@inproceedings{qi-etal-2022-ji,
    title = "基于多源知识融合的领域情感词典表示学习研究(Domain Sentiment Lexicon Representation Learning Based on Multi-source Knowledge Fusion)",
    author = "Qi, Ruihua  and
      Wei, Jia  and
      Shao, Zhen  and
      Guo, Xu  and
      Chen, Heng",
    booktitle = "Proceedings of the 21st Chinese National Conference on Computational Linguistics",
    month = oct,
    year = "2022",
    address = "Nanchang, China",
    publisher = "Chinese Information Processing Society of China",
    url = "https://aclanthology.org/2022.ccl-1.61",
    pages = "684--693",
    abstract = "{``}本文旨在解决领域情感词典构建任务中标注数据资源相对匮乏以及情感语义表示不充分问题,通过多源数据领域差异计算联合权重,融合先验情感知识和Fasttext词向量表示学习,将情感语义知识映射到新的词向量空间,从无标注数据中自动构建适应大数据多领域和多语言环境的领域情感词典。在中英文多领域公开数据集上的对比实验表明,与情感词典方法和预训练词向量方法相比,本文提出的多源知识融合的领域情感词典表示学习方法在实验数据集上的分类正确率均有明显提升,并在多种算法、多语言、多领域和多数据集上具有较好的鲁棒性。本文还通过消融实验验证了所提出模型的各个模块在提升情感分类效果中的作用。{''}",
    language = "Chinese",
}