A Dual-Query Reformulation Architecture for Temporal and Non-Temporal Information Retrieval

With the exponential growth of data on the World Wide Web, retrieving relevant information tailored to specific users' temporal and non-temporal needs has become increasingly challenging.This research advances the field of information retrieval by introducing a dualarchitecture system capable of addressing both temporal and non-temporal queries.In addition to handling traditional text-based queries, this study incorporates user speech, enhancing the accessibility and versatility of search interfaces.For temporal queries, the architecture retrieves web pages, trending hashtags, and tweets from designated sources.Contextual terms extracted from these tweets are integrated with the users' original query to form a 'Reformulated query'.By leveraging real-time data from social media trends and hashtags, the system dynamically reformulates temporal queries, ensuring the timely delivery of relevant information to users.Non-temporal queries are addressed using Enhanced Particle Swarm Optimization combined with pre-trained sentence embeddings.This novel optimization approach enhances retrieval accuracy across various content types.The system captures the contextual aspects of queries and documents using pre-trained sentence embedding models, ensuring an advanced level of context-aware retrieval for time-sensitive speech and text queries.Experimental results demonstrate an overall 23.07%improvement in MAP@K, 15.80 % in MRR, and 20.65% in NDCG@K compared to the baseline models.These findings highlight the system's potential for commercial applications, including search assistance systems, voice-enabled applications for mobile and smart devices, application tracking systems, and information collection across various domains.