SweetieChat: A Strategy-Enhanced Role-playing Framework for Diverse Scenarios Handling Emotional Support Agent

Jing Ye,Lu Xiang,Yaping Zhang,Chengqing Zong

2024-12-11

Abstract:Large Language Models (LLMs) have demonstrated promising potential in providing empathetic support during interactions. However, their responses often become verbose or overly formulaic, failing to adequately address the diverse emotional support needs of real-world scenarios. To tackle this challenge, we propose an innovative strategy-enhanced role-playing framework, designed to simulate authentic emotional support conversations. Specifically, our approach unfolds in two steps: (1) Strategy-Enhanced Role-Playing Interactions, which involve three pivotal roles -- Seeker, Strategy Counselor, and Supporter -- engaging in diverse scenarios to emulate real-world interactions and promote a broader range of dialogues; and (2) Emotional Support Agent Training, achieved through fine-tuning LLMs using our specially constructed dataset. Within this framework, we develop the \textbf{ServeForEmo} dataset, comprising an extensive collection of 3.7K+ multi-turn dialogues and 62.8K+ utterances. We further present \textbf{SweetieChat}, an emotional support agent capable of handling diverse open-domain scenarios. Extensive experiments and human evaluations confirm the framework's effectiveness in enhancing emotional support, highlighting its unique ability to provide more nuanced and tailored assistance.

Computation and Language

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are as follows: When current large - language models (LLMs) provide emotional support, their responses are often too long - winded or formulaic and cannot fully meet the diverse emotional support needs in the real world. Specifically, these problems include:

1. **Lack of response diversity**: The conversations generated by existing LLMs lack diversity, resulting in repetitive and predictable interaction content, which is difficult to meet the unique needs of different help - seekers.
2. **Poor implementation effect of strategies**: LLMs tend to use some common strategies, such as providing hope, affirmation, and advice, while ignoring deeper - level understanding and support strategies, which makes them insufficient in truly understanding and helping users.

To solve these problems, the author proposes an innovative **strategy - enhanced role - playing framework** aimed at simulating real - life emotional support conversations. This framework is achieved through the following two steps:

1. **Strategy - enhanced role - playing interaction**: Introduce three key roles - the Seeker, the Strategy Counselor, and the Supporter - to simulate diverse scenarios and promote broader conversations.
2. **Emotional support agent training**: Fine - tune the LLMs' emotional support capabilities by fine - tuning on a specially constructed dataset. Within this framework, the author has developed a dataset named ServeForEmo, which contains more than 3,700 multi - round conversations and more than 62,800 statements.

In addition, based on this framework, the author has also developed an emotional support agent named SweetieChat, which can handle diverse open - domain scenarios. Experimental results and human evaluations show that this framework has significant advantages in improving the effectiveness of emotional support and can provide more detailed and customized help.

### Formula presentation

To ensure the diversity of conversations and the effectiveness of strategies, the author uses the following formula to select appropriate strategies:

\[ o_i=\arg\max_{o\in O}P(o|D_i^j) \]

where:
- \( O \) represents the set of strategies,
- \( D_i^j \) represents the history of the \( j \) - th round of conversation, including the previous statement sequence \( \{ u_i^1,r_i^1,\ldots,r_i^{j - 1},u_i^j \} \),
- \( u_i^j \) is the \( j \) - th statement of the Seeker,
- \( r_i^{j - 1} \) is the most recent response of the Supporter.

In this way, the author ensures the quality of conversations and the rationality of strategies, thereby improving the effect of emotional support.