A recent newspaper corpus on Japanese verb phrase adverbials (Namba & Tamaoka, 2014) revealed that the canonical positions for manner and resultative adverbials are SOAdvV for resultative adverbials and SAdvOV and SOAdvV for manner adverbials. This difference of adverbial position is assumed to be derived from the strength of the selectional restriction between an adverbial and a verb. Since the adverbials have different positions, the selectional restriction for each adverbial type should be distinctly unique from each other. The current study investigated this claim by using a newspaper corpus to calculate type and token frequencies of the collocations of adverbial type with verbs and calculated two indexes of entropy and redundancy for each adverbial type. A cluster analysis revealed that while manner adverbials were likely to co-occur with various types of verbs found within a single cluster, resultative adverbials, on the other hand, were likely to occur with specific verb types classified in a separate cluster. From the collocation patterns of adverbials, manner adverbials were shown to have loose selectional restriction whereas resultative adverbials were strict.

View full abstract

Several techniques are discussed here with the idea that linguistic study can be promoted by visualization. Numerical data can be made visible by charts and maps. By this technique, the structure of the data is made easier to read, and new ideas are born. The inner structure of the data can be made clear by applying multivariate analysis technique, and once the structure becomes clear, simpler calculations can be used to indicate the inner structure. There are studies which aim to verify theories and hypotheses, but studies which make the obtained data talk for itself are also necessary. For this, both analysis techniques and representation (visualization) techniques are effective.

View full abstract