--- title: "Dense Slides: GraphRAG: The Marriage of Knowledge Graphs and RAG: Emil Eifrem" category: "slides" video_id: "knDDGYHnnSI" sourceLabels: ["Captured video frames", "Local OpenCV slide-region detection"] --- # Dense Slides: GraphRAG: The Marriage of Knowledge Graphs and RAG: Emil Eifrem ## Source Video [GraphRAG: The Marriage of Knowledge Graphs and RAG: Emil Eifrem](https://www.youtube.com/watch?v=knDDGYHnnSI) ## Method This deck is slide-only. The existing captured video frame set supplies candidate frames, then local OpenCV rejects sponsor/title/speaker-only frames, crops visible slide surfaces, deduplicates, and saves the cropped slide images. ## Cropped Visible Slides ![[assets/dense-slides/knDDGYHnnSI/slide-001.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-001.html) - AI slide classifier: `content_slide` confidence `0.97` - Text source: agent_vision. - OCR decision: ready — Dense screenshot of a web search homepage with small text. Slide text: > The Evolution of... Web Search ![[assets/dense-slides/knDDGYHnnSI/slide-002.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-002.html) - AI slide classifier: `content_slide` confidence `0.95` - Text source: advanced OCR `rapidocr-live/full`. - OCR decision: ready — Screenshot-heavy slide with small UI text and search results. Slide text: > AIE Moscone Center:Homepage FeredEEMwESCog2024/gDesgA SEMICONwESTCog2024/FgmDeigAuto Events > Floor Plans Mosc ore North.4 50 MosconeCenter > Explore the Neighborhood Nearty Nngroon 00d Aree A299ct > Directions andParking BART offers fat, co > Contact Us TUAG 1 s t os n ou faet Caelfornia Uanited Stanies. The complex cone od.Wkipeda s De lurgr9t con seg u xeyueo segree9 pe uo) siss of tvee main WT es te > People also ask1 Which BART stop for Moscone Center? Parlkcing: Pey Dosing neerly Aodne*s:747 H0e Phene:(415)974-4000 erCyd Cy fSnF wire St Sas.Fancisce, CA 94103 > Who owns the Moscone Center? 60:8157m Coses11PM > erved2024 > Microsoft smol aws ![[assets/dense-slides/knDDGYHnnSI/slide-003.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-003.html) - AI slide classifier: `content_slide` confidence `0.99` - Text source: agent_vision. Slide text: > Q: What...? > User > Your Application > Embedding Generator > Q > V > Vector search for V > Relevant Documents and Context > Knowledge Graph With Vector Support ![[assets/dense-slides/knDDGYHnnSI/slide-004.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-004.html) - AI slide classifier: `content_slide` confidence `0.95` - Text source: advanced OCR `rapidocr-live/bright-screen/opencv-adaptive`. - OCR decision: ready — Dense quote, diagram, and QR code are better handled by OCR. Slide text: > amamo Accuracy > AIE users is higher accuracy. The clearest driver of GraphRAG adoption amongst > Across all metrics, our method > demonstrates consistent > in MRR and by 0.32 in BLEU improvements. Notably, it surpasses the baseline by 77.6% > question-answering accuracy? retrieval efficacy and score,substantiating its superior > orved2023 Microsoft smolo aws ![[assets/dense-slides/knDDGYHnnSI/slide-005.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-005.html) - AI slide classifier: `content_slide` confidence `0.95` - Text source: advanced OCR `rapidocr-live/bright-screen/opencv-adaptive`. - OCR decision: ready — Dense quote card and blog screenshot are OCR-suitable. Slide text: > OHigher users is higher accuracy. Accuracy The clearest driverof GraphRAG adoption amongst Microsoft Research Blog 'By combining LLM-generated GraphRAG: Unlocking LLM discovery on narrative private data tro tt may pusor > knowledge graphs and graph > machine learning, GraphRAG > enables us to answer important > classes of questions that we cannot > attempt with baseline RAG alone. > Microsoft > 23. Neo4j Inc, All rlghts reserved 2023 ![[assets/dense-slides/knDDGYHnnSI/slide-006.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-006.html) - AI slide classifier: `content_slide` confidence `0.98` - Text source: agent_vision. Slide text: > Easier Development* > The second reason we hear people choose GraphRAG over vector-only RAG is easier development... once they've pushed through the initial learning curve. ![[assets/dense-slides/knDDGYHnnSI/slide-007.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-007.html) - AI slide classifier: `content_slide` confidence `0.99` - Text source: agent_vision. - OCR decision: ready — Diagram-heavy slide with small dense labels and a text table; OCR will read the body more reliably than manual transcription. Slide text: > ② Easier Development: Why? ![[assets/dense-slides/knDDGYHnnSI/slide-008.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-008.html) - AI slide classifier: `content_slide` confidence `0.97` - Text source: advanced OCR `rapidocr-live/bright-screen/opencv-adaptive`. - OCR decision: ready — Product UI screenshot with many small labels, rows, and buttons; OCR is appropriate for the table and controls. Slide text: > T+ ne04 0 > + n+o-l|+s/3to rb890 dhstatea.reo4ila7te7: Neo4J connectlon Disconect 202-0.1120514 > Drag& Drop O Show fles with status New G014.0 11011 14 > Lestructur ed le.l 0 Namo Th+ Four Wl... Status + Completed Uploid Status +Upedad 27.10 Siza (KB) 1X31 Sourco. YoUtuba 224 411110 > Youtube OperAl +Conpleied +Cpydad:&.07 TEXT wikipo > Tho Batch M.. Conplotad ◆Lpcadad 1507.79 locat fili > Wlklp+dlo > Amhion S3 > Showing t-3 of 3 results Show 5 √ > Gcs > Gp.to Preview Graph [3) Explore Gragh wtth Bloom Dolete Files(3)] ![[assets/dense-slides/knDDGYHnnSI/slide-009.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-009.html) - AI slide classifier: `content_slide` confidence `0.89` - Text source: none. ![[assets/dense-slides/knDDGYHnnSI/slide-010.jpg]] - Recreated text/layout view: [open HTML recreation](/assets/slide-recreations/dense/knDDGYHnnSI/slide-010.html) - AI slide classifier: `content_slide` confidence `0.93` - Text source: none. Classification audit: `raw/sources/slide-ai-classification/dense/knDDGYHnnSI/audit.json`