Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
126 tokens/sec
GPT-4o
47 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Narrating Causal Graphs with Large Language Models (2403.07118v1)

Published 11 Mar 2024 in cs.CL

Abstract: The use of generative AI to create text descriptions from graphs has mostly focused on knowledge graphs, which connect concepts using facts. In this work we explore the capability of large pretrained LLMs to generate text from causal graphs, where salient concepts are represented as nodes and causality is represented via directed, typed edges. The causal reasoning encoded in these graphs can support applications as diverse as healthcare or marketing. Using two publicly available causal graph datasets, we empirically investigate the performance of four GPT-3 models under various settings. Our results indicate that while causal text descriptions improve with training data, compared to fact-based graphs, they are harder to generate under zero-shot settings. Results further suggest that users of generative AI can deploy future applications faster since similar performances are obtained when training a model with only a few examples as compared to fine-tuning via a large curated dataset.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (39)
  1. Monica Agrawal “Large language models are few-shot clinical information extractors” In Proc. 2022 Conf. on Empirical Methods in Natural Language Processing, 2022, pp. 1998–2022
  2. Imane El Atillah “Man ends his life after an AI chatbot ‘encouraged’ him to sacrifice himself to stop climate change” In Euronews.next, 2023
  3. Svetlana Bialkova “I Want to Talk to You: Chatbot Marketing Integration” In Advances in Advertising Research XII Springer, 2023, pp. 23–36
  4. J Mark Bishop “Artificial intelligence is stupid and causal reasoning will not fix it” In Frontiers in Psychology 11 Frontiers, 2021, pp. 2603
  5. Tom Brown “Language models are few-shot learners” In Adv. Neural Inf. Process. 33, 2020, pp. 1877–1901
  6. “Kgpt: Knowledge-grounded pre-training for data-to-text generation” In arXiv preprint arXiv:2010.02307, 2020
  7. Helen Pinho “Mapping Complex Systems of Population Health” In Systems Science and Population Health Oxford University Press Oxford, UK, 2017, pp. 61–76
  8. Lydia Drasic and Philippe J Giabbanelli “Exploring the interactions between physical well-being, and obesity” In Can. J. Diabetes 39 Elsevier, 2015, pp. S12–S13
  9. “How human is human evaluation? Improving the gold standard for NLG with utility theory” In arXiv preprint arXiv:2205.11930, 2022
  10. “The WebNLG challenge: Generating text from RDF data” In Proc. 10th Int. Conf. on Natural Language Generation, 2017, pp. 124–133
  11. Philippe J Giabbanelli and Chirag X Vesuvala “Human Factors in Leveraging Systems Science to Shape Public Policy for Obesity: A Usability Study” In Information 14.3 MDPI, 2023, pp. 196
  12. Philippe J Giabbanelli, Ketra L Rice and Michael C Galgoczy “Pathways to suicide or collections of vicious cycles? Understanding the complexity of suicide through causal mapping” In Social network analysis and mining 12.1 Springer, 2022, pp. 1–21
  13. Alexander Hoyle, Ana Marasović and Noah Smith “Promoting graph awareness in linearized graph-to-text generation” In arXiv preprint arXiv:2012.15793, 2020
  14. Zhiting Hu and Li Erran Li “A causal lens for controllable text generation” In Adv. Neural Inf. Process. 34, 2021, pp. 24941–24955
  15. “Neural pipeline for zero-shot data-to-text generation” In arXiv preprint arXiv:2203.16279, 2022
  16. “Causal reasoning and large language models: Opening a new frontier for causality” In arXiv preprint arXiv:2305.00050, 2023
  17. Junyi Li “Few-shot knowledge graph-to-text generation with pretrained language models” In arXiv preprint arXiv:2106.01623, 2021
  18. Wei Li “Faithfulness in natural language generation: A systematic survey of analysis, evaluation and optimization methods” In arXiv preprint arXiv:2203.05227, 2022
  19. “Guided generation of cause and effect” In arXiv preprint arXiv:2107.09846, 2021
  20. Chong Ma, Zihao Wu and Jiaqi Wang “ImpressionGPT: an iterative optimizing framework for radiology report summarization with chatGPT” In arXiv preprint arXiv:2304.08448, 2023
  21. Manuel Mager “Gpt-too: A language-model-first approach for amr-to-text generation” In arXiv preprint arXiv:2005.09123, 2020
  22. “Gpt-3 models are poor few-shot learners in the biomedical domain” In arXiv preprint arXiv:2109.02555, 2021
  23. “Participatory modeling: A methodology for engaging stakeholder knowledge and participation in social science research” In Field Methods SAGE Publications Sage CA: Los Angeles, CA, 2022, pp. 1525822X221076986
  24. “Improving language understanding by generative pre-training” OpenAI, 2018
  25. “Investigating pretrained language models for graph-to-text generation” In arXiv preprint arXiv:2007.08426, 2020
  26. Thomas Scialom “QuestEval: Summarization Asks for Fact-based Evaluation” In arXiv preprint arXiv:2103.12693, 2021
  27. Yunzhou Shi “G2T: Generating Fluent Descriptions for Knowledge Graph” In Proc. 43rd Int. ACM SIGIR Conf. on Research and Development in Information Retrieval, 2020, pp. 1861–1864
  28. “Automatically explaining a model: Using deep neural networks to generate text from causal maps” In 2022 Winter Simulation Conf. (WSC), 2022, pp. 2629–2640 IEEE
  29. Alexey Voinov “Tools and methods in participatory modeling: Selecting the right tool for the job” In Environmental Modelling & Software 109 Elsevier, 2018, pp. 232–255
  30. Bao Wang and Philippe J Giabbanelli “Identifying informative features to evaluate student knowledge as causal maps” In Int. J. Artif. Intell. Educ. Springer, 2023, pp. 1–31
  31. “Empathetic response generation through graph-based multi-hop reasoning on emotional causality” In Knowledge-Based Systems 233 Elsevier, 2021, pp. 107547
  32. “Improving PLMs for Graph-to-Text Generation by Relational Orientation Attention” In Neural Processing Letters Springer, 2023, pp. 1–17
  33. Tianming Wang, Xiaojun Wan and Hanqi Jin “Amr-to-text generation with graph transformer” In Transactions of the Association for Computational Linguistics 8 MIT Press, 2020, pp. 19–33
  34. “ASDOT: Any-Shot Data-to-Text Generation with Pretrained Language Models” In arXiv preprint arXiv:2210.04325, 2022
  35. Zhengyuan Yang “An empirical study of gpt-3 for few-shot knowledge-based vqa” In Proc. AAAI Conf. on Artificial Intelligence 36.3, 2022, pp. 3081–3089
  36. Shaowei Yao, Tianming Wang and Xiaojun Wan “Heterogeneous graph transformer for graph-to-sequence learning” In Proc. 58th Annual Meeting of the Association for Computational Linguistics, 2020, pp. 7145–7154
  37. Hong Qing Yu “Dynamic causality knowledge graph generation for supporting the chatbot healthcare system” In Proc. Future Technologies Conf. (FTC) 2020, Volume 3, 2021, pp. 30–45 Springer
  38. Cheng Zhang “Causality in the time of LLMs: Round table discussion results of CLeaR 2023” In Proc. Machine Learning Research 1, 2023, pp. 7
  39. “Bertscore: Evaluating text generation with bert” In arXiv preprint arXiv:1904.09675, 2019
Citations (3)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now