Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Enhancing Personality Recognition in Dialogue by Data Augmentation and Heterogeneous Conversational Graph Networks (2401.05871v2)

Published 11 Jan 2024 in cs.CL

Abstract: Personality recognition is useful for enhancing robots' ability to tailor user-adaptive responses, thus fostering rich human-robot interactions. One of the challenges in this task is a limited number of speakers in existing dialogue corpora, which hampers the development of robust, speaker-independent personality recognition models. Additionally, accurately modeling both the interdependencies among interlocutors and the intra-dependencies within the speaker in dialogues remains a significant issue. To address the first challenge, we introduce personality trait interpolation for speaker data augmentation. For the second, we propose heterogeneous conversational graph networks to independently capture both contextual influences and inherent personality traits. Evaluations on the RealPersonaChat corpus demonstrate our method's significant improvements over existing baselines.

PDF HTML Abstract

Introduction

The ability to discern an individual's personality through dialogue is critical for enhancing human-robot interaction. Personality recognition technologies strive to identify characteristic patterns of feeling, thinking, and behaving, which are unique to each person. In the field of artificial intelligence, this capability opens the door to more nuanced and adaptable interactive systems, such as chatbots or personal assistants, capable of delivering responses tailored to the user's personality.

Challenges in Personality Recognition

Personality recognition from dialogue faces two primary obstacles. Firstly, the limited number of speakers in dialogue corpora restricts the development of speaker-independent models. This limitation poses a challenge as it does not adequately represent the diversity of individual personalities. Secondly, the complexity of dialogues, which include both the context of the conversation and the intrinsic personality traits of the speaker, presents a significant modeling challenge.

Solutions and Contributions

A novel solution has been proposed which addresses these two significant challenges. First, to mitigate the data scarcity problem, a data augmentation method named personality trait interpolation has been introduced. This technique enriches the dataset by creating synthetic dialogue and corresponding personality traits by blending two existing data points. Consequently, it enhances the speaker diversity in the training data, which is imperative for the robust recognition of unseen speakers' personalities.

Secondly, the paper introduces a sophisticated model, the heterogeneous conversational graph network (HC-GNN), designed to differentiate and analyze the unique interdependencies between conversants and the intra-dependencies within an individual speaker's dialogue contributions independently.

Experimental Findings

Researchers evaluated their methods using the RealPersonaChat corpus - a collection of dialogues with documented personality traits. The results are promising, demonstrating significant improvements in personality recognition across various metrics, outperforming existing baseline methods. These findings suggest that increasing speaker diversity is advantageous and that their proposed HC-GNN model can effectively capture the nuanced relationships within dialogue for enhanced personality recognition.

Future Directions

While the proposed methods have shown effectiveness, there is still room for exploration, particularly in the application of context in dialogue settings. Future research will focus on further refining dialogue-based personality recognition models to better understand and utilize the complexities of conversational data.

