A Hybrid NLP Framework for Topic Relevance and Advertisement Detection in Trending Hashtags on X (Twitter)

Trending hashtags on X (formerly Twitter) are widely used for real-time news tracking and public discourse, yet they are increasingly polluted by advertisements and off-topic posts that exploit hashtag visibility. This study proposes a hybrid natural language processing (NLP) framework for improving the reliability of hashtag streams through two complementary modules: 1) semantic topic relevance detection and 2) supervised advertisement classification. Using 9,671 Arabic tweets retained in the six final research datasets collected via the X API from multiple trending hashtags spanning diverse domains, we apply a dedicated Arabic preprocessing pipeline including noise removal and normalization. Topic relevance is assessed using AraBERT contextual embeddings by computing cosine similarity between each tweet and a concise hashtag topic description, followed by threshold-based labeling of on-topic versus off-topic content. In parallel, advertisement detection is formulated as a binary classification task using TF–IDF features and standard machine learning models (Logistic Regression, SVM, Random Forest, Naive Bayes, and SGD). Experimental results show that advertisement detection performance varies by hashtag and topic type: technical hashtags yield stronger and more stable F1-scores, while broad event hashtags are more challenging due to heterogeneous language and content. Across multiple hashtags, Random Forest and linear classifiers provide the most consistent trade-off between precision and recall. The proposed framework enables fine-grained quantification of promotional intrusion and semantic drift within hashtag discussions, providing a practical foundation for trustworthy hashtag analytics, improved trend monitoring, and downstream social media analysis in Arabic settings.

@article{Alshammari2026,
  title     = {A Hybrid NLP Framework for Topic Relevance and Advertisement Detection in Trending Hashtags on X (Twitter)},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {17},
  number    = {9},
  year      = {2026},
  publisher = {The Science and Information Organization},
  author    = {Aadil Alshammari and Khulud Alashjaee and Balsam Alzamam and Aryam Alshammari and Milaf Alshammari and Mohamed Ayari},
  doi       = {10.14569/IJACSA.2026.0170987},
  url       = {https://doi.org/10.14569/IJACSA.2026.0170987}
}