Replication Data for: Uncovering LLMs for Service-Composition: Challenges and Opportunities

Experimental results for the ICSOC 2023 AI-PA position paper "Uncovering LLMs for Service-Composition: Challenges and Opportunities."
<br>
<ul>
<li><i>Exemplars</i>: List of scenarios found in the Google Scholar literature search.
<li><i>Experiment 1 Service Discovery</i>: Chat history for experiment 1 asking ChatGPT for existing real services.</li>
<li><i>Experiment 2 Service Composition</i>: Chat history and service composition for experiment 2 asking ChatGPT for a service composition in Python using a natural language task and the list of services from experiment 1.</li>
<li><i>Experiment 3 Combined Service Discovery and Composition</i>: Chat history and service composition for experiment 3 asking ChatGPT for a service composition in Python using a natural language task without a list of services.</li>
</ul>
Each experiment in the dataset has its own folder (use the tree view to see the folder layout of the files). Chats in experiments 2 and 3 are accompanied by their service composition in Python from that chat as an extra file.

本数据集为ICSOC 2023大会AI-PA立场论文《面向服务组合的大语言模型（LLMs）探析：挑战与机遇》的实验结果。
- 示例样本（Exemplars）：谷歌学术（Google Scholar）文献检索所得的场景列表。
- 实验1：服务发现（Service Discovery）：实验1的对话历史，内容为向ChatGPT查询现有真实服务的交互记录。
- 实验2：服务组合（Service Composition）：实验2的对话历史与服务组合结果，该实验要求ChatGPT基于自然语言任务及实验1所得的服务列表，使用Python实现服务组合。
- 实验3：服务发现与组合联合任务（Combined Service Discovery and Composition）：实验3的对话历史与服务组合结果，该实验要求ChatGPT基于自然语言任务且不提供预设服务列表，使用Python实现服务组合。
本数据集内的每项实验均配有独立文件夹，可通过树形视图查看文件的文件夹布局。实验2与实验3的对话内容均附带对应对话中生成的Python服务组合代码文件作为额外附件。