| 赵又霖,林怡妮,石燕青.融合情感语义与句法结构的中文开放域事理图谱构建研究[J].数字图书馆论坛,2024,20(3):12~24 |
| 融合情感语义与句法结构的中文开放域事理图谱构建研究 |
| Construction of Chinese Open Domain Event Graph Integrating Sentiment Semantics and Syntactic Structure |
| 投稿时间：2023-12-18 |
| DOI：10.3772/j.issn.1673-2286.2024.03.002 |
| 中文关键词: 开放域；事理图谱；依存句法分析；语义依存分析；情感分析 |
| 英文关键词: Open Domain; Event Graph; Dependency Parsing; Semantic Dependency Parsing; Sentiment Analysis |
| 基金项目:本研究得到江苏省社会科学基金青年项目“社会感知数据驱动下的公共卫生事件时空演化研判机制研究”（编号：20TQC001）、中国博
士后科学基金特别资助“面向应急管理的时空数据语义模型构建及创新应用机理研究”（编号：2021T140311）、中国博士后科学基金面
上项目“环境污染突发事件的时空数据挖掘及协同治理机制研究”（编号：2019M650108）资助。 |
| 作者 | 单位 | | 赵又霖 | 河海大学商学院；南京大学信息管理学院 | | 林怡妮 | 河海大学商学院 | | 石燕青 | 南京农业大学信息管理学院 |
|
| 摘要点击次数: 417 |
| 全文下载次数: 842 |
| 中文摘要: |
| 为解决大规模开放域事理图谱构建过程中缺少标注数据以及事件类型未知导致的限定域事理图谱构建方法难以迁移的问题，利用规则匹配方法高效识别开放域文本中包含的多种事件逻辑关系，融合情感语义与句法结构信息分析提高事件抽取准确性，以更好完成事理图谱的构建任务。首先，总结并扩展因果、顺承、条件、转折等多种逻辑关系抽取模板，并基于规则模板、依存句法信息筛选逻辑关系事件句；其次，创新性地引入情感语义分析方法，在句法结构信息的基础上，通过捕获事件及事件间关系的情感语义精准识别事件类型，进而抽取事件论元；再次，计算语义相似度，进行事件融合，构建＜前序事件，事件逻辑关系，后序事件＞三元组，得到事件事理图谱，并进一步进行事件泛化以构建抽象事理图谱；最后，以事件发展较完整的“2022年猴痘事件”为数据源，通过实证分析证明开放域事理图谱构建方法可以实现不同类型事件的识别、事件间逻辑关系的揭露，其有效性、可行性得到验证。研究不仅弥补了现有事理图谱构建理论的不足，也为决策支持、事件发展预测等提供有力的数据支持。 |
| 英文摘要: |
| In the construction process of large-scale open domain event graph, lack of annotation data and unknown event types cause difficulties in the transfer of limited domain event graph construction method. To solve this problem, we utilize rule matching methods to efficiently identify multiple event logical relationships contained in open domain texts, and integrate sentiment semantics and syntactic structure information analysis to improve the accuracy of event extraction, in order to better complete the task of constructing event graphs. Firstly, we summarize and expand various logical relationship extraction templates such as cause and effect, succession, condition, transition, etc., and screen logical relationship event sentences based on rule templates and dependency parsing information. Secondly, we innovatively introduce the sentiment semantic analysis method to accurately identify event types by capturing the sentiment semantics of events and inter-event relations on the basis of syntactic structural infrmation, and then extract event arguments. Then, the semantic similarity is computed for event fusion, and the ＜preceding event, event logical relation, subsequent event＞ ternary is constructed to get the event graph, and further event generalization is performed to construct the abstract event graph. Finally, taking the “2022 Mpox Incident” event data as the data source, empirical analysis proves that the open domain event graph construction method can realize the identification of different types of events and reveal the logical relationships between events. Its effectiveness and feasibility are verified. The construction of the Chinese open domain event graph not only fills the gaps in the existing theories of event graph construction, but also provides powerful data support for decision-making and event development prediction. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
