Researchers have developed a new method for constructing annotated datasets for fine-grained aspect-based sentiment analysis, specifically for Korean e-commerce reviews. This approach, termed Semi-Automatic Symbolic Propagation (SSP), utilizes extensive linguistic resources formalized as Finite-State Transducers. The extended ABSA framework incorporates aspect values alongside topics and aspects, classifying them based on whether values are unary, binary, or multiple. When tested with KoBERT and KcBERT models, the dataset demonstrated strong performance, achieving F1 scores of 0.88 and 0.90, respectively, in recognizing aspect-value pairs. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Introduces a novel methodology for creating specialized datasets that can improve the accuracy of sentiment analysis models in specific domains.
RANK_REASON The cluster contains an academic paper detailing a new method for dataset construction and its application in sentiment analysis. [lever_c_demoted from research: ic=1 ai=1.0]