PENCARIAN CERDAS ANTAR-MODA : EVOLUSI TEKNOLOGI VIDEO-TEXT RETRIEVAL

This study aims to examine scientific trends and current approaches in the field of video-text retrieval through a bibliometric analysis approach. The urgency of this topic is driven by the surge in the growth of audiovisual data and the increasing need for a retrieval system that is able to understand the semantic relationship between text and video. Bibliometric data were collected using Publish or Perish Harzing software with Google Scholar as the main source, then visualized and analyzed using VOSviewer to identify thematic clusters, keyword distributions, and author collaboration patterns. The InternVid dataset was used as an exploration reference because it provides millions of video clips with semantically rich text annotations. The analysis results show five main clusters that illustrate thematic directions in this field, including the development of cross-modal representation models, retrieval performance evaluation, and large-scale dataset construction. Human perception-based benchmarks such as VBench are also utilized to enrich the evaluative perspective beyond quantitative metrics. This study contributes to mapping the knowledge structure of the multimodal retrieval field and opens up opportunities for the development of more contextual, adaptive, and user-oriented intelligent retrieval systems.