计算机科学
阿拉伯语
主题模型
自然语言处理
聚类分析
情报检索
层次分析法
背景(考古学)
人工智能
工作流程
变压器
现代标准阿拉伯语
过程(计算)
层次聚类
数据科学
清晰
质量(理念)
万维网
文档聚类
用户故事
预印本
出处
期刊:PeerJ
[PeerJ, Inc.]
日期:2026-05-01
卷期号:12: e3780-e3780
被引量:1
标识
DOI:10.7717/peerj-cs.3780
摘要
Background Female Technology (FemTech) apps are increasingly used worldwide, but perspectives from Arabic-speaking users remain underreported. Understanding user priorities in this context is critical for culturally relevant app development and health policy. Objective To identify and prioritize user-reported themes—both concerns and positive perspectives—in Arabic-language reviews of FemTech apps across the Middle East and North Africa (MENA) region. Methods FemTech apps were systematically identified across 16 Arabic-speaking countries in Google Play and Apple App Store. Public user reviews were collected, cleaned, and normalized from dialectal Arabic to Modern Standard Arabic (MSA) using a prompt-controlled workflow with manual verification. Sentiment was classified using a transformer-based Arabic model. Topics were extracted via Bidirectional Encoder Representations from Transformers Topic Modeling (BERTopic), using Arabic Bidirectional Encoder Representations from Transformers (AraBERTv2) embeddings and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) clustering, then coded by three experts with substantial inter-rater agreement. Themes were prioritized using Analytic Hierarchy Process (AHP) to weight four criteria—frequency, user importance strength, recency, and app-version spread—and ranked via Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). Results After screening, 51 apps (37 Android, 14 iOS) and 26,151 Arabic reviews were included. Sentiment aligned with star ratings. Topic modeling yielded 50 topics grouped into eight themes. Frequency (0.380) and user importance (0.344) received the highest AHP weights. Positive perspectives—widespread user approval and easy-to-use tracking—ranked highest overall, while poor Arabic language support emerged as the leading concern. Mid-ranked themes addressed content quality and long-term engagement; lower-ranked concerns involved monetization barriers and prediction accuracy. Conclusions Arabic reviews show predominantly positive perspectives, yet language support remain significant concerns. The proposed pipeline has demonstrated a reproducible way to mine app-store reviews in underrepresented languages, offering actionable priorities for developers, researchers, and policymakers.
科研通智能强力驱动
Strongly Powered by AbleSci AI