Theory and Operational Definitions in Computational Memory Models: A Response to Glenberg and Robertson

Extracting semantic representations from word co-occurrence statistics: A computational study

“…They reject the solution offered for this problem by Landauer and Dumais (1997), of encoding co-occurrence between perceptual events and words or other perceptual events, because this has not yet been implemented in approaches such as HAL or LSA. Burgess (2000), in his reply to Glenberg and Robertson (2000), champions models where meaning is represented as high dimensional vectors derived from word co-occurrence for being explicit and transparent. He reasons that Glenberg & Robertson's experimental data showing that one implementation of LSA cannot account for flexible judgments (such as the plausibility of filling a sweater with leaves as a substitute for a pillow, as against filling a sweater with water) are unfair tests because the LSA vectors had not been derived from relevant "experiences."…”

Modelos y teorías computacionales de la memoriahumana: un estado de la cuestión y análisis crítico

“…La tercera clasificación trata sobre modelos de tipos de memoria o funciones de la memoria específicas (Baddeley, 1986;Burgess, 2000;Burgess y Hitch, 2005;Byrne, 1996;Chartier, Renaud y Boukadoum, 2008;Collins, 1973;Denhière y Lemaire, 2004;Foltz, 1991;Howard, 2009;Leake, 1994;Norman, Detre y Polyn, 2008;O'Reilly, Braver y Cohen, 1999;Rizzuto, 2002;Rogers, 2008;Shastri, 2001). En este tipo de modelo se aisla alguna característica o función de la memoria para ser estudiada aparte de los otros procesos cognitivos.…”

Use of latent semantic analysis for predicting psychological phenomena: Two issues and proposed solutions

“…In addition, we would expect less similarity across students, and we might not expect the level of similarity to predict essay quality. In other words, LSA may not be appropriate for analyses of reasoning phenomena (for related discussions, see Burgess, 2000, andW. Kintsch, 2001), or at least, simple similarity indices may not be.…”

“…In the example to follow, we show how we derived LSA-based indices that were predictive of students' reasoning. LSA, as with people, knows only about words and concepts to which it has been exposed (Burgess, 2000;Landauer & Dumais, 1997). The ability of LSA to match human semantic relatedness judgments is dependent on LSA's having exposure to texts comparable to ones human judgment makers would have had exposure to.…”

“…They used a general knowledge space that may not have had enough "experience" with leaves and water in appropriate contexts to make the distinctions needed to distinguish between the sensibility of the two sentence pairs (Burgess, 2000).…”