Personality Profiling for Literary Character Dialogue Agents with Human Level Attributes
Authors: Rusnachenko, N. and Liang, H.
Journal: Lecture Notes in Computer Science
Volume: 15509 LNCS
Pages: 240-254
eISSN: 1611-3349
ISSN: 0302-9743
DOI: 10.1007/978-3-031-82484-5_18
Abstract:Equipping personalities to dialogue agents can help to better engage end-users. However, how to profile personality remains an open research question due to the difficulties of obtaining real human data. As classic literary characters often encapsulate typical human personality traits, literature books has been used as a high quality data source to construct personality profiles for dialogue agents. Existing work mainly focuses on using external reviews and human experts’ annotations to profile character personalities. The in-text comments about the personality of characters in a literature book itself have been ignored. In this paper, we propose a new NLP task called character comments annotation to annotate the in-text comments about the personality of characters including dialogue utterances and surrounding text, paragraphs mentioning a character. We constructed new personality annotated dialogue datasets based on Gutenberg literature book project. We propose a workflow to automatically profile literary characters from literature novel books. Two personality profiling models have been proposed, including (i) psychological personality traits vocabulary-based spectrum (spectrums) approach and (ii) a tf-idf based words selection as a baseline approach. We applied the proposed personality models in dialogue response prediction tasks with ranking-based and generative dialogue agents. The results show that the fine-tuned dialogue agents with spectrums profiles surpass those trained without them by 2.5% (Hits 1@20) for ranking-based, and by 8% (Rouge-1) for generative agents. The implementation of the workflow with study-related resources is publicly available: https://github.com/nicolay-r/book-persona-retriever
Source: Scopus
Personality Profiling for Literary Character Dialogue Agents with Human Level Attributes
Authors: Rusnachenko, N. and Liang, H.
Journal: MACHINE LEARNING, OPTIMIZATION, AND DATA SCIENCE, LOD 2024, PT II
Volume: 15509
Pages: 240-254
eISSN: 1611-3349
ISBN: 978-3-031-82483-8
ISSN: 0302-9743
DOI: 10.1007/978-3-031-82484-5_18
Source: Web of Science (Lite)
Personality Profiling for Literary Character Dialogue Agents with Human Level Attributes
Authors: Rusnachenko, N. and Huizhi, L.
Conference: The 10th International Conference on Machine Learning, Optimization, and Data Science – September 22 – 25, 2024
Dates: 22-25 September 2024
Publisher: Springer
Abstract:Equipping personalities to dialogue agents can help better engage end-users. In the domain of fictional characters from literature novel books, the choice of the related profiles and sources for their completion remains an open research question. At present, the major amount of dialogue agents relies on profiles preparation from third party resources with such as reviews and crowd-sources data primary. However, counting the books itself as a source of the related information on literary characters profiling is still remains poorly covered. This paper proposes the workflow of automatic profiling fictional character from literature novel books. The workflow is aimed at character personalities construction by solely rely on their comments in book: dialogue utterances and surrounding text, paragraphs. We proposed two personality profiling models including (i) psychological personality traits vocabulary-based spectrum (spectrums) approach and (ii) a tf-idf based words election as a baseline approach. We applied the proposed personality models to dialogue agents in following tasks: character response prediction (E1) and character response generation (E2). For experiments conduction, the personality annotated dialogue datasets based on Gutenberg literature book service were constructed. The experimental results show that fine-tuning dialogues agents with our proposed personality profiles surpasses those trained without them by 2.5% (Hits 1@20) in E1 and by 8% (Rouge-1) in E2. The workflow implementation and supplementary resources are publicly available: https://github.com/nicolay-r/book-persona-retriever
Source: Manual