整理一套包含提示词框架、指令提示、角色提示等23种ChatGPT提问技巧的完整指南,从基础到进阶逐章示例,适合入门用户系统掌握高质量提问方法。
中文版提示词
请根据以下23个章节,为我系统讲解ChatGPT的高质量提问技巧,每章给出定义和示例: 第1章:提示词技术概论。提示词框架由任务(希望完成的目标)、指令(应遵循的规则)、角色(应扮演的身份)三个元素组成。示例:你现在是一个翻译家(角色)。我希望你将下面中文翻译成简洁清晰的英文:xxx(任务)。现在开始翻译(指令)。 第2章:指令提示。按指示生成符合法律法规的租房合同。 第3章:角色提示。你作为一名离婚律师,生成离婚协议。 第4章:标准提示。按"任务+说明+角色+需突出点"公式组织,如:你是一名技术专家,写一篇客观且信息丰富的产品评论,强调新笔记本电脑的强大功能。 第5章:零样本、单样本和少量样本。零样本直接生成;单样本给一个示例;少量样本给多个示例。 第6章:让我们思考一下。如:让我们思考一下气候变化对农业的影响。 第7章:自我一致性提示。生成与给定产品信息一致的产品评论。 第8章:关键词。用关键词引导内容方向。 第9章:知识生成提示。生成新的原创信息。 第10章:知识整合提示。将新信息与既有知识整合。 第11章:多项选择提示。判断文本是消极、积极还是客观中立。 第12章:特定风格。如:以王小波小说的风格写一篇关于恐龙的小说,200字左右。 第13章:特定要求。在特定风格基础上加入指定词汇。 第14章:问答提示。定义某个词汇。 第15章:摘要和总结。如:300字左右总结《红楼梦》。 第16章:对话。生成人物之间的对话。 第17章:对抗性提示。生成难以被翻译的内容。 第18章:分类。将词汇按类别分类。 第19章:强化学习。使用强化学习生成风格一致的文本。 第20章:课程学习提示。使用课程学习完成翻译任务。 第21章:情感分析提示。对文本进行积极/消极/中立的情感分类。 第22章:实体识别。识别人名、机构、地点和日期。 第23章:文本分类。将邮件按内容分类,如垃圾邮件、重要邮件、紧急邮件。
英文版提示词
Based on the following 23 chapters, systematically teach me high-quality ChatGPT prompting techniques, giving a definition and example for each chapter: Chapter 1: Introduction to prompting. A prompt framework consists of three elements: task (the goal), instruction (the rules to follow), and role (the identity to assume). Example: You are now a translator (role). I want you to translate the following Chinese into concise, clear English: xxx (task). Begin translating now (instruction). Chapter 2: Instruction prompting. Generate a legally compliant rental contract as instructed. Chapter 3: Role prompting. As a divorce lawyer, generate a divorce agreement. Chapter 4: Standard prompting. Organize using "task + description + role + key points" formula, e.g., you are a tech expert writing an objective and informative product review highlighting the new laptop's powerful features. Chapter 5: Zero-shot, one-shot, and few-shot. Generate directly with zero examples, one example, or several examples. Chapter 6: "Let's think about it." E.g., let's think about the impact of climate change on agriculture. Chapter 7: Self-consistency prompting. Generate a product review consistent with the given product information. Chapter 8: Keywords. Use keywords to steer content direction. Chapter 9: Knowledge generation prompting. Generate new, original information. Chapter 10: Knowledge integration prompting. Integrate new information with existing knowledge. Chapter 11: Multiple-choice prompting. Determine whether a text is negative, positive, or neutral. Chapter 12: Specific style. E.g., write a novel about dinosaurs in the style of Wang Xiaobo, around 200 words. Chapter 13: Specific requirements. Add designated vocabulary on top of a specific style. Chapter 14: Q&A prompting. Define a term. Chapter 15: Summary and abstract. E.g., summarize "Dream of the Red Chamber" in about 300 words. Chapter 16: Dialogue. Generate a conversation between characters. Chapter 17: Adversarial prompting. Generate content that is difficult to translate. Chapter 18: Classification. Classify words into categories. Chapter 19: Reinforcement learning. Use reinforcement learning to generate stylistically consistent text. Chapter 20: Curriculum learning prompting. Use curriculum learning for a translation task. Chapter 21: Sentiment analysis prompting. Classify text as positive, negative, or neutral. Chapter 22: Entity recognition. Identify names, organizations, places, and dates. Chapter 23: Text classification. Classify emails into categories such as spam, important, or urgent.

◯ 评论 0