AIミュージアム

出典

検索

• Google, How Search Works “Organizing information”
• Srinivasan, Google The Keyword (2025-03-03)
• Barroso, Dean & Hölzle, IEEE Micro 23(2) (2003)
• Hölzle, Official Google Blog (2009-01)
• Berners-Lee, “Information Management: A Proposal” (1989)
• CERN, “A short history of the Web”
• Gray, “Web Growth Summary” (1996)
• Netcraft, Web Server Survey (2006-11, 2014-09, 2026-09)
• Pinkerton, WWW2 (1994)
• Koster, WWW1 (1994)
• Reuters (1995-12-15)
• Dean, WSDM keynote (2009)
• Nayak, Google The Keyword (2019-10-25)
• McIlroy, A Research UNIX Reader (1987)
• Knuth, Morris & Pratt, SIAM J. Comput. 6(2) (1977)
• Boyer & Moore, CACM 20(10) (1977)
• Deutsch, “archie - An Electronic Directory Service for the Internet” (1991)
• Manning, Raghavan & Schütze, Introduction to Information Retrieval (2008)
• Rouse & Rouse, Archivum Fratrum Praedicatorum 44 (1974)
• Rockwell & Sinclair, DHQ 14(3) (2020)
• Gull, College & Research Libraries (1953)
• Robertson, in Fifty Years of Information Progress (1994)
• Spärck Jones, Journal of Documentation 28 (1972)
• Robertson & Spärck Jones, JASIS 27 (1976)
• Robertson & Walker, SIGIR (1994)
• Robertson et al., “Okapi at TREC-3” (1994)
• Brin & Page, Computer Networks 30 (1998)
• Harris, Word 10 (1954)
• Deerwester et al., JASIS 41 (1990)
• McBryan, WWW1 (1994)
• Mikolov et al., arXiv:1301.3781 (2013)
• Reimers & Gurevych, EMNLP (2019)
• Karpukhin et al., EMNLP (2020)
• Bengio et al., JMLR 3 (2003)
• Mikolov et al., Interspeech (2010)
• Press & Wolf, EACL (2017)
• Vaswani et al., NeurIPS (2017)
• Radford et al. (2018)
• Yang et al., ICLR (2025)

検索：データとモデル

• データ: 日本語版 Wikipedia のダンプ（2026-11-01）。
• ベクトル: EmbeddingGemma 2（点の図、ベクトルの検索）、word2vec（単語の点の図、言語モデルの始まり）。
• 言語モデル: Gated DeltaNet 6層と注意2層（Qwen3.5 と同じ 3:1 の組み方）、8層・640次元、日本語版 Wikipedia 約9.7億語で学習。