Identifying Generic Noun Phrases

Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 40–49, Uppsala, Sweden, 11-16 July 2010. c⃝2010 Association for Computational Linguistics Identifying Generic Noun Phrases Nils Reiter and Anette Frank Department of Computational Linguistics Heidelberg University, Germany {reiter,frank}@cl.uni-heidelberg.de Abstract This paper presents a supervised approach for identifying generic noun phrases in context. Generic statements express rule- like knowledge about kinds or events. Therefore, their identification is important for the automatic construction of know- ledge bases. In particular, the distinction between generic and non-generic state- ments is crucial for the correct encoding of generic and instance-level information. Generic expressions have been studied ex- tensively in formal semantics. Building on this work, we explore a corpus-based learning approach for identifying generic NPs, using selections of linguistically mo- tivated features. Our results perform well above the baseline and existing prior work. 1 Introduction Generic expressions come in two basic forms: generic noun phrases and generic sentences. Both express rule-like knowledge, but in different ways. A generic noun phrase is a noun phrase that does not refer to a specific (set of) individual(s), but rather to a kind or class of individuals. Thus, the NP The lion in (1.a)1 is understood as a ref- erence to the class “lion” instead of a specific in- dividual. Generic NPs are not restricted to occur with kind-related predicates as in (1.a). As seen in (1.b), they may equally well be combined with predicates that denote specific actions. In contrast to (1.a), the property defined by the verb phrase in (1.b) may hold of individual lions. (1) a. The lion was the most widespread mam- mal. b. Lions eat up to 30 kg in one sitting. 1All examples are taken from Wikipedia unless stated oth- erwise. Generic sentences are characterising sentences that quantify over situations or events, expressing rule-like knowledge about habitual actions or situ- ations (2.a). This is in contrast with sentences that refer to specific events and individuals, as in (2.b). (2) a. After 1971 [Paul Erd˝os] also took am- phetamines. b. Paul Erd˝os was born [...] on March 26, 1913. The genericity of an expression may arise from the generic (kind-referring, class-denoting) inter- pretation of the NP or the characterising interpre- tation of the sentence predicate. Both sources may concur in a single sentence, as illustrated in Ta- ble 1, where we have cross-classified the exam- ples above according to the genericity of the NP and the sentence. This classification is extremely difficult, be- cause (i) the criteria for generic interpretation are far from being clear-cut and (ii) both sources of genericity may freely interact. S[gen+] S[gen-] NP[gen+] (1.b) (1.a) NP[gen-] (2.a) (2.b) Table 1: Generic NPs and generic sentences The above classification of generic expressions is well established in traditional formal semantics (cf. Krifka et al. (1995))2. As we argue in this paper, these distinctions are relevant for semantic processing in computational linguistics, especially for information extraction and ontology learning and population tasks. With appropriate semantic analysis of generic statements, we …

Content truncated to 500 words