Verbvaluizing

Verbvaluizing is a term used in corpus linguistics and natural language processing to describe the systematic assignment of structured values to verbs. The practice seeks to capture both the syntactic behavior of a verb—its valency and subcategorization frames—and its semantic properties, such as thematic roles, plus, in some schemes, evaluative or affective dimensions. The result is a feature vector for each verb that can be used in computational models and cross-linguistic comparisons. It is a neologism with no single standard definition, and different researchers may emphasize different components.

Implementation typically involves annotation schemas that map verbs to a finite set of categories: syntactic frames (e.g., transitive, ditransitive, intransitive), argument structure labels (agent, patient, recipient), semantic roles (Agent, Theme, Goal), and optional polarity or intensity scores. Tools such as FrameNet, PropBank, or VerbNet-inspired schemes are often leveraged or extended. The process requires clear guidelines and inter-annotator agreement to ensure reliability.

Applications include improving parsing in NLP, aiding machine translation, enhancing lexical resources, and supporting sentiment or stance analysis for event detection. Challenges include subjectivity in assigning evaluative values, cross-linguistic variation in valency, data sparsity for less common verbs, and the lack of consensus on the exact feature set.

Example: In a verbvaluizing resource, the verb give might be annotated with frames: syntactic (Agent + Recipient + Theme), semantic roles (Agent, Recipient, Theme), and a positive polarity score for contexts where the transfer of possession is favorable; while the verb run might be annotated with syntactic (Subject) and semantic roles (Agent, Theme) with neutral polarity. Such annotations enable models to reason about argument structure and valence across languages.