b'@online{Singhania2411.12449,'b'\nTITLE = {Neon: News Entity-Interaction Extraction for Enhanced Question Answering},\nAUTHOR = {Singhania, Sneha and Cucerzan, Silviu and Herring, Allen and Jauhar, Sujay Kumar},\nLANGUAGE = {eng},\nURL = {https://arxiv.org/abs/2411.12449},\nEPRINT = {2411.12449},\nEPRINTTYPE = {arXiv},\nYEAR = {2024},\nMARGINALMARK = {$\\bullet$},\nABSTRACT = {Capturing fresh information in near real-time and using it to augment<br>existing large language models (LLMs) is essential to generate up-to-date,<br>grounded, and reliable output. This problem becomes particularly challenging<br>when LLMs are used for informational tasks in rapidly evolving fields, such as<br>Web search related to recent or unfolding events involving entities, where<br>generating temporally relevant responses requires access to up-to-the-hour news<br>sources. However, the information modeled by the parametric memory of LLMs is<br>often outdated, and Web results from prototypical retrieval systems may fail to<br>capture the latest relevant information and struggle to handle conflicting<br>reports in evolving news. To address this challenge, we present the NEON<br>framework, designed to extract emerging entity interactions -- such as events<br>or activities -- as described in news articles. NEON constructs an<br>entity-centric timestamped knowledge graph that captures such interactions,<br>thereby facilitating enhanced QA capabilities related to news events. Our<br>framework innovates by integrating open Information Extraction (openIE) style<br>tuples into LLMs to enable in-context retrieval-augmented generation. This<br>integration demonstrates substantial improvements in QA performance when<br>tackling temporal, entity-centric search queries. Through NEON, LLMs can<br>deliver more accurate, reliable, and up-to-date responses.<br>},\n}\n'