wordbearing

Wordbearing is a proposed metric in linguistics and information retrieval that measures the extent to which a word contributes to the meaning or communicative content of a text in a given context. It aims to capture both how common a word is and how informative it is about the topic or content of the passage. The concept is analogous to information content and to keyword-weighting schemes used in search and summarization.

Measurement approaches vary. A simple formulation pairs word frequency with a contextual distinctiveness score, yielding a wordbearing score W(w) that increases for terms that occur with sufficient frequency within a document while remaining relatively distinctive compared to a reference corpus. Common variants draw on tf-idf, pointwise mutual information, or contextual surprisal. In practice, W(w) can be computed over a sliding window or per document, depending on the analysis goal.

Applications include keyword extraction, document summarization, topic modeling, and search ranking. Wordbearing helps prioritize terms that most accurately reflect a text’s subject matter and reduce the influence of stopwords or overly common terms.

Limitations include dependence on the reference corpus, language, and domain; polysemy and syntactic ambiguity can affect measures of contextual distinctiveness; results vary with window size and pre-processing choices.

See also: information content, tf-idf, keyword extraction, term frequency, mutual information.