| dc.contributor.advisor | 胡毓忠 | zh_TW |
| dc.contributor.advisor | Hu, Yuh-Jong | en_US |
| dc.contributor.author (Authors) | 何世堯 | zh_TW |
| dc.contributor.author (Authors) | Ho, Shih-Yao | en_US |
| dc.creator (作者) | 何世堯 | zh_TW |
| dc.creator (作者) | Ho, Shih-Yao | en_US |
| dc.date (日期) | 2025 | en_US |
| dc.date.accessioned | 3-Nov-2025 14:45:17 (UTC+8) | - |
| dc.date.available | 3-Nov-2025 14:45:17 (UTC+8) | - |
| dc.date.issued (上傳時間) | 3-Nov-2025 14:45:17 (UTC+8) | - |
| dc.identifier (Other Identifiers) | G0109971020 | en_US |
| dc.identifier.uri (URI) | https://nccur.lib.nccu.edu.tw/handle/140.119/160073 | - |
| dc.description (描述) | 碩士 | zh_TW |
| dc.description (描述) | 國立政治大學 | zh_TW |
| dc.description (描述) | 資訊科學系碩士在職專班 | zh_TW |
| dc.description (描述) | 109971020 | zh_TW |
| dc.description.abstract (摘要) | 本研究探討在房地產資訊系統中，結合大型語言模型（Large Language Model, LLM）與引導式對話是否能有效降低使用者的認知負擔，並提升需求澄清與推薦的精準度。研究以 Conversation Routines 為提示設計框架，在AWS Bedrock 建置一套分工式架構：由 Bedrock Agent 依 Conversation Routines 規範與槽位填充機制主導互動流程，完成地區、預算、生活機能與其他條件的蒐集；其後透過 RAG 與外部即時搜尋整合知識，並交由 Foundation Model 統整產生推薦結果與可解釋文字。本研究設計六個真實購屋任務情境（明確 ×3、模糊 ×3），在互動式資訊檢索（Interactive Information Retrieval, IIR）觀點下，與產業代表系統「永慶 AI 特助」進行比較；指標包含結果符合度、互動輪數與需求澄清能力。結果顯示：在明確需求情境下，兩系統皆能於單輪內產出可用結果；在模糊需求情境下，本研究系統能透過多輪追問收斂條件，維持推薦品質，並於「完全模糊」情境中仍提供具體建議與物件，而比較系統則無法回應。雖然本系統在模糊情境下通常需要較多互動輪數，但換得較高的需求釐清與建議品質。綜合而言，本研究驗證了引導式對話於高度動態之房地產場域的可行性與實用性，並提供一套可複用的對話設計準則與系統化實作流程。 | zh_TW |
| dc.description.abstract (摘要) | This study investigates whether integrating Large Language Models (LLMs) with guided dialogue in real estate information systems can effectively reduce users’cognitive load and improve requirement clarification and recommendation accuracy. The research adopts Conversation Routines as the prompt design framework and builds a modular architecture on AWS Bedrock: the Bedrock Agent, guided by Conversation Routines and a slot-filling mechanism, manages the interaction process to collect region, budget, amenities, and other conditions. Subsequently, knowledge is integrated through Retrieval-Augmented Generation (RAG) with AWS OpenSearch Service and real-time external search via the Perplexity API, with the results consolidated by the Foundation Model (Claude 3.7 Sonnet) into recommendations and explanatory text. Six real-world housing purchase scenarios (three clear, three ambiguous) were designed and compared with an industry benchmark system, “Yung-Ching AI Assistant,”under the perspective of Interactive Information Retrieval (IIR). Evaluation indicators included result relevance, number of dialogue turns, and requirement clarification ability. The results show that in clear-demand scenarios, both systems can generate usable results within a single turn; in ambiguous-demand scenarios, our system refines conditions through multiple prompts, maintaining recommendation quality, and even in “completely ambiguous”cases it can still provide concrete suggestions and listings, whereas the comparison system could not respond. Although our system typically requires more dialogue turns in ambiguous scenarios, it yields superior requirement clarification and recommendation quality. Overall, this study validates the feasibility and practicality of guided dialogue in the dynamic real estate domain and provides a reusable set of dialogue design guidelines and a systematic implementation framework. | en_US |
| dc.description.tableofcontents | 致謝 i
摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 vii
第一章 緒論 1
第一節 研究動機 1
第二節 研究目的 1
第三節 研究架構 3
第二章 文獻探討 4
第一節 大型語言模型文獻回顧 4
第二節 檢索增強生成文獻回顧 4
第三節 引導式對話 5
第四節 引導式對話與大型語言模型結合 7
第五節 大型語言模型於房地產應用 7
第三章 研究方法 9
第一節 提示設計框架 9
第二節 系統模組 11
第四章 系統設計 18
第一節 系統功能設計 19
第二節 使用者流程 26
第三節 系統流程 26