The benefits of these advancements in personality recognition extend beyond human-robot interaction. They could be instrumental in various applications, from personalized education and healthcare to customer service and entertainment, bringing about more personalized and engaging AI-driven experiences.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (33)
  1. Using linguistic cues for the automatic recognition of personality in conversation and text. J. Artif. Int. Res., 30(1):457–500, nov 2007. ISSN 1076-9757.
  2. An introduction to the five-factor model and its applications. Journal of personality, 60(2):175–215, 1992.
  3. Linguistic styles: language use as an individual difference. Journal of personality and social psychology, 77(6):1296, 1999.
  4. Text-based automatic personality prediction using kgrat-net: a knowledge graph attention network classifier. Scientific Reports, 12(1):21453, 2022.
  5. Desprompt: Personality-descriptive prompt tuning for few-shot personality recognition. Information Processing & Management, 60(5):103422, 2023.
  6. Chalearn lap 2016: First round challenge on first impressions-dataset and results. In Computer Vision–ECCV 2016 Workshops: Amsterdam, The Netherlands, October 8-10 and 15-16, 2016, Proceedings, Part III 14, pages 400–418. Springer, 2016.
  7. Can chatgpt’s responses boost traditional natural language processing? IEEE Intelligent Systems, 38(05):5–11, sep 2023. ISSN 1941-1294. doi: 10.1109/MIS.2023.3305861.
  8. PANDORA talks: Personality and demographics on Reddit. In Proceedings of the Ninth International Workshop on Natural Language Processing for Social Media, pages 138–152, June 2021.
  9. Personality recognition in conversations using capsule neural networks. In IEEE/WIC/ACM International Conference on Web Intelligence, pages 180–187, 2019.
  10. Automatic text-based personality recognition on monologues and multiparty dialogues using attentive networks and contextual embeddings (student abstract). In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 13821–13822, 2020.
  11. CPED: A large-scale chinese personalized and emotional dialogue dataset for conversational ai. arXiv preprint arXiv:2205.14727, 2022. URL https://arxiv.org/abs/2205.14727.
  12. Personality in its natural habitat: manifestations and implicit folk theories of personality in daily life. Journal of personality and social psychology, 90(5):862, 2006.
  13. RealPersonaChat: A realistic persona chat corpus with interlocutors’ own personalities. In Proceedings of the 37th Pacific Asia Conference on Language, Information and Computation, 2023.
  14. Nl-augmenter: A framework for task-sensitive natural language augmentation. arXiv preprint arXiv:2112.02721, 2021.
  15. EDA: Easy data augmentation techniques for boosting performance on text classification tasks. In EMNLP-IJCNLP, pages 6382–6388, November 2019.
  16. Paraphrase augmented task-oriented dialog generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 639–649, July 2020.
  17. Improving neural machine translation models with monolingual data. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 86–96, August 2016.
  18. Graph attention networks. arXiv preprint arXiv:1710.10903, 2017.
  19. How attentive are graph attention networks? arXiv preprint arXiv:2105.14491, 2021.
  20. A survey on image data augmentation for deep learning. Journal of big data, 6(1):1–48, 2019.
  21. BPE-dropout: Simple and effective subword regularization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1882–1892, July 2020.
  22. A task in a suit and a tie: paraphrase generation with semantic augmentation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 33, pages 7176–7183, 2019.
  23. Noising and denoising natural language: Diverse backtranslation for grammar correction. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 619–628, June 2018.
  24. Generalizing back-translation in neural machine translation. In Proceedings of the Fourth Conference on Machine Translation (Volume 1: Research Papers), pages 45–52, August 2019.
  25. Data augmentation for low resource neural machine translation for sotho-tswana languages. In Proceedings of Southern African Conference for AI Research (SACAIR 2023), Johannesburg, South Africa, 2023.
  26. Jason Wei. Good-enough example extrapolation. In Marie-Francine Moens, Xuanjing Huang, Lucia Specia, and Scott Wen-tau Yih, editors, Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 5923–5929, November 2021.
  27. CONSK-GCN: conversational semantic-and knowledge-oriented graph convolutional network for multimodal emotion recognition. In 2021 IEEE International Conference on Multimedia and Expo (ICME), pages 1–6. IEEE, 2021.
  28. Context-and knowledge-aware graph convolutional network for multimodal emotion recognition. IEEE MultiMedia, 29(3):91–100, 2022.
  29. Hag: Hierarchical attention with graph network for dialogue act classification in conversation. In ICASSP, pages 1–5. IEEE, 2023.
  30. Orders are unwanted: dynamic deep graph convolutional network for personality detection. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 13896–13904, 2023.
  31. Modeling relational data with graph convolutional networks. In The Semantic Web: 15th International Conference, ESWC 2018, Heraklion, Crete, Greece, June 3–7, 2018, Proceedings 15, pages 593–607. Springer, 2018.
  32. Luke: deep contextualized entity representations with entity-aware self-attention. arXiv preprint arXiv:2010.01057, 2020.
  33. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907, 2016.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Yahui Fu (8 papers)
  2. Haiyue Song (18 papers)
  3. Tianyu Zhao (73 papers)
  4. Tatsuya Kawahara (61 papers)
Citations (1)
View on Semantic Scholar
X Twitter Logo Streamline Icon: https://streamlinehq.com

Tweets