非监控类别学习的分类策略与表征

Representation and Strategy in Unsupervised Category Learning

类别学习是人类对不同类别加以归类的过程。类别信息的表征、分类策略运用的特点一直是类别学习研究的重点。非监控类别学习可分为直接的非监控类别学习和间接的非监控类别学习。直接的非监控类别学习(非限制任务,限制任务)中被试的分类策略具有分类"单维度倾向"策略特点,类别变异程度会影响类别表征;间接的非监控类别学习更倾向形成相似性表征,直接的非监控类别学习则为基于规则表征。现有的非监控类别学习的理论对分类策略和表征的解释仍显薄弱,不同学习任务下类别迁移和知识效应的研究还存在不足,未来研究还需要进一步验证知识效应对非监控类别学习的认知加工过程的影响、探索影响类别表征形成的因素等问题。

Categorization is a process in which human beings learn to classify different categories. The representation of categorical information and the implication of classification strategy are hot topics in the field of categorization. Category learning includes both supervised and unsupervised category learning. Whereas previous literature has had a lot of introductions on supervised category learning, this article mainly introduced unsupervised category learning in which how human beings represent information and apply strategy by direct vs. indirect ways. In the direct unsupervised category learning （constrained categorization tasks, unconstrained categorization tasks）, individuals has a tendency to classify the strategy in a ＂one-dimension＂ way. Moreover, both within-category variances and between-category distance can influence category representation. Indirect categories of unsupervised learning are more likely to form a similar representation. By contrast, the representation of direct unsupervised category learning is rule-based. The current theories of unsupervised category learning can＇t completely explain category strategy and representation. The researches on category transfer and knowledge effect in different learning tasks are not yet sufficient. Further studies can test some issues such as the influence of knowledge effect on the cognitive process of unsupervised category learning and explore some factors that influence category representation formation.