An annotated English child language database

The use of large-scale naturalistic data has been opening up new investigative possibilities for language acquisition studies, providing a basis for empirical predictions and for evaluations of alternative acquisition hypotheses. One widely used resource is CHILDES (MacWhinney, 1995) with transcriptions for over 25 languages of interactions involving children, with the English corpora available in raw, part-of-speech tagged, lemmatized and parsed formats (Sagae et al., 2010; Buttery and Korhonen, 2005). With a recent increase in the availability of lexical and psycholinguistic resources and robust natural language processing tools, it is now possible to further enrich childlanguage corpora with additional sources of information. In this paper we describe the English CHILDES Verb Database (ECVD), which extends the original lexical and syntactic annotation of verbs in CHILDES with information about frequency, grammatical relations, semantic classes, and other psycholinguistic and statistical information. In addition, these corpora are organized in a searchable database that allows the retrieval of data according to complex queries that combine different sources of information. This database is also modular and can be straightforwardly extended with additional annotation levels. In what follows, we discuss the tools and resources used for the annotation (§2), and conclude with a discussion of the implications of this initial work along with directions for future research (§3).