Abstract Cardinality estimation is a crucial component in query optimizers. After decades of research, employing autoregressive models for cardinality estimation has demonstrated remarkable accuracy. However, when queries involve attributes with large domain sizes, autoregressive model-based estimators struggle to accurately capture the data distribution, leading to poor performance. Furthermore, these models often exhibit significant errors when handling queries with low-selectivity predicates. To address these challenges, we propose a self-adaptive cardinality estimator named AdaCard. Initially, we employ a self-adaptive smoothing factor selection strategy to variably adjust the original data, thereby mitigating the impact of large domain sizes. Secondly, to correct errors stemming from Monte Carlo sampling, we utilize resampling to refine the handling of low-selectivity predicates, thereby improving accuracy. Through evaluation using four real-world benchmarks, we compared AdaCard with mainstream baselines. The final results show that our estimator has the lowest tail estimation error and improves accuracy by nearly 10$\times $ over the second-best method, with similar latency and model size.

Aspect sentiment triplet extraction (ASTE) is an emerging sentiment analysis task that aims to extract sentiment triplets from review sentences. Each sentiment triplet consists of an aspect, corresponding opinion, and sentiment. Although extensive research has been conducted on the ASTE task, existing methods use the span representations to predict the relationship between spans, failing to consider the interrelation between span pairs. On the other hand, early fusion of entity information is critical for sentiment classification. In this paper, we propose an Enhanced Packed Marker with Entity Information (EPMEI) framework for ASTE task to address the above limitations of the existing works. Specifically, EPMEI consists of entity recognition and sentiment classification models. The entity information is obtained from the entity recognition model first. After that, we insert solid markers with entity information at the input layer of the sentiment classification model to highlight the subject span and improve subject span representation. Furthermore, we introduce a subject-oriented packing strategy, which packs each subject span and all its levitated markers of object spans to model the interrelation between the same-subject span pairs. Extensive experimental results on four ASTE benchmark datasets demonstrate that EPMEI achieves the state-of-the-art baseline. Our code can be found in https://github.com/MKMaS-GUET/EPMEI.