Key terms and questions Terms: amodal , symbolic and linguistic as - PowerPoint PPT Presentation

Key terms and questions ● Terms: – amodal , symbolic and linguistic as antonyms for modal , embodied , and perceptual (De Vega et al., 2008) – usage: symbolic = amodal linguistic , embodied = perceptual Grounded cognition ● Questions: Can computational algorithms Symbol Interdependency Hypothesis 1. extract meaning from language? 2. advance theories of human cognition? Igor Farkaš Centre for Cognitive Science Faculty of Mathematics, Physics and Informatics Comenius University in Bratislava M. Louwerse: Symbol Interdependency in Symbolic and Embodied Cognition. Príprava štúdia matematiky a informatiky na FMFI UK v anglickom jazyku Topics in Cognitive Science, 2010. ITMS: 26140230008 1 2 Claims of the paper Symbolic cognition - LSA ● LSA goes beyond the simple word–context frequency matrix 1. Symbolic cognition account tends to place more emphasis on the algorithm than on linguistic regularities. ● first-order (the same) vs higher-order (similar) co-occurrences 2. Embodied cognition account underestimates the importance of ● many successful applications (better than keyword-based linguistics in general (meaning is modal in nature). methods) 3. Embodied representations are directly mapped onto language ● Similarity estimates derived by LSA depend on a powerful because language encodes embodied relations. mathematical analysis that is capable of correctly inferring much deeper relations. 4. Central claim: „The support for language comprehension and language production is vested neither in the brain of the ● How is meaning extracted? language user, its computational processes, nor in embodied – Landauer: via computational algorithm representations, but outside the user, the process, and the representation, in language itself.“ (Deacon, 1997). – Louwerse: from the language itself – these two options are not mutually exclusive 3 4

Arguments for embodied cognition Symbol interdependency hypothesis ● Simple covariation of amodal symbolic patterns basically is not ● Proposal: language comprehension is both embodied and much more than a symbolic merry-go-round (Harnad, 1990) . symbolic ● A wealth of experimental evidence exists that comprehension ● SIH emphasizes the importance of the language structures, must go beyond symbol manipulation (e.g. Pulvermueller) . without discarding the notion of symbol grounding. ● Nonlinguistic representations are tightly coupled to language ● Prediction: language encodes perceptual information (Zwaan, Spivey,...) ● SIH has implications for both accounts: ● Amodal linguistic symbols must always activate embodied – Symbolic: LSA results can be obtained from language representations whose meshing only constitutes meaning surface structure as well (Glenberg, 1997). – Embodied: results attributed to perceptual simulations can be traced back to language itself. ● shallow (underspecified and incomplete) and deep (specified and complete) language processing (Louwerse & Jeuniaux, 2010) 5 6 SIH relations Evidence for SIH ● based on hierarchy of signs (Deacon, 1997) ← semiotic theories ● Language might in fact not be as arbitrary as the embodied (Peirce, 1923) cognition account suggests (Christiansen & Chater, 2008): ● humans as symbolic species, can make link b/w symbols ● Short words tend to describe objects and events that are (Deacon, 1997) frequent (Zipf, 1935). ● SIH is related to Construction Integration model (Kintsch, 1998) , ● Cross-linguistically, phonological features alone can determine does not emphasize propositional (but surface) structure grammatical category (Monaghan, Chater, and Christiansen, 2005) . ● Similar to Dual Coding theory (Paivio, 1971, 1986) ● Human ability to detect the difference between real and unreal language – different activation in Broca’s area (Musso et al., 2003) – Levels of meaning: representational, referential, associative ● has similarities with perceptual symbols theory (Barsalou, 1999) ● shift from syntagmatic to paradigmatic relations (in development) ● SIH assumption: language encodes perceptual information and ● incidental statistical learning (Saffran, Newport, …) that language users make use of these linguistic cues. ● Even when there are no statistical regularities in incoming ● Evidence? information, humans try to find a pattern. 7 8

Semantic regularities in language Semantic regularities in language (ctd) ● LSA vs first-order methods ● Categorization of semantic knowledge ● Strength of LSA lies in extracting meaning from small texts, ● computational model that simulates (the emergence of) human whereas first-order co-occurrences fail because of data sparsity. categorization of concepts (Rogers and McClelland, 2004, 2008) ● LSA analysis is latent, whereas a word co-occurrence analysis ● Similar to LSA – if origin of features is irrelevant is overt ● LSA categorization analysis: two findings ● If humans are exposed to a large amount of language, then 1. First-order co-occurrence analyses yielded very similar statistical learning humans can rely, in theory, on the surface results as the LSA analyses. structure of language. 2. Perceptual features assigned to verbal descriptors yielded a ● Example might indicate that language (exposure 200–500 mil. grouping of concept categories. word tokens) overtly encodes some of the relations that LSA reveals in a latent analysis. 9 10 Perceptual information in language Discussion ● Argument: Embodied cognition results obtained using linguistic ● According to SIH, meaning can be induced by symbol stimuli should also be considered from a symbolic cognition grounding, as well as by bootstrapping meaning through relations between the symbols themselves. perspective. ● Bootstrapping process is facilitated by language having encoded ● Modality switching embodied information. ● Affordances ● Proposal for an alternative to SGP: „you know some basics of ● Iconic relations the language, then symbolic cognition helps to bootstrap meaning that is obtained through embodied cognition (because ● Geographical information humans are skilled at picking up linguistic regularities that ● Motor resonance (cues for motor affordances) encode embodied relations).“ ● … all encoded in language ● An embodied component should not be abandoned altogether, but neither should a symbolic component. The two are mutually reinforcing. 11 12