基于融合特征与语法规则的流式文档理解方法

Reflowable document comprehension method based on fusion features and grammar rules

针对流式文档结构理解中构件识别特征分析的不足,提出一种基于融合特征的构件识别方法。首先建立格式向量表示字体等构件格式特征,提取文档构件中关键字等内容特征作为内容向量,分别计算待识别构件两种特征与候选构件的得分并对其加权计算,得出候选的构件标签;结合自顶向下和自底向上的结构识别方法,得到文档的逻辑结构。通过实验验证了该方法能有效提高文档构件识别的准确率,同时提高了文档结构识别的准确率。

In order to achieve adequate component identification in reflowable document structure understanding, a new method to understand documents is proposed based on fusion features and grammatical rules. Two vectors are used in the method. One is the format vector representing the format features, such as fonts;the other is the content vector representing text features such as keywords. Then the components to be identified are compared with the candidates by measuring the distance between the vectors with different weights. Finally, based on the candidate labels and grammatical rules, the logic structure of the document is recognized by applying the top-down and bottom-up algorithm. The experiment results show that this method can effectively improve the accuracy of component identification, and in turn improve the accuracy of whole document structure recognition.