A Multi-word-agent Autonomous Learning Model for Regulating Word Combination Strength

Words are basic structural units of language that combine with each other to form sentences. The learning strength of combinative relations between words is of key importance in sentence structure analysis. Inspired by the analogies between words and lymphocytes, a multi-word-agent autonomous learning model based on an artificial immune system is proposed to learn word combination strength. The model is constructed via Cellular Automation, and words are modeled as B cell word agents and as antigen word agents. The language network is then simulated as an immune network. Meanwhile, Spreading Activation is employed to simulate idiotypic interactions between B cells. This research provides a completely new perspective on language and words and introduces biologically inspired processes from immune systems into the proposed model. The most significant advantage of the model is the ability of continuous learning and the concise implementation method. According to the graph-based dependency parsing method, the syntax dependency tree of a sentence can be predicted based on word combination strength in a bottom-up paradigm, from pairs of smaller structures to larger structures. Therefore, the effectiveness of the model can be verified by sentence dependency parsing. The experimental results on the Penn Chinese Treebank 5.1 indicate that our model can effectively and continuously learn word combination strength.