大型語言模型（LLM）具備優異的查詢理解與文本生成能力，有效解決傳統檢索系統中的語意挑戰。然而，其高度依賴預訓練知識，容易產生過時或虛構的資訊，特別是在即時新聞檢索的應用情境中更為明顯。

相較於其他資料類型，新聞資料具備時效性高、語意重複率高、資訊片段化等特性，進一步加深了檢索與生成任務的挑戰。因此，本研究選擇以新聞資料作為基礎，實作一套檢索增強生成系統，並將情境設定於新聞專業人員使用內部查詢系統時的多樣需求，藉此觀察RAG技術是否能有效降低幻覺現象，提升資訊正確性。

研究首先自行爬取ETtoday新聞資料作為知識來源，並設計實驗比較四種模型配置：Llama3-8B、Llama3-8B 搭配 RAG、GPT-4、以及 GPT-4 搭配 RAG。測試任務涵蓋事實問答、事件綜合與新聞摘要等多個面向，並針對生成內容的正確性與完整性等進行評分。結果顯示，引入RAG技術可顯著提升模型回應的事實準確性，並有效減少生成幻覺的情況。

在系統開發方面，涵蓋從資料爬取、前處理、文件切分至向量資料庫設計，最終以 Gradio 實作互動介面。研究過程中亦強調檢索品質的追蹤與回饋機制，確保最終生成答案的可靠性。

Large language models (LLMs) excel at query understanding and text generation but are prone to outdated or hallucinated content, especially in real-time news retrieval. News data poses unique challenges such as high redundancy, temporal sensitivity, and fragmented information.
This study develops a retrieval-augmented generation (RAG) system tailored for internal news query scenarios and evaluates its effectiveness using ETtoday news data. Four model setups—LLaMA3-8B, LLaMA3-8B+RAG, GPT-4, and GPT-4+RAG—were tested across tasks like factual Q&A and summarization.
The evaluation focused on factual accuracy and completeness. Results show that RAG significantly improves output reliability while reducing hallucinations. The system implementation includes data crawling, data preprocessing, word chunking, and vector database building, with an interactive frontend built using Gradio.
Throughout the experiment, emphasis was placed on monitoring retrieval quality and incorporating feedback checking to ensure the reliability of final outputs.

Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P.,
Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., . . . Amodei, D. (2020). Language Models are Few-Shot Learners. Neural Information Processing Systems, 33, 1877–1901. https://proceedings.neurips.cc/paper/2020/file/1457c0d6bfcb4967418bfb8ac142f64a-Paper.pdf
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the Dangers of Stochastic Parrots. On The Dangers of Stochastic Parrots: Can Language Models Be Too Big?, 610–623. https://doi.org/10.1145/3442188.3445922
Chase, L. (2022). LangChain: Building applications with LLMs through composability. GitHub Documentation.
Gharge, S., & Chavan, M. (2017). An integrated approach for malicious tweets detection using NLP. 2017 International Conference on Inventive Communication and Computational Technologies (ICICCT). https://doi.org/10.1109/icicct.2017.7975235
Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” in Advances in Neural Information Processing Systems 27, Montreal, Quebec, Canada, 2014, pp. 2672−2680.
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. In MIT Press eBooks. https://dl.acm.org/citation.cfm?id=3086952
Guu, K., Lee, K., Tung, Z., Pasupat, P., & Chang, M. (2020, February 10). REALM: Retrieval-Augmented Language Model Pre-Training. arXiv.org. https://arxiv.org/abs/2002.08909
Hearst, M. A. (2009). Search user interfaces. http://ci.nii.ac.jp/ncid/BA91702558
Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735
Hochreiter, S. (1998). The vanishing gradient problem during learning recurrent neural nets
and problem solutions. International Journal of Uncertainty Fuzziness and Knowledge-Based Systems, 06(02), 107–116. https://doi.org/10.1142/s0218488598000094
Holtzman, A., Buys, J., Du, L., Forbes, M., & Choi, Y. (2020). The curious case of neural text degeneration. arXiv (Cornell University). https://arxiv.org/pdf/1904.09751.pdf
Izacard, G., & Grave, E. (2020, July 2). Leveraging Passage Retrieval with
Generative Models for Open Domain Question Answering. arXiv.org. https://arxiv.org/abs/2007.01282
Johnson, J., Douze, M., & Jégou, H. (2017, February 28). Billion-scale similarity search with GPUs. arXiv.org. https://arxiv.org/abs/1702.08734
Karpukhin, V., Oguz, B., Min, S., Lewis, P., Wu, L., Edunov, S., Chen, D., & Yih, W. (2020). Dense Passage Retrieval for Open-Domain Question Answering. Dense Passage Retrieval for Open-domain Question Answering. https://doi.org/10.18653/v1/2020.emnlp-main.550
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444.
https://doi.org/10.1038/nature14539
Maynez, J., Narayan, S., Bohnet, B., & McDonald, R. (2020, May 2). On faithfulness and factuality in abstractive summarization. arXiv.org. https://arxiv.org/abs/2005.00661
Manning, C. D., Raghavan, P., & Schütze, H. (2009). Introduction to information retrieval.
Choice Reviews Online, 46(05), 46–2715. https://doi.org/10.5860/choice.46-2715
Mitra, B., & Craswell, N. (2018). An Introduction to Neural Information Retrieval t.
Foundations and Trends® in Information Retrieval, 13(1), 1–126. https://doi.org/10.1561/1500000061
Mikolov, T., Karafiát, M., Burget, L., Černocký, J., & Khudanpur, S. (2010). Recurrent
neural network based language model. Interspeech 2022. https://doi.org/10.21437/interspeech.2010-343
Masterman, T., Besen, S., Sawtell, M., & Chao, A. (n.d.). The landscape of emerging AI agent architectures for reasoning, planning, and tool calling: a survey. arXiv.org. https://arxiv.org/abs/2404.11584
Nogueira, R., & Cho, K. (2019, January 13). Passage Re-ranking with BERT. arXiv.org. https://arxiv.org/abs/1901.04085
Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations and
Trends in Information Retrieval Vol. 2, No 1-2, 1–135 http://dx.doi.org/10.1561/1500000011
Robertson, S., & Zaragoza, H. (2009). The Probabilistic Relevance Framework: BM25 and
beyond. Foundations and Trends® in Information Retrieval, 3(4), 333–389. https://doi.org/10.1561/1500000019
Ramit Sawhney, Harshit Joshi, Saumya Gandhi, and Rajiv Ratn Shah. 2020.
A Time-Aware Transformer Based Model for Suicide Ideation Detection on Social Media. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7685–7697, Online. Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.emnlp-main.619
Shuster, K., Poff, S., Chen, M., Kiela, D., & Weston, J. (2021). Retrieval augmentation reduces hallucination in conversation. Empirical Methods in Natural Language Processing, 3784–3803. https://aclanthology.org/2021.findings-emnlp.320/
Trabelsi, M., Chen, Z., Davison, B. D., & Heflin, J. (2021). Neural ranking models for
document retrieval. Information Retrieval, 24(6), 400–444. https://doi.org/10.1007/s10791-021-09398-0
Thorne, J., & Vlachos, A. (2018). Automated fact checking: task formulations, methods and future directions. arXiv (Cornell University), 3346–3359. https://arxiv.org/pdf/1806.07687.pdf
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N.,
Kaiser, L., & Polosukhin, I. (2017). Attention is All you Need. Advances in neural information processing systems, 30.
https://doi.org/10.48550/arXiv.1706.03762
Yasunaga, M., Ren, H., Bosselut, A., Liang, P., & Leskovec, J. (2021). QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. https://doi.org/10.18653/v1/2021.naacl-main.45
Zhang, Y., Ni, A., Mao, Z., Wu, C. H., Zhu, C., Deb, B., Awadallah, A. H., Radev, D., & Zhang, R. (2021, October 16). SUMM^N: a Multi-Stage summarization framework for long input dialogues and documents. arXiv.org. https://arxiv.org/abs/2110.10150
Zhang, T., Ladhak, F., Durmus, E., Liang, P., McKeown, K., & Hashimoto, T. B. (2023, January 31). Benchmarking large language models for news summarization. arXiv.org. https://arxiv.org/abs/2301.13848