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

Area of Science:

• Computer Science
• Information Retrieval
• Internet of Things

Background:

• Rapid growth of Internet of Things (IoT) deployments creates challenges in discovering smart devices within heterogeneous environments.
• Conventional IoT search methods struggle with diverse device types, locations, statuses, metadata, and interfaces, limiting user intent interpretation in Web of Things (WoT) contexts.

Purpose of the Study:

• To present an intent-based Web of Things (WoT) query framework for enhanced smart device discovery.
• To convert natural-language queries into structured search constraints by extracting key device attributes like type, operational status, and geographic location.

Main Methods:

• Developed a prototype framework utilizing Flask, TextBlob, Google Maps Geocoding API, and Apache Solr.
• Implemented modular components for user interaction, intent extraction, geospatial resolution, query construction, retrieval, and multi-factor ranking.
• Evaluated the framework using an intent-robustness benchmark, a large metadata corpus with noise, and external standards-backed validation.

Main Results:

• Achieved high intent-extraction accuracy (0.968) and F1 score for noisy metadata retrieval (0.978).
• Demonstrated superior performance in retrieval (NDCG@10 of 0.950) compared to baseline methods like keyword matching and TF-IDF.
• External validation showed a query success rate of 0.946 and NDCG@10 of 0.905.

Conclusions:

• The intent-based WoT query framework significantly improves smart device discovery in complex IoT environments.
• The system effectively interprets user intent from natural language, outperforming conventional search approaches.
• Future work includes addressing spelling sensitivity and integrating the full ranking module into the live interface.