Anaphora is one of the most frequently encountered phenomena in natural languages, and its resolution is important to understand sentences and discourses. Anaphora resolution has been one of the hottest topics studied not only in Linguistics but also other related fields such as Philosophy, Artificial Intelligence, Machine Translation, and so on. The goal of this dissertation is to provide anaphora resolution algorithms in which categories play crucial roles. In those algorithms, category combinatorics of constituents triggers resolution algorithms, and these ideas will be implemented with specific computational algorithms and flowcharts. In those analyses and implementations, all of the anaphoric relations will be represented by λ-expressions, and all the anaphoric expressions will be resolved ultimately by λ-conversions.; As theoretical bases, this dissertation takes Categorial Grammar and Chierchia's Binding Theory. Steedman's Combinatory Categorial Grammar (CCG) will be slightly modified so that Chierchia's ideas can be implemented easily and effectively, which will be called a CCG-like system. Two different kinds of anaphora resolution algorithms will be developed. One is English Anaphora Resolution Algorithms, which will be abbreviated an EARA system. The other is Korean Anaphora Resolution Algorithms, which will be named a KARA system. In both systems, the relevant anaphoric properties are instantiated as features of lexical items. These features are percolated up until the anaphoric expressions meet suitable antecedents. When anaphoric expressions meet antecedents, the features are interpreted with those of the antecedents.; As an implementational basis, this dissertation employs a bottom-up-style Shift-Reduce parsing technology. All the operations on categories are included in the Reduce action. All the anaphora resolution algorithms are also performed in this action. In those implementations, we will have no parse tree, and it will speed up the processing time reducing computational complexity considerably.; In sum, this dissertation develops faster and more efficient computational algorithms for anaphora resolution, in which category combinatorics determines the actions. In those algorithms, various kinds of anaphoric expressions are resolved similarly. Through the analyses and their implementations, I hope this research can give us more opportunity to understand characteristics of anaphoric phenomena, which is one of the important phenomena that natural languages demonstrate.
展开▼
