04 AI Triple Generation
neode-notebookstutorials
Export
🤖 AI-Powered Triple Generation
This cookbook covers how to automatically generate knowledge graph triples from unstructured text using Neode's AI capabilities.
What You'll Learn
- Generating triples from text automatically
- Processing articles and documents
- Building extraction pipelines
[ ]
Basic Triple Generation
Pass any text context to the AI and it will extract structured triples:
[3]
📝 Input Text: Albert Einstein was a German-born theoretical physicist who developed the theory of relativity. He was born in Ulm, Germany in 1879. Einstein won the Nobel Prize in Physics in 1921 for his discovery of the law of the photoelectric effect. He later became a US citizen and worked at Princeton University until his death in 1955. 🔗 Generated Triples: ------------------------------------------------------------ Albert Einstein → instance_of → theoretical physicist Albert Einstein → born_on → 1879-03-14 Albert Einstein → born_in → Ulm, Württemberg, Germany Albert Einstein → died_on → 1955-04-18 Albert Einstein → died_in → Princeton, New Jersey, United States Albert Einstein → developed → special theory of relativity Albert Einstein → developed → general theory of relativity Albert Einstein → awarded → Nobel Prize in Physics Albert Einstein → Nobel_Prize_in_Physics_award_year → 1921 Nobel Prize in Physics 1921 → prize_motivation → discovery of the law of the photoelectric effect
Processing News Articles
[3]
📰 Article Triples: ------------------------------------------------------------ GPT-4 → announced_on → 2023-03-14 GPT-4 → developed_by → OpenAI GPT-4 → is_a → large multimodal model GPT-4 → accepts_input_modality → text GPT-4 → accepts_input_modality → images GPT-4 → emits_output_modality → text OpenAI → headquartered_in → San Francisco, California, United States Sam Altman → ceo_of → OpenAI GPT-4 → integrated_into → Bing Chat GPT-4 → available_in → ChatGPT
Generate and Store Triples
Generate triples and immediately add them to your knowledge graph:
[4]
✅ Extracted and stored 10 triples!
Processing Multiple Documents
[5]
📄 Document 1: Generated 10 triples 📄 Document 2: Generated 10 triples 📄 Document 3: Generated 10 triples ✅ Total triples generated: 30 🔗 All Generated Triples: ------------------------------------------------------------ SpaceX → founded_by → Elon Musk SpaceX → incorporated_on → 2002-03-14 SpaceX → operates_facility_in → Hawthorne, California SpaceX → headquartered_in → Starbase, Texas Falcon 9 → designed_and_manufactured_by → SpaceX Falcon 9 → first_launched_on → 2010-06-04 Dragon (SpaceX spacecraft family) → developed_by → SpaceX Crew Dragon Demo-2 → launched_on → 2020-05-30 SpaceX → sent_astronauts_to → International Space Station SpaceX Hawthorne facility → has_address → 1 Rocket Road, Hawthorne, California Tesla, Inc. → incorporated_on → 2003-07-01 Tesla Motors, Inc. → incorporated_by → Martin Eberhard Tesla Motors, Inc. → incorporated_by → Marc Tarpenning Elon Musk → became_chairman_of → Tesla, Inc. Elon Musk → became_chairman_in → 2004-02 Elon Musk → named_chief_executive_officer_of → Tesla, Inc. Elon Musk → became_ceo_in → 2008 Tesla, Inc. → headquartered_in → Austin, Texas Tesla, Inc. → manufactures_vehicle_model → Tesla Model S Tesla, Inc. → manufactures_vehicle_model → Tesla Model 3 Neuralink → is_a → American neurotechnology company Neuralink → founded_on → 2016-06-21 Elon Musk → cofounded → Neuralink Neuralink → headquarters_location → Fremont, California, United States Neuralink → develops → implantable brain-computer interfaces Jared Birchall → chief_executive_officer_of → Neuralink Neuralink → website → https://neuralink.com/ Neuralink → headquarters_address → 7400 Paseo Padre Pkwy, Fremont, CA 94555 Neuralink → received_FDA_approval_for_human_clinical_trials_in → May 2023 Neuralink → began_first_human_trials_in → September 2023
Building a Document Processing Pipeline
[6]
🔄 Processing documents... Processed 'python_wiki': 10 triples Processed 'js_wiki': 10 triples Processed 'rust_wiki': 10 triples ✅ Total triples extracted: 30
💾 Saved 30 triples to knowledge graph
Best Practices for Triple Generation
1. Chunk Large Documents
[7]
Split into 8 chunks Chunk 1: 10 triples Chunk 2: 10 triples
2. Deduplicate Generated Triples
[8]
Original: 3 triples After deduplication: 2 triples
Next Steps
- 05_graphs_and_entities.ipynb - Organize knowledge into separate graphs and manage entities