文章摘要

宋宁远,王晓光.基于情节本体的叙事性文本语义结构化表示方法研究[J].中国图书馆学报,2020,(2):96~113
基于情节本体的叙事性文本语义结构化表示方法研究
Semantic Structured Representation Method of Narrative Text Based on Plot Ontology
投稿时间:2019-01-03  修订日期:2019-06-12
DOI:
中文关键词: 叙事性文本  情节本体  语义结构  知识表示  数字人文
英文关键词: Narrative texts Plot ontology Semantic structure Knowledge representation Digital humanities
基金项目:本文系教育部人文社科重点研究基地重大项目“大数据资源语义表示与组织”(编号:16JJD870002)的研究成果之一
作者单位E-mail
宋宁远 武汉大学信息管理学院 湖北 武汉430072  
王晓光 武汉大学信息管理学院 湖北 武汉430072 whu_wxg@126.com,whu_wxg@126.com 
摘要点击次数: 427
全文下载次数: 298
中文摘要:
      
叙事性文本是对现实世界的记录,也是人文学者的研究对象与基础,对叙事性文本的阐释更是人文学术研究的基本任务。随着数字人文的发展,人文学者对远距离阅读与策略型阅读的需求与日俱增。作为构成叙事性文本的基本单元,事件和情节是文本解读的重点。为了更好地适应人文学者阅读需求的转变,建构叙事性文本智慧数据,支撑数字人文研究,以情节为视角,以事件为基础,对广泛存在的叙事性文本进行深层次语义表示与结构化处理显得尤为重要。但现阶段较为成熟的文本表示方法更适用于结构相对清晰、统一的科学文本,不能很好地表示叙事性文本中存在的复杂叙事结构,因此,以知识表示与知识组织为出发点,构建符合叙事性文本特征的语义结构化表示方法具有十分重要的意义。本文在梳理叙事学基本理论与概念,以及事件知识表示、事件本体与叙事本体等理论研究的基础上,对叙事文本中存在的事件、事件间关系进行语义建模与表示,同时完善了对情节的定义,构建了基于事件的情节本体EBPO,提出了叙事性文本的“层次—网络”结构模型,用以实现对叙事性文本的语义结构化表示。初步标注的实验结果表明,借助情节本体及结构模型,可以实现对叙事性文本的语义结构化处理。图8。表4。参考文献61。


英文摘要:

Narration is the record, deduction and re creation of the development of the human world. Narrative text is a kind of rich semantic text, which uses symbolic systems such as words and images to represent real or fictional events and plots. Because of the different purposes, styles and genres of the text, it has various kinds and rich contents. The interpretation of narrative texts has always been the core and basic work of humanists. With the rise of digital humanities, a new research paradigm has emerged in the field of Humanities and Social Sciences; with more emphasis on data-driven research methods, the way of understanding narrative texts has changed, and new reading methods and research needs of narrative texts, including strategic reading and distant reading, have emerged. In order to fully meet this demand, structuring, semanticizing and standardizing narrative text content and constructing narrative text data infrastructure adapted to the digital humanities research paradigm have gradually become the key to future work.
As a special type of text, the complexity of narrative text lies in its emphasis on events as the basic component and the core structure. Events and plots all have abundant correlations. Therefore, narrative texts are dynamic and complex. This brings challenges to the full expression of narrative text features, the description of narrative structure, and the realization of semantic structured representation for narrative text.
In order to solve the problem, this study started from the basic theory of narratology, summarizes and collates the related statements about events, plots and narratives. Firstly, the characteristics of narrative texts were summarized, and an event knowledge representation scheme for narrative texts was proposed. Followed by the definition of plot, the ontology EBPO (Event Based Plot Ontology, EBPO) was designed and developed. The ontology defines 10 classes and 29 properties, which can express plots and events more clearly. Based on this ontology, a general process of deep semantic annotation oriented to the characteristics of narrative texts was proposed, and a preliminary deep semantic annotation experiment was carried out. In the semantic annotation experiment, five classical novel fragments were selected, such as Pride and Prejudice , Moon and Sixpence , Living , and so on. Natural language processing tool GATE 8.5.1 was used for annotating. With the results of semantic annotation, this paper proposed a hierarchical-network model for narrative texts. Taking Pride and Prejudice , Moon and Sixpenceas examples, the annotation results were reorganized and correlated, and the semantic structured representation of narrative texts was realized.
The results show that the semantic structured representation of narrative text based on plot ontology can achieve more accurate definition on the events and plots. Bases on EBPO ontology and annotation ontology, the large-scale, high-quality, rich semantic narrative text corpus could be built, so the machine-aided event retrieval and reading methods including plot review, character relationship overview, plot development path can be realized. Future research will focus on the definition of plot, narrative mode, narrative style and other issues in order to achieve more accurate content representation of narrative text. 8 figs. 4 tabs. 61 refs.
查看全文   查看/发表评论  下载PDF阅读器