Winograd schema challenge

The Winograd schema challenge (WSC) is a test of machine intelligence proposed in 2012 by <a href="/facts/Hector_Levesque/xXYRyBLg">Hector Levesque</a>, a computer scientist at the <a href="/facts/University_of_Toronto/FpGjQNm9">University of Toronto</a>. Designed to be an improvement on the <a href="/facts/Turing_test/wEvf9DIr">Turing test</a>, it is a multiple-choice test that employs questions of a very specific structure: they are instances of what are called Winograd schemas, named after <a href="/facts/Terry_Winograd/RsMmt7R9">Terry Winograd</a>, professor of computer science at <a href="/facts/Stanford_University/JEQeF9z2">Stanford University</a>.
On the surface, Winograd schema questions simply require the resolution of <a href="/facts/Anaphora_(linguistics)/gm1WHVBF">anaphora</a>: the machine must identify the <a href="/facts/Antecedent_(grammar)/FdtdTPdG">antecedent</a> of an ambiguous <a href="/facts/Pronoun/UsbSNRSl">pronoun</a> in a statement. This makes it a task of <a href="/facts/Natural_language_processing/1hjMKsSN">natural language processing</a>, but Levesque argues that for Winograd schemas, the task requires the use of knowledge and <a href="/facts/Commonsense_reasoning/HfL7zQe4">commonsense reasoning</a>.
The challenge is considered defeated in 2019 since a number of <a href="/facts/Transformer_(machine_learning_model)/cDbjx6a8">transformer</a>-based <a href="/facts/Language_model/FntSpg0j">language models</a> achieved accuracies of over 90%.

Winograd schema challenge open-in-new

Winograd schema challenge