Investigating Lexical Effects in Syntax with Regularized Regression (Lasso)

Within usage-based theory, notably in construction grammar though also elsewhere, the role of lexicon and lexically-specific patterns morphosyntax is well recognized. The methodology, however, not always sufficiently suited to get at details, as lexical effects are difficult study under what currently standard methods for investigating empirically. In this short article, we propose a method from machine learning: regularized regression (Lasso) with k-fold cross-validation, compare its performance Distinctive Collexeme Analysis.