Subcategorization Frames (WP6. 1) Subcategorization frames (SCFs) define the potential of predicates to choose their argument slots in syntax. Most work on SCF acquisition has focused on verbs, although nouns and adjectives can also subcategorize. A knowledge of SCFs implies the ability to distinguish, given a predicate in raw text and its co-occurring phrases, which of those phrases are arguments (obligatory or optional) and which adjuncts. For example, in the sentence ▇▇▇▇ hit the fence with a stick in the morning, the NP the fence is an obligatory argument, the instrumental PP with a stick is an optional argument, and the PP in the morning is an adjunct. SCFs describe the syntactic, not semantic, behaviour of predicates. Thus ▇▇▇▇▇▇▇'▇ well-known example Colorless green ideas sleep furiously involves a violation of the selectional preferences of sleep but not its SCF, whereas the sentence The parent slept the child violates the SCF of sleep. Access to an accurate and comprehensive SCF lexicon is useful for parsing (▇▇▇▇▇▇▇ and ▇▇▇▇▇▇▇, 1997; ▇▇▇▇▇▇▇, 1997; ▇▇▇▇▇▇▇ et al., 1998; ▇▇▇▇ and ▇▇▇▇▇▇, 2005) as well as other NLP tasks such as Information Extraction (▇▇▇▇▇▇▇▇ et al., 2003) and Machine Translation (▇▇▇▇▇ et al., 2002). SCF induction is also important for other (computational) linguistic tasks such as automatic verb classification, selectional preference acquisition, and psycholinguistic experiments (▇▇▇▇▇▇▇ im ▇▇▇▇▇, 2000; ▇▇▇▇▇▇ et al., 2001; ▇▇▇▇▇▇▇ im ▇▇▇▇▇ and ▇▇▇▇, 2002; ▇▇▇▇▇▇▇▇, 2001; ▇▇▇▇▇▇▇▇ and ▇▇▇▇▇▇▇, 2003; ▇▇▇ et al., 2008a, 2008b). All methods of SCF acquisition share a common objective: given corpus data, to identify (verbal) predicates in this data and record the types of SCFs taken by these predicates, and often
Appears in 2 contracts
Sources: Grant Agreement, Grant Agreement