Search

As online reviews play a crucial role in purchasing decisions, businesses are increasingly incentivized to generate positive reviews, sometimes resorting to fake reviews or opinion spam. Detecting opinion spam requires well-trained models, but obtaining annotated training data in the same domain (e.g., hotels) can be challenging. Transfer learning addresses this by leveraging training data from a similar domain (e.g., restaurants). This paper examines three popular transformer models—BERT, RoBERTa, and DistilBERT—to evaluate how training data from different domains, including imbalanced datasets, impacts Transformer model performance. Notably, our evaluation of hotel opinion spam detection achieved an AUC of 0.927 using RoBERTa trained on YelpChi restaurant data.

Hyperledger Fabric is an open-source project designed to deploy permissioned blockchains. Performance characteristics of Hyperledger Fabric v2.2, such as the impact of ordering services, bottleneck, and scalability) are difficult to understand due to the performance challenges of distributed systems. In this paper, we evaluate the performance of each phase in Hyperledger Fabric's new execute-order-validate architecture. We also assessed the performance of ordering services (Solo, Kafka, and Raft) on the OR and the AND endorsement policies. We found the execution phase was more scalable using the OR endorsement policy than the AND endorsement policy. While all three ordering services (Solo, Raft, and Kafka) performed relatively well, we discuss why Raft may be the best choice for most organizations. Last, we focus on the performance capabilities of the Raft ordering service using different transaction settings. This evaluation helps to understand some of the trade-offs in Hyperledger Fabric v2.x.