Abstract
Large image collections are typically organized around basic metadata and keyword tags, making content discovery challenging for users seeking specific visual information. Although images may be accompanied by descriptive text, traditional retrieval systems often struggle to bridge the semantic gap between textual descriptions and visual content. In this demo, we present ImageSeek, a hybrid text-to-image retrieval system designed to enhance search effectiveness by combining text and image-based retrieval methods through an asymmetric score adjustment mechanism. The system leverages multilingual CLIP models to encode both visual and textual information, creating unified representations for cross-modal retrieval. Users can search through natural language queries in any supported language, with results ranked using a hybrid approach that treats image-based retrieval as a reliable baseline while harmonizing text-based scores through position-dependent adjustments. The demonstration system operates on a dataset of 42,333 images from the Portuguese Presidency website, providing an appropriate testbed for multimodal retrieval performance. The web application enables direct comparison between conventional CLIP-based retrieval and our hybrid approach, supporting image searches under the same conditions on external platforms, including Google Images and the Arquivo.pt image search system, enabling comparative analysis of the results. To evaluate its effectiveness, ImageSeek allows users to experience differences between retrieval modes while exploring domain-specific visual content.

Abstract

Abstract

Abstract

Abstract
[No abstract available]

Abstract
The authors regret that a minor inconsistency was identified in Algorithm 1 of our published paper during subsequent experiments conducted to further improve the G16 constructive heuristic. Specifically, the original implementation of G16 did not distinguish between regular and merged earliest time windows when computing [Formula presented], which could, in some cases, affect the consistency of [Formula presented], [Formula presented], [Formula presented], [Formula presented], and [Formula presented] for requests simultaneously served at the same location, leading to infeasible routes under specific configurations. The correction is as follows (Algorithm 1, Line 15): Original: [Formula presented] Corrected: [Formula presented] where [Formula presented] denotes the merged earliest time window when a merged time service is applied; otherwise, it equals [Formula presented]. As a result, the total costs obtained with the corrected version of G16 slightly deviate, either positively or negatively, from those originally published. The average percentage gaps between the published and corrected G16 results are 2.55, 0.61, -0.64, 0.42, and -2.27% for instances with 50, 100, 200, 400, and 700 requests, respectively. Complementarily, a Spearman correlation (p = 0.98) and a Wilcoxon signed-rank test (p = 0.106) revealed no statistically significant difference between both sets of results. Therefore, the overall performance patterns and comparative findings discussed in the original paper remain valid. Updated computational results are available in the same Mendeley Data repository (DOI: https:/doi.org/10.17632/7fb9jn2wcs.1). The authors would like to apologise for any inconvenience caused. © 2025 Elsevier B.V.