AUTOMATED ANALYSIS OF MULTIMODAL SOCIAL MEDIA DATA BASED ON MODELS FOR UNSTRUCTURED INFORMATION PROCESSING
Social media generates massive volumes of unstructured data every day. These data contain valuable information, but their heterogeneity and complexity make analysis difficult when using standard methods.The aim of the study is to develop and test an approach for automatic analysis of multimodal social media data using modern models for processing unstructured information. The work employs deep learning technologies, including transformers and models that combine different types of data (text and image).A prototype was developed that integrates textual and visual features using neural network architectures. Experiments were conducted on open datasets containing user posts with images and textual captions; video content was not included in the study. The results show that the use of multimodal models improves the accuracy of sentiment analysis and enhances data interpretation.The proposed approach can be applied in SMM analytics, marketing, user behavior prediction, and public opinion analysis. It helps automate the processing of complex data and supports decision-making based on comprehensive information.The conclusions confirm that combining modern methods of unstructured information analysis is effective for working with multimodal data in conditions of large scale and diverse sources.
Сведения
- Опубликовано
- 2026 · № 1 (69), С. 309-320
- Язык
- Английский
- ISSN
- 2309-6977 · 2958-003X · 2958-0048
- Издатель
- Non-profit limited company "Manash Kozybayev North Kazakhstan University
- Аффилиации
- META University