Fake News Detection via an Adaptive Feature Matching Optimization Framework
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: Multimodal, fake news detection, simulated annealing, explainable AI, adaptive optimization
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: Adaptive Feature Matching Optimization
Abstract: The rampant proliferation of fake news across online platforms has become a significant cause for concern, necessitating the creation of robust detection techniques. Within the confines of this investigation, we present an optimization methodology built upon salient attributes tailored for the identification of fake news, spanning both unimodal and multimodal data sources. By harnessing the capabilities inherent in a diverse array of modalities, ranging from textual to visual elements, we are able to comprehensively apprehend the multifaceted nature of falsified news stories. Primarily, our methodology introduces an unprecedented array of features, encompassing word-level, sentence-level, and contextual features. This infusion bestows upon it a robust capacity to adeptly accommodate a wide spectrum of textual content. Subsequently, we integrate a feature-centric optimization technique grounded in the principles of simulated annealing. This approach enables us to ascertain the most optimal fusion of features, thereby mitigating potential conflicts and interferences arising from the coexistence of textual and visual components. Empirical insights garnered from exhaustive dataset experimentation decisively underscore the efficacy of our proposed methodology. Our approach outperforms standalone modalities as well as traditional single-classifier models, as evidenced by its superior detection capabilities. This research underscores the indispensable role played by the integration of multimodal data sources and the meticulous optimization of feature amalgamations. These factors collectively contribute to the creation of a resilient framework tailored for the identification of fake news within the intricate landscape of our contemporary, data-rich environment.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7754
Loading