the root. In this case, LinearConstructor is used to construct structure. If you use your custom configuration, look to the documentation of StructureConstructorComposition