PST 2.0 – Corpus of Polish Spatial Texts

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2167–2174 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2167 PST 2.0 – Corpus of Polish Spatial Texts Michał Marci´nczuk, Marcin Oleksy, Jan Wieczorek Department of Computational Intelligence Wrocław University of Science and Technology Wrocław, Poland {michal.marcinczuk, marcin.oleksy, jan.wieczorek}@pwr.edu.pl Abstract In the paper, we focus on modeling spatial expressions in texts. We present the guidelines used to annotate the PST 2.0 (Corpus of Polish Spatial Texts) – a corpus designed for training and testing the tools for spatial expression recognition. The corpus contains a set of texts gathered from texts collected from travel blogs available under Creative Commons license. We have defined our guidelines based on three existing specifications for English (SpatialML, SpatialRole Labelling from SemEval-2013 Task 3 and ISO-Space1.4 from SpaceEval 2014). We briefly present the existing specifications and discuss what modifications have been made to adapt the guidelines to the characteristics of the Polish language. We also describe the process of data collection and manual annotation, including inter-annotator agreement calculation and corpus statistics. In the end, we present detailed statistics of the PST 2.0 corpus, which include the number of components, relations, expressions, and the most common values of spatial indicators, motion indicators, path indicators, distances, directions, and regions. Keywords: information extraction, spatial expressions, data resources, annotated corpus, Polish 1. Introduction Spatial information refers to a physical location of an ob- ject, which can be encoded using some absolute values in a coordinate system or by relative references to other entities in the space – spatial relations. The spatial relations can be expressed directly by spatial expressions (Kolomiyets et al., 2013) or indirectly by a chain of semantic relations (LDC, 2008). A comprehensive recognition of spatial re- lations between objects in a text requires a manually anno- tated corpus, which may be used as a training and testing data. Recognition and interpretation of spatial expressions is crucial to understand and reason about spatial relations between object in scene description (Chang et al., 2015) or dialogue systems (Williams et al., 2016; Marge and Rud- nicky, 2019). In the paper, we focus on modeling spatial expressions in texts for Polish. We present the guidelines used to anno- tate the PST 2.0 – corpus of Polish Spatial Texts. In Sec- tion 2, we present existing spatial annotation schemes for English (SpatialML, SpatialRole Labelling from SemEval- 2013 Task 3 and ISO-Space1.4 from SpaceEval 2014). In Section 3 we discuss what modifications have been made to adapt the guidelines to the characteristics of the Polish language and we define basic componentes and relations used to model spatial expressions. Section 4 is an overview of the Polish Spatial Texts corpus, including the descrip- tion of data source, manual annotation process and corpus statistics. 2. Existing Spatial Specifications The specification language for spatial information in lan- guage evolves over the years. There are several schemes, which more or less correspond to each other. We based our approach …

Content truncated to 500 words