领域语言模型及其在中文输入系统中的应用

摘要

The paper presents a strategy of constructing special-domain language model, and a model fusion method based on cosine similarity as well in order to integrate the special-domain language model into dynamic self-adaptive language model. Such language model can be applied to Chinese input system. Experimental results show that the Chinese input system with the special-domain language model loaded can reduce average input code-length, and can also improve hit rate of the first candidate character.%提出一种构建领域语言模型的策略,并基于余弦相似度提出一种模型融合的方法,以便将领域语言模型融入动态自学习语言模型,该语言模型可应用于中文输入系统.实验表明,加载了领域语言模型的中文输入系统不仅降低了平均码长,而且显著提高了首字命中率.