Discursive quads: new kinds of lexical co‐occurrence data with linguistic concept modelling

This paper introduces linguistic concept modelling, a new computational approach to humanities-driven analysis of meaning in large text collections, and presents illustrative examples of the approach applied to over one billion words of printed Early Modern English contained in Early English Books Online (Text Creation Partnership edition). Linguistic concept modelling methods and innovations are described in detail, and justified as a unique, new, powerful means for studying meaning in texts in relation to individual lexical items, sets of lexical items and entire text collections. Linguistic concept modelling is compared to established approaches of distributional semantics and topic modelling, to show how linguistic concept modelling overcomes the limitations of those approaches to support humanities research. Examples demonstrate how our approach can be used to explore and analyse semantic, pragmatic and discursive meaning in texts, highlighting the novel observations that our approach affords, and its benefits for humanities scholars.