SpRL-CWW: Spatial Relation Classification with Independent Multi-class Models

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 895–901, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics SpRL-CWW: Spatial Relation Classification with Independent Multi-class Models Eric Nichols Fadi Botros Honda Research Institute Japan Co., Ltd. University of Calgary e.nichols@jp.honda-ri.com fnmbotro@ucalgary.ca Abstract In this paper we describe the SpRL-CWW en- try into SemEval 2015: Task 8 SpaceEval. It detects spatial and motion relations as de- fined by the ISO-Space specifications in two phases: (1) it detects spatial elements and spa- tial/motion signals with a Conditional Ran- dom Field model that uses a combination of distributed word representations and lexico- syntactic features; (2) given relation candidate tuples, it simultaneously detects relation types and labels the spatial roles of participating ele- ments by using a combination of syntactic and semantic features in independent multi-class classification models for each relation type. In evaluation on the shared task data, our system performed particularly well on detection of el- ements and relations in unannotated data. 1 Introduction Understanding human language about location and motion is important for many applications includ- ing robotics, navigation systems, and wearable com- puting. Shared tasks dedicated to the problem of representing and detecting spatial and motion rela- tions have been organized for SemEval 2012 (Ko- rdjamshidi et al., 2012), 2013 (Kolomiyets et al., 2013), and 2015. In this paper we present SpRL-CWW, our entry to SemEval 2015 Task 8: SpaceEval, and present extended evaluation of our system to inves- tigate the impact of the task annotations and system configurations on task performance. 2 SpaceEval Task Definition Kordjamshidi et al. (2011) proposed the task of Spatial Role Labeling (SpRL) to detect spatial and motion relations in text. SpRL was modeled after semantic role labeling (see (Fillmore et al., 2003; Màrquez et al., 2008)), with spatial indicators in- stead of predicates signaling the presence of rela- tions, and spatial roles instead of semantic roles. A canonical example of a spatial relation from (Kordjamshidi et al., 2011) is: (1) Give me the [grey book]TR [on]SP the [large table]LM. The spatial indicator (SP) on indicates that there is a spatial relation between the trajector (TR; pri- mary object of spatial focus) and the landmark (LM; secondary object of spatial focus). SpRL was formalized as a task of classifying tuples of < wSP , wTR, wLM > as spatial relations or not. The SpRL task was reformulated and reintro- duced in SpaceEval1 using the ISO-Space annota- tion specifications (Pustejovsky et al., 2012). The biggest change was the decoupling of the semantic type and role of spatial relation arguments. A taxon- omy of Spatial Element (SE) types was introduced to describe the meaning of arguments independent of their participation in relations, and spatial roles were treated as instance-specific annotations on spa- tial and motion relations. The SE types introduced are: SPATIAL_ENTITY, PATH, PLACE, MOTION, NON_MOTION_EVENT, and MEASURE. Two types were also introduced to rep- resent expressions that indicated the presence of re- lations: SPATIAL_SIGNAL and MOTION. Spatial and motion relations were …

Content truncated to 500 words