第五章 研究實作 29
第一節 實驗設計 29
第二節 實驗結果 33
第三節 小結 38
第六章 結論與未來展望 40
第一節 研究結論 40
第二節 未來展望 41
參考文獻 42
附錄 A 本研究系統實驗結果補充 44
附錄 B 永慶 AI 助理實驗結果補充 47 | zh_TW |
| dc.format.extent | 19333481 bytes | - |
| dc.format.mimetype | application/pdf | - |
| dc.source.uri (資料來源) | http://thesis.lib.nccu.edu.tw/record/#G0109971020 | en_US |
| dc.subject (關鍵詞) | 引導式對話 | zh_TW |
| dc.subject (關鍵詞) | 大型語言模型 | zh_TW |
| dc.subject (關鍵詞) | 槽位填充 | zh_TW |
| dc.subject (關鍵詞) | 檢索增強生成 | zh_TW |
| dc.subject (關鍵詞) | AWS Bedrock | zh_TW |
| dc.subject (關鍵詞) | Guided dialogue | en_US |
| dc.subject (關鍵詞) | Large Language Models | en_US |
| dc.subject (關鍵詞) | Slot filling | en_US |
| dc.subject (關鍵詞) | Retrieval-Augmented Generation | en_US |
| dc.subject (關鍵詞) | AWS Bedrock | en_US |
| dc.title (題名) | 建構引導式對話功能於大語言模型上的房地產資訊系統 | zh_TW |
| dc.title (題名) | Construct a Guided Dialogue Function on Large Language Model for Real Estate Information System | en_US |
| dc.type (資料類型) | thesis | en_US |
| dc.relation.reference (參考文獻) | [1] 劉婷慧. 房屋交易應用程式與使用者經驗研究. 世新大學碩士論文 (2018).
[2] Romal Thoppilan, Daniel De Freitas, Jamie Hall, et al. LaMDA: Language Models for Dialog Applications. arXiv preprint arXiv:2201.08239 (2022).
[3] OpenAI. GPT-4 Technical Report. arXiv (2023).
[4] Anthropic. Claude 1 and 2 Product Overview. Anthropic Technical Report(2025).
[5] Ashish Vaswani, Noam Shazeer, Niki Parmar, et al. Attention is All You Need. Advances in Neural Information Processing Systems (2017).
[6] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. arXiv preprint arXiv:1810.04805 (2018).
[7] Tom B. Brown, Benjamin Mann, Nick Ryder, et al. Language Models are Few-Shot Learners. Advances in Neural Information Processing Systems (2020).
[8] Patrick Lewis, Ethan Perez, Alex Piktus, et al. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv preprint arXiv:2005.11401 (2020).
[9] Daniel Jurafsky and James H. Martin. Speech and Language Processing (3rd Edition). Draft Manuscript (2023).
[10] Stefan Larson, Kevin Leach, Jonathan Kim, et al. An Evaluation Dataset for Intent Classification and Out-of-Scope Prediction. EMNLP-IJCNLP (2019).
[11] Yu Zhao and Haoxiang Gao. Utilizing Large Language Models for Information Extraction from Real Estate Transactions. arXiv preprint arXiv:2404.18043 (2024).
[12] Giorgio Robino. Conversation Routines: A Prompt Engineering Framework for Task-Oriented Dialog Systems. arXiv preprint arXiv:2403.17809 (2024).
[13] Kyle Blocksom, John Baker, Sudip Dutta, Maira Ladeira Tanke, and Mark Roy. Getting Started with Amazon Bedrock Agents Custom Orchestrator. Amazon Bedrock Agents Blog (2024).
[14] Amazon Web Services. Anthropic’s Claude 3.7 Sonnet: The First Hybrid Reasoning Model is Now Available in Amazon Bedrock. AWS Blog (2025).
[15] Stefan Larson and Kevin Leach. A Survey of Intent Classification and Slot-Filling Datasets for Task-Oriented Dialog. arXiv preprint arXiv:2207.13211 (2022).
[16] OpenSearch Project. k-NN plugin for OpenSearch: Vector search techniques. OpenSearch Documentation (2023).
[17] Mohammad Mujaheed Hassan, Nobaya Ahmad, and Ahmad Hariza Hashim. Factors Influencing Housing Purchase Decision. International Journal of Academic Research in Business and Social Sciences (2021).
[18] Lizawati Abdullah, Ilyana Bazlin Mohd Nor, Norhaslina Jumadi, and Huraizah Arshad. First-Time Home Buyers: Factors Influencing Decision Making. International Conference on Innovation and Technology for Sustainable Built Environment (ICITSBE 2012) (2012).
[19] 吳聲煌. 影響購屋意願因素之研究. 國立政治大學碩士論文 (2020).
[20] Diane Kelly. Methods for Evaluating Interactive Information Retrieval Systems with Users. Foundations and Trends in Information Retrieval (2009).
[21] 永慶房屋. 永慶 AI 特助. 永慶房屋官方網站 (2025). | zh_TW |
