摘要
语气词是语气的主要表达方式,能表现说话人的情感和真实意图,识别与分类语气词有助于把握句子的真实含义与情感倾向,从而提升机器翻译、情感分析等自然语言处理任务的性能。语气词用法较为灵活,同一语气词往往可以表示多种语气,使得语气词的自动标识面临着挑战。为此,提出一种融合语句特征的WBA模型对语气词进行识别与分类。首先,使用WoBERT对语料文本进行预训练以获得具有上下文信息的特征向量;其次,在特征向量的基础上融合词性、依存句法关系、语义依存关系等外部特征增强语义信息;最后,使用BiLSTM模型捕捉上下文的语义依赖关系,并使用自注意力机制筛选特征,以进一步提升模型性能。实验表明,该方法在语气词识别与十二分类上取得了较好的效果,宏F1值分别为98.19%、88.74%,相较于不同模型进一步证明了该方法的有效性。
Modal particles are the main way of expressing mood,which can express the speaker's emotions and true intentions.Identifying and classifying modal particles in sentences helps to grasp the true meaning and emotional tendency of the sentence,thereby improving the performance of natural language processing tasks such as machine translation and sentiment analysis.The usage of modal particles is relatively flexible,and the same modal particle can often represent multiple moods,which poses a challenge to the automatic identification of modal particles.Therefore,a WBA model that integrates sentence features is proposed to recognize and classify modal particles.Firstly,WoBERT is used to pre train the corpus text to obtain feature vectors with contextual information;Secondly,external features such as part of speech,dependency syntactic relationships,and semantic dependency relationships are integrated on the basis of feature vectors to enhance semantic information;Finally,the BiLSTM model is used to capture the semantic dependencies of the context,and the self attention mechanism is used to filter features to further improve the model performance.The experiment shows that this method has achieved good results in modal particle recognition and twelve classification,with macro F1 values of 98.19%and 88.74%,respectively.Compared with different models,it further proves the effectiveness of this method.
作者
杨进才
郭清创
吕文杰
沈显君
YANG Jincai;GUO Qingchuang;LYU Wenjie;SHEN Xianjun(School of Computer Science,Central China Normal University,Wuhan 430079,China)
出处
《软件导刊》
2025年第7期135-142,共8页
Software Guide
