Czech Legal Text Treebank 2.0

Czech Legal Text Treebank 2.0 Vincent Kr´ıˇz and Barbora Hladk´a Charles University Faculty of Mathematics and Physics Prague, Czech Republic {kriz, hladka}@ufal.mff.cuni.cz Abstract The Czech Legal Text Treebank 2.0 (CLTT 2.0) contains texts that come from the legal domain and are manually syntactically annotated. The syntactic annotation in CLTT 2.0 is more elaborate than in CLTT 1.0. In addition, CLTT 2.0 contains two new annotation layers, namely the layer of entities and the layer of semantic entity relations. In total, CLTT 2.0 consists of two legal documents, 1,121 sentences and 40,950 tokens. Keywords: treebank, legal, long sentences, syntactic annotations, named entities, semantic relations 1. Introduction We have been developing approaches and systems for de- tecting and extracting semantic relations from unstructured texts. We have developed the RExtractor system (Kr´ıˇz et al., 2014; Kr´ıˇz and Hladk´a, 2015). This system implements an extraction pipeline which processes input texts by linguistically-aware tools and extracts entities and relations using queries over dependency trees. The language used for testing RExtractor is Czech and the legal domain was chosen to be explored in detail. We surveyed existing syntactically annotated corpora and only a few of them contain texts from the legal domain, e.g., the Universal Dependencies v2.1 To have a gold- standard data for the RExtractor evaluation, we created the Czech Legal Text Treebank 1.0 (Kr´ıˇz et al., 2016). In total, 1,121 sentences from the Collection of Laws of the Czech Republic were annotated morphologically and syntactically in accordance with the Prague Dependency Treebank annotation framework. In this paper, we introduce the next version of CLTT with more elaborate syntactic annotations and enriched with two annotation layers. The remainder of this paper is organized as follows: Section 2. presents a brief description of CLTT 2.0. Modifications in the syntactic annotation are described in Section 3. Section 4. describes the layer of entities and Section 5. presents the layer of semantic relations. Finally, Section 6. provides more details about getting CLTT 2.0. 2. Czech Legal Text Treebank 2.0 We provide basic characteristics of CLTT 2.0 with a spe- cial attention paid to the differences between CLTT 1.0 and CLTT 2.0. 2.1. Annotation Layers Both CLTT 1.0 and CLTT 2.0 annotation principles fit the framework originally formulated in the Prague Dependency Treebank project (PDT, (Hajiˇc et al., 2018)).2 According to 1http://universaldependencies.org/ 2http://hdl.handle.net/11234/1-2621 this annotation framework, dependency trees are annotated on the three layers: • Word Layer (w-layer) A text is segmented into documents and paragraphs and individual tokens are recognized and associated with unique identifiers. • Morphological layer (m-layer) A sequence of tokens of the word layer is divided into sentences. Annotation of a sentence consists of attach- ing several attributes to the tokens of the w-layer, the most important ones are morphological lemma and tag. • Analytical layer (a-layer) A sentence is represented as a rooted ordered tree with labeled nodes and edges. One token from the morpho- logical layer is represented by exactly one node in the tree and the dependency relation between two nodes is …

Content truncated to 500 words