紀要論文 Graph-theoretic and Physical Account of Binary versus Ternary Branching Structure in Human Natural Language Computation
Graph-theoretic and Physical Account of Binary versus Ternary Branching Structure in Human Natural Language Computation

有川, 康二

内容記述
We demonstrate that information loss (error) is minimized in a binary branching network from linear-algebraic (graph-theoretic) and physical perspectives. Nature organizes the network currents to minimize heat loss. Nature has created the human brain, and a set of computational procedures of human natural language (CHL) exists in the human brain. A natural object CHL conforms with the principle of minimal computation (MC) and chooses binary branching structures rather than ternary branching structures. MC prefers efficient (less costly) computation. More information is conserved in the binary branching network. Binary branching structures, and not ternary branching structures, conform with the conservation law. Information means energy, forces, or heat from an engineering perspective. Moreover, we have graphtheoretic (linear-algebraic) and physical reasons for concluding that the syntactic operation merge in CHL must be binary, and not ternary.
本文を読む

https://stars.repo.nii.ac.jp/?action=repository_action_common_download&item_id=613&item_no=1&attribute_id=22&file_no=1

このアイテムのアクセス数:  回

その他の情報