Chinese Sentence Compression Algorithm Based on Deep Analysis of Sentence Hierarchy in Multiple Application Scenarios

摘要

In recent years, people have paid more and more attention to how to obtain high-value information from massive data, and the demand for efficiency and accuracy has also increased. As a basic technology to solve this problem, sentence compression technology can be widely used in machine translation, question answering system, and generating abstracts. Given that the proposed strategy for Chinese sentence compression is less than English sentence compression, the actual value of Chinese-related sentence compression is self-evident. In the current practical context, the problems encountered in sentence compression are as follows: First, the traditional sentence compression methods are mostly based on parallel corpora of "original sentences-compressed sentences". Parallel corpora are usually difficult to build or obtain. The corpus is even more scarce. Second, sentence compression is relatively rigid, and you cannot choose a flexible compression strategy based on the application. Based on the above practical background, this paper proposes a Chinese sentence compression algorithm that does not rely on parallel corpus and deep analysis of sentence hierarchy in multiple application scenarios. The three compression methods are compared: manual compression, only compression using rules, and compression based on deep analysis of sentence hierarchies. The experimental results show that the Chinese sentence compression algorithm based on deep analysis of sentence hierarchies can be used to some extent according to different application scenarios, reducing or increasing the compression ratio flexibly and appropriately within the acceptable range to obtain a better / slightly lower amount of information and grammatical retention. The compression effect under the comprehensive index is significantly better than the previous two compression methods, and can bring more flexible and realistic compression effects according to the actual application scenario.