'arbor est causa proxima fructus'
Documentation
The annotation style of the Index Thomisticus Treebank (and of the portion of the Latin Dependency Treebank made available here) is grounded on Functional Generative Description (FGD), a dependency-based theoretical framework developed in Prague and intensively tested and applied while building the Prague Dependency Treebank of Czech (PDT).
Beside the raw text (the so called 'word-layer': w), the annotation style of the Index Thomisticus treebank features a three-layer structure, ordered as follows:
- a morphological layer (m): lemmatisation and full morphological annotation;
- an 'analytical layer' (a): dependency-based superficial (surface) syntactic annotation;
- a 'tectogrammatical layer' (t): dependency-based annotation of the underlying meaning of the sentence (connnected with the valency lexicon for Latin VALLEX).
The development of each layer requires the availability of the previous one(s). The analytical and the tectogrammatical layers
are represented by dependency tree-graphs, respectively named 'analytical tree structures' (ATS) and 'tectogrammatical tree-structures' (TGTS).
The main features of ATS and TGTS are the following:
- in ATS every word and punctuation mark of the sentence is represented as a node of a rooted dependency tree. The edges of the tree correspond to (surface) syntactic dependency relations, such as Subject, Object, etc.;
- TGTS describe the underlying structure of the sentence, conceived as the semantically relevant counterpart of the grammatical means of expression. The nodes of TGTS represent autosemantic words only (function words are excluded), labeled with semantic role labels (called 'functors') divided into 'inner participants' (i.e. arguments), and 'free modifications' (i.e. adjuncts). The two dimensions of TGTS represent the syntactic structure of the sentence (the vertical dimension) and its information structure ('topic-focus articulation', TFA), based on the underlying word order (the horizontal dimension). In FGD, TFA deals with the opposition between contextual boundness (the 'given' information, on the left) and contextual unboundness (the 'new' information, on the right). Ellipsis and anaphora resolution and coreferential relations are provided.
Please, find below the documentation for each of the annotation layers of the Index Thomisticus Treebank (plus the valency lexicon for Latin VALLEX):
Morphological layer:
- the morphological tagset of the Index Thomisticus Treebank. See the README file for details about the moprhological tags in the CoNLL files of the IT-TB
- the morphological tagset of the Latin Dependency Treebank
Analytical layer:
- PDT guidelines
- Guidelines for the Syntactic annotation of Latin Treebanks
- Guidelines for the treatment of specific constructions -- work in progress --
Tectogrammatical layer:
- PDT guidelines
- Guidelines for the treatment of specific constructions -- work in progress --
- Video of a talk by Marco Passarotti about the treatment of a number of Latin constructions at tectogrammatical layer