基于人工神经网络的微博投诉句识别

Identification of Micro-blog Complaint Sentences through ANN

近年来，微博投诉成为用户维权的新途径。面对海量微博投诉，企业如何在有限资源下实现微博投诉自动诊断和处理，引发了学术和业界的高度关注。微博投诉句识别是微博投诉自动化处理的前提。本文针对微博投诉句识别问题，设计了微博投诉句识别框架，提出了投诉句识别的MCSI模型；应用人工神经网络方法，基于投诉信息特征、语句位置、句式、句长、词频5个属性，识别语句是否为投诉句。与多种基线模型相比，本文提出的模型可以提高投诉句识别性能。

Nowadays, microblogging complaints have become a new way for users to express their rights and dissatisfactions. How to achieve automatic diagnosis and treatment of mieroblogging complaints with limited resource is causing great concern in the academic and practice. Recognition of complaint sentences from microblogging complaint is a prerequisite for automated processing of microblogging complaints and this work designs a Microblogging Complaint Sentence Recognition （MCSR） model for the task. The model uses artificial neural network method, based on 5 features including the complaint information feature, sentence position, syntactical structure, sentence length, and word frequency to identify whether the sentence is a complaint sentence. In comparison with a variety of baseline models, the proposed model can improve the performance of microblogging complaint sentence recognition.