Evaluating named entity recognition tools for extracting social networks from novels

Area of Science:

• Computational Linguistics
• Digital Humanities
• Social Network Analysis

Background:

• Computer-assisted analysis of literary works is growing.
• Social network extraction from novels offers insights into community structures and interactions.
• Existing tools often lack domain specificity for literature and focus on older texts.

Purpose of the Study:

• To evaluate natural language processing (NLP) tools for automatic social network extraction from novels.
• To analyze the network structures derived from literary texts.
• To assess the suitability of current NLP techniques for both historical and modern literature.

Main Methods:

• Applying NLP tools to identify named entities and relations in novels.
• Constructing social networks based on extracted information.
• Analyzing the resulting network structures.
• Comparing results across different literary periods.

Main Results:

• No significant differences were found in social network extraction between 19th/early 20th-century novels and modern novels.
• Both historical and modern literary texts demonstrated considerable variance in network structures.
• Identified challenges in named entity recognition within the novel corpus.
• Proposed methods to address these named entity recognition issues.

Conclusions:

• Current NLP tools are applicable to a range of literary periods, but results vary.
• Further research is needed to refine NLP for literary analysis.
• This work contributes to developing more culturally-aware artificial intelligence systems.