Scholarlink

<p>The rapid growth of Internet of Things (IoT) deployments has made smart device discovery difficult in heterogeneous environments where devices differ in type, location, operational status, metadata structure, and communication interface. Conventional IoT search approaches often rely on keyword matching, static indexing, or predefined metadata fields, which limit their ability to interpret user intent in Web of Things (WoT) environments. This paper presents an intent-based WoT query framework for smart device discovery. The framework converts natural-language queries into structured search constraints by extracting device type, operational status, and geographic location. The prototype uses Flask, TextBlob for lightweight preprocessing and tokenization, Google Maps Geocoding API, and Apache Solr, with modular components for user interaction, robust intent extraction, geospatial resolution, query construction, retrieval, and multi-factor ranking. The evaluation was extended using a 380-query intent-robustness benchmark, a 10,000-record enriched WoT-like metadata corpus with controlled noise conditions, multi-factor ranking ablation, stronger lexical and semantic baselines, and external standards-backed validation on combined W3C Thing Description 1.1-style and OGC SensorThings-style metadata. Results show overall intent-extraction accuracy of 0.968, normalized noisy-metadata retrieval F1 of 0.978, NDCG@10 of 0.950 for full multi-factor ranking versus 0.446 for timestamp-only sorting, and clear superiority over keyword, BM25, fuzzy lexical, and TF-IDF baselines. External validation reached query success rate 0.946 and NDCG@10 0.905 on 55 labelled queries over a reproducible standards-backed corpus. Remaining limitations include spelling-error sensitivity, synthetic data constraints, the modest size of the external standards corpus, and the need to integrate the full ranking module into the live Flask/Solr interface.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design and evaluation of an intent-based web of things query framework for smart device discovery

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月9日

查看全文
  • Salman Tahir,
  • Najia Saher,
  • Muhammad Rehan Faheem,
  • Muhammad Nabeel Asghar,
  • Abid Iqbal,
  • Saad Arif

摘要

The rapid growth of Internet of Things (IoT) deployments has made smart device discovery difficult in heterogeneous environments where devices differ in type, location, operational status, metadata structure, and communication interface. Conventional IoT search approaches often rely on keyword matching, static indexing, or predefined metadata fields, which limit their ability to interpret user intent in Web of Things (WoT) environments. This paper presents an intent-based WoT query framework for smart device discovery. The framework converts natural-language queries into structured search constraints by extracting device type, operational status, and geographic location. The prototype uses Flask, TextBlob for lightweight preprocessing and tokenization, Google Maps Geocoding API, and Apache Solr, with modular components for user interaction, robust intent extraction, geospatial resolution, query construction, retrieval, and multi-factor ranking. The evaluation was extended using a 380-query intent-robustness benchmark, a 10,000-record enriched WoT-like metadata corpus with controlled noise conditions, multi-factor ranking ablation, stronger lexical and semantic baselines, and external standards-backed validation on combined W3C Thing Description 1.1-style and OGC SensorThings-style metadata. Results show overall intent-extraction accuracy of 0.968, normalized noisy-metadata retrieval F1 of 0.978, NDCG@10 of 0.950 for full multi-factor ranking versus 0.446 for timestamp-only sorting, and clear superiority over keyword, BM25, fuzzy lexical, and TF-IDF baselines. External validation reached query success rate 0.946 and NDCG@10 0.905 on 55 labelled queries over a reproducible standards-backed corpus. Remaining limitations include spelling-error sensitivity, synthetic data constraints, the modest size of the external standards corpus, and the need to integrate the full ranking module into the live Flask/Solr interface.