The ACQDIV Database: Min(d)ing the Ambient Language

The ACQDIV Database: Min(d)ing the Ambient Language Steven Moran, Robert Schikowski, Danica Pajovi´c, Cazim Hysi and Sabine Stoll Department of Comparative Linguistics, University of Zurich Plattenstrasse 54, 8032 Zurich, Switzerland {steven.moran, robert.schikowski, danica.pajovic, cazim.hysi, sabine.stoll}@uzh.ch Abstract One of the most pressing questions in cognitive science remains unanswered: what cognitive mechanisms enable children to learn any of the world’s 7000 or so languages? Much discovery has been made with regard to specific learning mechanisms in specific languages, however, given the remarkable diversity of language structures (Evans and Levinson, 2009; Bickel, 2014) the burning question remains: what are the underlying processes that make language acquisition possible, despite substantial cross-linguistic variation in phonology, morphology, syntax, etc.? To investigate these questions, a comprehensive cross-linguistic database of longitudinal child language acquisition corpora from maximally diverse languages has been built. Keywords: language acquisition, corpus, cross-linguistic, database 1. Introduction In this paper, we present the ACQDIV database com- piled for the European Research Council-funded project: Acquisition processes in maximally diverse languages: min(d)ing the ambient language.1 During the project’s first year, we have compiled together a set of ten electronic cor- pora from very linguistically diverse languages. In doing so, we have had to solve issues including: • Compiling disparately formatted and annotated cor- pora and making them technologically and linguisti- cally interoperable • Creating workflows for extracting, transforming and loading the data into standardized formats and inter- faces for analysis • Producing a single unified database structure that al- lows us to mine patterns in naturally spoken language at the utterance, word and morpheme levels With this unified cross-linguistic resource in place, our project is beginning to investigate questions involving lan- guage acquisition, including: • Are there underlying patterns in the ambient language (i.e. child-directed speech) that children use in acquir- ing language? • If patterns exist, do they verify or refute current theo- ries of child language acquisition (e.g. universalist ap- proaches vs rule-based learning)? • How do children, say between the ages of 2-3, acquire language-specific features that are known to vary so dramatically across languages in form and content?2 1Primary investigator, Prof. Sabine Stoll, Department of Com- parative Linguistics, Psycholinguistics Lab, University of Zurich, Switzerland. ERC award period: 2014-2019. Website: http: //www.acqdiv.uzh.ch/. 2We focus on grammatical negation and aspect strategies. We describe the language sample and how it was selected in Section 2. The sample is not only geographically, cul- turally and linguistically diverse, but the corpora are tech- nologically and theoretically heterogeneous. Therefore we briefly explain our approaches toward interoperability and the resulting database, which to our knowledge, is one of the single largest spoken and unified language corpora to date in Section 3.3 Given the unified ACQDIV database, we are now starting phase two of the project: addressing the pertinent research questions mentioned above. We show some initial data explorations in Section 4. And we present our preliminary research findings in Section 5. 2. The Language Sample A fuzzy clustering algorithm that takes as input a set of lan- guages and their …

Content truncated to 500 words