CAJ | 학술논문

文本的生成模型是基于文本向量空间表示基础上，针对文本数据的概率表示，针对文本的产生过程进行统计建模的方法。生成模型的优势在于在准确描述表示文本的前提下，达到提高文本挖掘效率的目的。同时，根植于概率统计的生成模型，拥有一整套用于模型学习、推论、模型比较、选择的理论和方法，是其他非统计方法所无法比拟的。本文介绍了用于描述可交换的离散数据集合主要的几种概率生成模型，在向量空间表示的基础上能够捕捉不同程度的词与词之间的上下文（主题）依赖性，因而在英文文本的挖掘上得到了一定程度的应用和肯定。
문본적생성모형시기우문본향량공간표시기출상，침대문본수거적개솔표시，침대문본적산생과정진행통계건모적방법。생성모형적우세재우재준학묘술표시문본적전제하，체도제고문본알굴효솔적목적。동시，근식우개솔통계적생성모형，옹유일정투용우모형학습、추론、모형비교、선택적이론화방법，시기타비통계방법소무법비의적。본문개소료용우묘술가교환적리산수거집합주요적궤충개솔생성모형，재향량공간표시적기출상능구포착불동정도적사여사지간적상하문（주제）의뢰성，인이재영문문본적알굴상득도료일정정도적응용화긍정。
The generation model of the text is a statistical modeling method for the generation process according to the probability representation of text data and on the basis of the text vector space representation. Advantage of generation model is that it can accurately describe the representation text and achieve the purpose of improving the efficiency of text mining. At the same time, the generation model rooted in probability statistics has a set of theories and methods of for model learning, reasoning, model comparison and choice, which is the other non-statistical methods can not match. This article describes several major probability generation models used to describe the exchengeable discrete data collection, which can capture the context (theme) dependence between words based on the vector space representation and has obtain a certain degree of application and affirm in English text mining.