Uzbek language Aspect Based Sentiment Analysis

We evaluated our proposed methodology on a comprehensive Aspect-Based Sentiment Analysis (ABSA) dataset consisting of restaurant reviews in the Uzbek language. The dataset focuses on two primary sub-tasks: Aspect Term Extraction (ATE) and Aspect Category Sentiment Analysis (ACSA).

For the ATE task, the dataset is annotated using the BIO (Beginning, Inside, Outside) tagging scheme. As seen in the training samples (e.g., \texttt{test.json}), tokens representing aspect terms are labeled as \texttt{B-ASP} or \texttt{I-ASP}, while non-aspect tokens are labeled \texttt{O}.

For the ACSA task, we utilized standard and ``hard'' test sets (e.g., \texttt{acsa\_hard\_test.json}). Each entry consists of a sentence, a specific category (e.g., \textit{ovqat} (food), \textit{xizmat} (service)), and a sentiment label (Positive, Negative, Neutral, or Conflict).