Cognition depends on knowledge. High-level cognition, in particular, as in text and discourse comprehension and problem solving, is shaped in every respect by the nature and limitations of its knowledge base. Hence, models of high-level cognition must include a workable model of human knowledge. One could argue that such a model is a necessary prerequisite to the study of higher level cognition. However, we do not have such a model today-a model of human knowledge representation that does justice to the vast scale of human experiences and their intricate structure-and cognitive scientists cannot afford to wait with their work for such a model. Thus, cognitive scientists interested in building models of cognition have been forced to take shortcuts, supplying their models of cognitive processes with hand-coded, on-demand knowledge structures. This has been a general practice in cognitive psychology. For instance, models of text comprehension (including Kintsch, 1974, 1998) and models of analogy making (e.g., Forbus, Gentner, and Law, 1995; Holyoak and Thagard, 1989) have relied on hand-coded propositional representations and knowledge structures. This was a defensible practice because the alternative was not to attempt to model cognition at all. Moreover, the use of ad hoc knowledge structures in these models yielded many valuable insights about comprehension and analogy making, as it did in all of the other areas where such shortcuts were employed. But today, we are moving beyond hand-coding knowledge and toward actually modeling it. Among the first attempts to do so has been LSA. We judge it to have been highly successful at this point. We judge this success on the basis of its contribution to cognitive theory, as well as to real-world applications. But we also believe that LSAis just the beginning of a widespread surge of models of human knowledge that will revolutionalize the study of higher level cognitive processes in the coming decades. We have attempted to communicate in this book what LSAhas achieved already, but the goal was also to foreshadow the exciting possibilities for LSA-like systems in the future.
|Original language||English (US)|
|Title of host publication||Handbook of Latent Semantic Analysis|
|Publisher||Taylor and Francis|
|Number of pages||13|
|State||Published - Jan 1 2007|
ASJC Scopus subject areas