On Learning Subtypes of the Part-Whole Relation: Do Not Mix Your Seeds

Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 1328–1336, Uppsala, Sweden, 11-16 July 2010. c⃝2010 Association for Computational Linguistics On Learning Subtypes of the Part-Whole Relation: Do Not Mix your Seeds Ashwin Ittoo University of Groningen Groningen, The Netherlands r.a.ittoo@rug.nl Gosse Bouma University of Groningen Groningen, The Netherlands g.bouma@rug.nl Abstract An important relation in information ex- traction is the part-whole relation. On- tological studies mention several types of this relation. In this paper, we show that the traditional practice of initializ- ing minimally-supervised algorithms with a single set that mixes seeds of different types fails to capture the wide variety of part-whole patterns and tuples. The re- sults obtained with mixed seeds ultimately converge to one of the part-whole relation types. We also demonstrate that all the different types of part-whole relations can still be discovered, regardless of the type characterized by the initializing seeds. We performed our experiments with a state-of- the-art information extraction algorithm. 1 Introduction A fundamental semantic relation in many dis- ciplines such as linguistics, cognitive science, and conceptual modelling is the part-whole rela- tion, which exists between parts and the wholes they compise (Winston et al., 1987; Gerstl and Pribbenow, 1995). Different types of part-whole relations, classified in various taxonomies, are mentioned in literature (Winston et al., 1987; Odell, 1994; Gerstl and Pribbenow, 1995; Keet and Artale, 2008). The taxonomy of Keet and Ar- tale (2008), for instance, distinguishes part-whole relations based on their transitivity, and on the semantic classes of entities they sub-categorize. Part-whole relations are also crucial for many in- formation extraction (IE) tasks (Girju et al., 2006). Annotated corpora and semantic dictionaries used in IE, such as the ACE corpus1 and WordNet (Fell- baum, 1998), include examples of part-whole re- lations. Also, previous relation extraction work, 1http://projects.ldc.upenn.edu/ace/ such as Berland and Charniak (1999) and Girju et al. (2006), have specifically targeted the discovery of part-whole relations from text. Furthermore, part-whole relations are de-facto benchmarks for evaluating the performance of general relation ex- traction systems (Pantel and Pennacchiotti, 2006; Beamer et al., 2008; Pyysalo et al., 2009). How- ever, these relation extraction efforts have over- looked the ontological distinctions between the different types of part-whole relations. They as- sume the existence of a single relation, subsuming the different part-whole relation types. In this paper, we show that enforcing the onto- logical distinctions between the different types of part-whole relations enable information extraction systems to capture a wider variety of both generic and specialised part-whole lexico-syntactic pat- terns and tuples. Specifically, we address 3 major questions. 1. Is information extraction (IE) harder when learning the individual types of part-whole relations? That is, we determine whether the performance of state-of-the-art IE systems in learning the individual part-whole relation types increases (due to more coherency in the relations’ linguistic realizations) or drops (due to fewer examples), compared to the tra- ditional practice of considering a single part- whole relation. 2. Are the patterns and tuples discovered when focusing on a specific …

Content truncated to 500 words