With the presence of a massive amount of digitally recorded data, an automated computation can be preferable over the manual approach to evaluate sentiments within given textual fragments. Artificial neural network (ANN) is preferred for sentiment analysis (SA) because of its learning ability and adaptive nature towards diverse data. Handling negation in SA is a challenging task, and to address the same, we propose a specific order of preprocessing (PPR) steps to enhance the performance of SA using ANN. Typically, ANN weights are randomly initialized (R-ANN), which may not give the desired performance. As a potential solution, we propose a novel approach named Matching features with output label based Advanced Technique (MAT) to initialize the ANN weights (MAT-ANN). Simulation results conclude the superiority of the proposed approach PPR+MAT-ANN compared to the existing approach EPR+R-ANN i.e., integrating existing preprocessing (EPR) steps with R-ANN. Moreover, PPR+MAT-ANN architecture is significantly simpler than the existing deep learning-based approach named the NeuroSent tool and gives better performance when evaluated upon the Dranziera protocol.

Clara Vania, Phu Mon Htut, William Huang, Dhara Mungra, Richard Yuanzhe Pang, Jason Phang, Haokun Liu, Kyunghyun Cho, Samuel R. Bowman. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Performance on the Winograd Schema Challenge (WSC), a respected English commonsense reasoning benchmark, recently rocketed from chance accuracy to 89% on the Super-GLUE leaderboard, with relatively little corroborating evidence of a correspondingly large improvement in reasoning ability.We hypothesize that much of this improvement comes from recent changes in task formalizationthe combination of input specification, loss function, and reuse of pretrained parametersby users of the dataset, rather than improvements in the pretrained model's reasoning ability.We perform an ablation on two Winograd Schema datasets that interpolates between the formalizations used before and after this surge, and find (i) framing the task as multiple choice improves performance by 2-6 points and (ii) several additional techniques, including the reuse of a pretrained language modeling head, can mitigate the model's extreme sensitivity to hyperparameters.We urge future benchmark creators to impose additional structure to minimize the impact of formalization decisions on reported results.