基于BiLSTM模型的定义抽取方法  

Definition Extraction Method Based on BiLSTM Model

作　　者：阳萍 谢志鹏[1] YANG Ping;XIE Zhipeng(School of Computer Science,Fudan University,Shanghai 201203,China)

机构地区：[1]复旦大学计算机科学技术学院,上海201203

出　　处：《计算机工程》2020年第3期40-45,共6页Computer Engineering

基　　金：国家重点研发计划(2018YFC0830900)。

摘　　要：定义抽取是从非结构化文本中自动识别定义句的任务,定义抽取问题可建模为句子中术语及相应定义的序列标注问题,并利用标注结果完成抽取任务。针对传统的定义抽取方法在抽取定义特征过程中费时且容易造成错误传播的不足,提出一个基于双向长短时记忆(BiLSTM)的序列标注神经网络模型,对输入文本进行自动化定义抽取。通过将原始数据输入到BiLSTM神经网络中,完成输入句的特征表示,并采用基于LSTM的解码器进行解码得到标注结果。在Wikipedia英文数据集上的实验结果表明,该方法的精确率、召回率和F1值分别为94.21%、90.10%和92.11%,有效提升了基准模型效果。Definition extraction is the task that automatically identifying definition sentences from unstructured text.The definition extraction problem can be modeled as the sequence labeling problem of a sentence term and its corresponding definition,in which the extraction task can be accomplished by using the labeling results.Aiming at the shortcomings that the traditional definition extraction method can easily cause error propagation while defining features,this paper proposes a sequence labeling neural network model based on Bidirectional Long Short Term Memory(BiLSTM)to automatically execute definition extraction for input text.By inputting the original data into the BiLSTM neural network,this model completes the feature representation of the input sentences.Then,by using the LSTM based decoder for decoding,the labeling results are obtained.Experimental results on the Wikipedia English dataset show that the accuracy,recall and F1 value of the proposed method are 94.21%,90.10%and 92.11%respectively and it can effectively improve the effectiveness of the benchmark model.

关 键 词：定义抽取 双向长短时记忆模型 序列标注 LSTM模型 深度神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]