[Paper Review] A Survey on Sentiment and Emotion Analysis for Computational Literary Studies

Evgeny Kim, Roman Klinger|arXiv (Cornell University)|Aug 9, 2018

Sentiment Analysis and Opinion MiningComputer Science211 references63 citations

TL;DR

This survey provides a comprehensive overview of sentiment and emotion analysis methods in computational literary studies, focusing on how computational techniques can model emotional dynamics in literary texts. It evaluates dictionary-based, machine learning, and deep learning approaches, emphasizing challenges such as contextual interpretation, character-level emotion tracking, and the need for domain-specific annotation, while advocating for integrated psychological and narrative models to advance the field.

ABSTRACT

Emotions are a crucial part of compelling narratives: literature tells us about people with goals, desires, passions, and intentions. Emotion analysis is part of the broader and larger field of sentiment analysis, and receives increasing attention in literary studies. In the past, the affective dimension of literature was mainly studied in the context of literary hermeneutics. However, with the emergence of the research field known as Digital Humanities (DH), some studies of emotions in a literary context have taken a computational turn. Given the fact that DH is still being formed as a field, this direction of research can be rendered relatively new. In this survey, we offer an overview of the existing body of research on emotion analysis as applied to literature. The research under review deals with a variety of topics including tracking dramatic changes of a plot development, network analysis of a literary text, and understanding the emotionality of texts, among other topics.

Motivation & Objective

• To map and analyze existing computational methods for sentiment and emotion analysis in literary texts.
• To identify methodological gaps in current approaches, especially regarding narrative context and character-level emotion tracking.
• To highlight the unique challenges in literary emotion analysis compared to other domains like social media or news.
• To advocate for integrating psychological models and narrative structures into computational emotion analysis for literature.
• To support interdisciplinary research by clarifying terminology and methodological choices in emotion analysis for literary studies.

Proposed method

• The survey employs a systematic literature review of peer-reviewed publications in computational literary studies and digital humanities.
• It categorizes methods into dictionary-based, feature-based machine learning, and representation-learning/deep learning approaches.
• The analysis evaluates techniques such as TF-IDF, bag-of-words, SVM, Naïve Bayes, decision trees, and neural embeddings for emotion classification.
• It emphasizes supervised learning with annotated data, noting the scarcity of fine-grained, narrative-level annotations in literary corpora.
• The survey compares methodological transparency and interpretability, favoring lexicon-based methods for explainability despite lower accuracy.
• It discusses the integration of narrative features such as character networks, plot progression, and reader perception into emotion modeling.

Experimental results

Research questions

• RQ1 How do different computational methods for emotion analysis perform when applied to literary texts compared to other domains?
• RQ2 What are the key methodological challenges in analyzing emotions in literature, particularly regarding context, narrative structure, and character development?
• RQ3 To what extent can existing emotion analysis tools be adapted for literary texts without losing interpretability or accuracy?
• RQ4 How can psychological models of emotion be integrated into computational frameworks for literary analysis?
• RQ5 What role does domain expertise play in annotation and model development for literary emotion analysis?

Key findings

• Dictionary-based methods are favored in digital humanities for transparency, despite lower contextual accuracy, due to the need for interpretability in literary analysis.
• Machine learning and deep learning models show strong performance on sentiment classification but require large, fine-grained annotated corpora that are currently scarce in literary studies.
• Emotion analysis in literature is more complex than in social media or news, due to indirect emotional expressions, stylistic devices, and embedded narrative contexts.
• Current methods often fail to model long-range dependencies in emotional trajectories across characters and plot developments.
• There is a growing need for multi-level annotations—covering characters, relationships, and plot arcs—to enable context-aware emotion modeling.
• The integration of psychological models of emotion with narrative structures is essential for advancing computational literary studies.