{"pk":29439,"title":"Reconstructing Maps from Text","subtitle":null,"abstract":"Previous research has demonstrated that Distributional\nSemantic Models (DSMs) are capable of reconstructing maps\nfrom news corpora (Louwerse & Zwaan, 2009) and novels\n(Louwerse & Benesh, 2012). The capacity for reproducing\nmaps is surprising since DSMs notoriously lack perceptual\ngrounding (De Vega et al., 2012). In this paper we investigate\nthe statistical sources required in language to infer maps, and\nresulting constraints placed on mechanisms of semantic\nrepresentation. Study 1 brings word co-occurrence under\nexperimental control to demonstrate that direct co-occurrence\nin language is necessary for traditional DSMs to successfully\nreproduce maps. Study 2 presents an instance-based DSM that\nis capable of reconstructing maps independent of the frequency\nof co-occurrence of city names.","language":"eng","license":{"name":"","short_name":"","text":null,"url":""},"keywords":[{"word":"semantic memory; spatial cognition; embodiment"}],"section":"Language and Meaning","is_remote":true,"remote_url":"https://escholarship.org/uc/item/1xd9145s","frozenauthors":[{"first_name":"Johnathan","middle_name":"E.","last_name":"Avery","name_suffix":"","institution":"Indiana University","department":""}],"date_submitted":null,"date_accepted":null,"date_published":"2020-01-01T10:00:00-08:00","render_galley":null,"galleys":[{"label":"PDF","type":"pdf","path":"https://journalpub.escholarship.org/cognitivesciencesociety/article/29439/galley/19299/download/"}]}