Where does sense collide?

Gottlob Frege's 1892 paper Über Sinn und Bedeutung, translated in 1948 as Sense and Reference, drew a line between what a name refers to and the way that referent is presented to a mind. The word Frege coined, given in English as the term 'sense', is that mode of presentation: 'the morning star' and 'the evening star' pick out the same planet yet present it differently, so they share a referent while carrying different modes of presentation. Here the referent is an abstract feature of linguistic meaning, a property of expressions studied by logicians, and it has nothing to do with a channel of perception, an entry chosen by a program, or a strand of DNA.[1][2]

Unit

One expression's mode of presenting its referent

Recovery means

Not applicable

Measured by

Analysed by logicians through substitution tests in which co-referring expressions behave differently

Charles Sherrington's 1906 book The Integrative Action of the Nervous System set out a modern classification of the body's channels of perception, grouping the receptors that face the outside world, those that report on the internal organs, and those that track the position of the muscles and joints, a category he named proprioception. In this meaning a 'sense' is a distinct route by which the nervous system gathers information, such as vision, hearing, touch, taste, or smell, each with its own receptors and pathways. The referent is a biological faculty of an organism, measured by thresholds and acuity, and it is unrelated to a mode of presenting a referent in language, a meaning listed for a word, or a coding strand of a gene.[3]

Unit

One channel of perception with its own receptors and pathway

Recovery means

Receptors adapt and recover their responsiveness between stimuli

Measured by

Quantified by detection thresholds, acuity, and just-noticeable differences

Warren Weaver's 1949 memorandum Translation, published in 1955, framed a problem for machine translation: a single written word often carries several unrelated meanings, and only its context reveals which one is intended. The computational tradition that followed came to call each of those distinct meanings a word's 'sense', and David Yarowsky's 1995 paper Unsupervised word sense disambiguation rivaling supervised methods gave an influential way to choose the intended one automatically. Here the referent is a discrete entry in an inventory of meanings that an algorithm selects among, a unit of lexicography and language technology, and it is not a faculty of perception, a philosophical mode of presentation, or a strand of nucleic acid.[4][5]

Unit

One discrete meaning of a word in a defined inventory

Recovery means

Not applicable

Measured by

Scored by disambiguation accuracy against human-annotated corpora