FusionMind -- Improving question and answering with external context fusion (2401.00388v1)
Abstract: Answering questions using pre-trained LLMs (LMs) and knowledge graphs (KGs) presents challenges in identifying relevant knowledge and performing joint reasoning.We compared LMs (fine-tuned for the task) with the previously published QAGNN method for the Question-answering (QA) objective and further measured the impact of additional factual context on the QAGNN performance. The QAGNN method employs LMs to encode QA context and estimate KG node importance, and effectively update the question choice entity representations using Graph Neural Networks (GNNs). We further experimented with enhancing the QA context encoding by incorporating relevant knowledge facts for the question stem. The models are trained on the OpenbookQA dataset, which contains ~6000 4-way multiple choice questions and is widely used as a benchmark for QA tasks. Through our experimentation, we found that incorporating knowledge facts context led to a significant improvement in performance. In contrast, the addition of knowledge graphs to LLMs resulted in only a modest increase. This suggests that the integration of contextual knowledge facts may be more impactful for enhancing question answering performance compared to solely adding knowledge graphs.
- “EDM3: Event Detection as Multi-task Text Generation” In ArXiv abs/2305.16357, 2023 URL: https://api.semanticscholar.org/CorpusID:258947208
- “Instruction Tuned Models are Quick Learners” In ArXiv abs/2306.05539, 2023 URL: https://api.semanticscholar.org/CorpusID:259129868
- “TarGEN: Targeted Data Generation with Large Language Models” In ArXiv abs/2310.17876, 2023 URL: https://api.semanticscholar.org/CorpusID:264555527
- “Context-NER : Contextual Phrase Generation at Scale”, 2021 URL: https://api.semanticscholar.org/CorpusID:265039305
- “KagNet: Knowledge-Aware Graph Networks for Commonsense Reasoning” In CoRR abs/1909.02151, 2019 arXiv: http://arxiv.org/abs/1909.02151
- “RoBERTa: A Robustly Optimized BERT Pretraining Approach” In CoRR abs/1907.11692, 2019 arXiv: http://arxiv.org/abs/1907.11692
- “Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering”, 2018 arXiv:1809.02789 [cs.CL]
- “LongBoX: Evaluating Transformers on Long-Sequence Clinical Tasks” In ArXiv abs/2311.09564, 2023 URL: https://api.semanticscholar.org/CorpusID:265221168
- “Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer” In CoRR abs/1910.10683, 2019 arXiv: http://arxiv.org/abs/1910.10683
- “SQuAD: 100,000+ Questions for Machine Comprehension of Text”, 2016 arXiv:1606.05250 [cs.CL]
- “InstructABSA: Instruction Learning for Aspect Based Sentiment Analysis” In ArXiv abs/2302.08624, 2023 URL: https://api.semanticscholar.org/CorpusID:257020097
- Robyn Speer, Joshua Chin and Catherine Havasi “ConceptNet 5.5: An Open Multilingual Graph of General Knowledge”, 2018 arXiv:1612.03975 [cs.CL]
- “Improving Natural Language Inference Using External Knowledge in the Science Questions Domain” In CoRR abs/1809.05724, 2018 arXiv: http://arxiv.org/abs/1809.05724
- Zhen Wang “Modern Question Answering Datasets and Benchmarks: A Survey”, 2022 arXiv:2206.15030 [cs.CL]
- “Fusing Context Into Knowledge Graph for Commonsense Reasoning” In CoRR abs/2012.04808, 2020 arXiv: https://arxiv.org/abs/2012.04808
- “QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering” In CoRR abs/2104.06378, 2021 arXiv: https://arxiv.org/abs/2104.06378
- “Retrieving and Reading: A Comprehensive Survey on Open-domain Question Answering”, 2021 arXiv:2101.00774 [cs.AI]
- Shreyas Verma (7 papers)
- Manoj Parmar (2 papers)
- Palash Choudhary (2 papers)
- Sanchita Porwal (1 paper)