Jiangping Chen

Oral histories serve as vital records of lived experience, particularly within communities affected by systemic injustice and historical erasure. Effective and efficient annotation and analysis of oral history archives can promote access and use of the oral histories. However, large scale analysis of these archives remains limited due to their unstructured format, emotional complexity, and the high cost of manual annotation. This paper presents a scalable LLM-based framework to automate semantic and sentiment annotation for oral history archives, with a focus on Japanese American Incarceration Oral History (JAIOH). Using large language models (LLMs), this study seeks to construct a high-quality dataset, systematically evaluate the performance of multiple LLMs, and investigate effective prompt engineering strategies on annotation in historically sensitive contexts. Our multiphase approach combines expert annotation, prompt design, and LLM evaluation using ChatGPT, Llama, and Qwen. We labeled 558 sentences from 15 narrators for sentiment and semantic classification, then developed prompts and evaluated across zero shot, few shot, and retrieval augmented generation (RAG) strategies based on the labeled data. On this benchmark, ChatGPT achieved the highest macro-F1 score for semantic classification (88.71%), followed by Llama (84.99%) and Qwen (83.72%). For sentiment analysis, Llama performed slightly better (82.87%) than Qwen (82.66%) and ChatGPT (82.29%), with all models showing comparable results. Based on the overall evaluation, we selected the best performing prompt configurations for each task and used them to automatically annotate 92,191 sentences from 1,002 interviews in the JAIOH collection. This study develops an LLM-based framework for automatic annotation and analysis of oral history collections. The evaluation results indicate that the proposed LLM-based annotation framework provides promising performance on semantic and sentiment annotation of the JAIOH collection. It contributes a reusable pipeline and practical guidance for applying LLMs in culturally sensitive archival analysis. All code, annotated data, prompt templates, and experimental details are available at the Repository . ¹

Introduction. We propose a conceptual framework to illuminate the connections between misinformation and elder fraud, especially the role of misinformation in elder fraud. Both topics have been explored in multiple areas. The connections between misinformation and elder fraud, however, are under investigated. Also, the contributions of Library and Information Science communities are limited to help older adults avoid being victims of fraud. Method. Following an integrative literature review approach, related literature on misinformation and elder fraud were collected. Analysis. This paper applies concept analysis and thematic analysis on recent literature of misinformation and elder fraud. The analysis focuses on groups of factors that moderate or mediate misinformation to elder fraud. Results. A conceptual framework on misinformation and elder fraud is developed. The framework describes the role of misinformation in elder fraud as well as related factors, including preventive factors to misinformation, antecedents to elder fraud, mediating mechanisms between misinformation and elder fraud, and protective factors to elder fraud. Conclusion(s). There are pathways through which misinformation connects to elder fraud. Strategies to detect and mitigate misinformation may help to reduce elder fraud. More investigation is needed to refine and validate the proposed framework.

Objective: Medical Concept Normalization (MCN) maps informal health phrases to formal clinical concepts. It is a critical task for pharmacovigilance, patient record analysis, and health-related text mining. Extensive MCN research has mainly relied on single-source datasets and overlooked data quality (DQ) issues. This study aims to develop a quality-aware data fusion framework for MCN using Large Language Models (LLMs). Methods: The methods consist of DQ evaluation and enhancement, LLM-based data augmentation, and multi-source MCN dataset fusion using enriched concept-phrase pairs. We analyze six widely used MCN datasets—AskAPatient, CADEC, COMETA, PsyTAR, TwADR-S, and TwiMed—each collected from social media and mapped to SNOMED-CT. Our evaluation metrics include correctness, concept validity, coverage, semantic variety, and class imbalance. For DQ enhancement and data augmentation, we use Gemini for zero-shot and few-shot learning to increase the semantic variety and phrase count for rare concepts. We then perform data fusion based on shared medical concepts. Results: The DQ evaluation reveals substantial issues, including incorrect mappings, invalid concepts, low-variety redundant phrases, and long-tail concept-phrase distribution. After augmentation, phrase counts increase by 172.1%, and by 450.0% after fusion. To directly investigate model performance improvement on rare cases, we introduce concept-level macro metrics. SapBERT trained on the fused dataset achieves significant gains in accuracy, recall, precision, and F1 over single-source baselines, up to 30.88%.Conclusion: Our study finds that existing MCN benchmarks present data quality issues and underexplored data fusion potential. Data quality enhancement and LLM-based controlled-variety data augmentation help alleviate overlapping phrases and long-tail issues. Moreover, quality-aware data fusion can expand conceptual coverage, improving MCN performance. This work highlights data quality evaluation and fusion strategies are effective in advancing MCN. We hope these contributions could further reduce MCN dataset development costs, support more reliable biomedical text mining, and strengthen downstream applications. Availability: https://github.com/yhZHOU515/DataFusion4MCN

Exploratory search is considered challenging for users due to its open-ended nature, and ambiguity of information needs. Clarifying exploratory searchers’ information needs is fundamental to design information environments that provide adaptive support. Since tasks are a representation of information needs, this study investigates the characteristics of exploratory tasks in different fields through multiple attributes to reveal exploratory searchers’ information needs. Repeated surveys were conducted in three representative fields: healthcare, artificial intelligence, and integrated circuit. Content analysis was used to identify 16 attributes. Learning and investigation differ in 14 attributes, and exploratory subtasks differ significantly in Source, Domain knowledge, Cognitive complexity and Search frequency . Different composition of user knowledge background leads to differences in two attributes of Motivation and Cognitive complexity , and difference in the correlation between Motivation and Interdependence . The findings help researchers design distinct exploratory tasks with domain characteristics, and help practitioners develop task- and domain-adaptive exploratory search systems.

Purpose: This paper aims to understand the characteristics of current misinformation detection studies, including the datasets used by researchers, the computational models or algorithms being developed or applied, and the performance of misinformation detection models or algorithms.Design/methodology/approach: We first identified articles from the Scopus database with inclusion and exclusion criteria. Then a coding scheme was derived from the articles based on research questions. Next, datasets, models, and performance were coded. The paper concluded with answers to research questions and future research directions.Findings: From 115 relevant articles published during 2019–2023 on misinformation detection. We found that most studies used previously existing datasets. Twitter (now X) has been the most widely used source for collecting social media misinformation data. The ten most frequently used datasets are identified. Most studies (96.1%) developed or applied machine learning, especially deep learning models. The most advanced current misinformation detection models could achieve pretty high performance. For example, among 104 studies reporting performance with accuracy, 44.2% achieved an accuracy of 0.95 or higher, and 24.0% achieved 0.90–0.94 on accuracy.Research limitations/implications: Our study only reviewed English articles from 2019–2023 that are included in the Scopus database. Articles that are not included in the Scopus database are not reviewed.Practical implications: The high performance of misinformation detection indicates that social media should be able to detect most misinformation if they are willing to do it. However, no system or algorithm could achieve 100% misinformation on performance. Due to the complexity of misinformation, users of social media still need to improve their capabilities of evaluating information on the Internet.Social implications: This study provides evidence to policymakers that social media platforms have the capability of detecting most misinformation posted. These platforms are responsible for alerting to suspicious postings with misinformation.Originality/value: This study identifies datasets, computer models, and performance of models from current misinformation detection research. The findings will help social media companies, computer scientists, and information system designers improve their misinformation detection systems. It will also help students in information science and computer science to study the latest models and algorithms. Information professionals may work with computer scientists to improve datasets used for misinformation detection.