Figure 4 - uploaded by Qingqing Zhao
Content may be subject to copyright.
The heat values of the final sensory layer in SensoryT5 and the encoder layer in T5 for the sentence 'I get so mad when I see or hear about kids getting bullied...' sourced from the Empathetic Dialogues training dataset.
Source publication
In traditional research approaches, sensory perception and emotion classification have traditionally been considered separate domains. Yet, the significant influence of sensory experiences on emotional responses is undeniable. The natural language processing (NLP) community has often missed the opportunity to merge sensory knowledge with emotion cl...
Context in source publication
Context 1
... conducted a focused case study on the SensoryT5 model using a sentence from the Empathetic Dialogues dataset: "I get so mad when I see or hear about kids getting bullied..." In Figure 4, attention heatmaps display the model's focus during processing. The SensoryT5 heatmap shows the aggregate attention for each token in the sensory layer, while the T5 section compiles attention weights across all encoder layers, subsequently averaging them to reveal the model's overall focus. ...
Similar publications
Simple Summary
Seizures are one of the most common and severe symptoms of meningioma, leading to increased morbidity and mortality in the affected patients. Therefore, seizure prevention represents an important goal in the treatment of meningioma patients. For this purpose, our study aims to identify predictors for the occurrence of preoperative an...
Citations
... BERT and GPT-3 have great context awareness but their word relation preservation and being able to deal with class imbalance in real-world datasets is a hard problem. The text-to-text basis method, T5 has sufficiently been shown applicable to multilingual NLP tasks because it can generate structurally constructed responses instead of just classifying input text (9). In particular, T5 does not inherently model hierarchical dependencies in sentiment expressions, which are key to performing accurate classification. ...
As the e-commerce platforms grew exponentially, the volume of multilingual customer reviews increased, indicating that sentiment analysis is a priceless tool for finding consumer sentiment, enhancing marketing strategies, and improving customer experience. Nevertheless, emotion classification in multilingual reviews is very hard, and for one causes the variability of the language, the ambiguity of the sentiment, hierarchical word dependencies, and class imbalance which can skew traditional models. In order to resolve such challenges, this paper introduces a T5-CapsNet meta-ensembles model, which combines the T5 transformer for context embedded feature extraction and with Capsule Networks (CapsNet) for hierarchical sentiment learning. Furthermore, the model is further enhanced by a GAN-based data augmentation technique which increases the number of minority class reviews in a dataset by adding synthetic minority class reviews in an effort to correct dataset imbalance and promote classification fairness. As a meta-ensemble fusion strategy, weighted voting and stacking ensemble learning are used to improve sentiment prediction by making good use of the advantages of T5 and CapsNet. Experimental evaluations on the Multilingual Amazon Reviews Corpus (MARC) confirm that the proposed model surpasses the best sentiment classifier to reach an accuracy of 97.56% and an F1-score of 95.5%. It turns out that this hybrid deep learning approach very well captures the complex sentiment structures, or to put it differently, the multilingual e-commerce sentiment analysis is largely benefited from such a hybrid deep learning approach. The findings from this study will be a foundation for building more advanced emotion classification models, that can assist in improving customer sentiment analysis, automated feedback systems as well as decision-making in global e-commerce ecosystems.
... Extensive studies have investigated how linguistic components trigger synesthesia. Recent studies (Zhao et al. 2019, Xia et al. 2024) have delved into the mechanisms and influences underlying forms of synaesthesia, providing insights into the cognitive processes that govern sensory associations and linguistic representation. In the realm of linguistics, synaesthesia can manifest in various ways. ...
... Day (1996) posits that temperature is perceived as a distinct sensation separate from the sense of touch. In addition to the above-mentioned eight senses, sensory perception and emotion classification have traditionally been regarded as distinct fields (Xia et al., 2024). Hence, this study uses these nine senses to identify and interpret synaesthetic metaphors in Vietnamese love letters. ...
Synaesthesia has been a research interest from various disciplines, including linguistics. This study examines how synaesthetic metaphors work in the Vietnamese language and what language universality and relativity can be explained based on this. A collection of 300 love letters written by Trinh Cong Son, a well-known figure in Vietnamese music and songwriting, from 1964 to 2001, was the dataset of this study. The directionality and hierarchy of senses Ullmann (1957), Williams (1976), and Lien (1994) serve as the theoretical framework. The findings show that (1) distinct synaesthetic mappings, with dimension being the most prevalent source domain, (2) sound being the most prevalent destination domain. and (3) there are reciprocal transfers between dimension and emotion, emphasizing the dynamic interaction between sensory experience and language. The result validates Ullmann's (1957) hierarchy of senses, where lower senses act as foundational domains for higher senses. Nevertheless, significant disparities are seen in relation to prior studies. The differences can be ascribed to the variation in language and culture, highlighting the necessity for more cross-linguistic investigation to comprehensively grasp the subtleties of synaesthetic language.
