In recent years, there has been a renewed interest in Emotion-Cause Analysis (ECA) related Natural Language Processing (NLP) tasks. Most of these are classification tasks in nature with little to no emphasis on the rapidly advancing field of text generation with machine learning models. In this paper, we propose a new generative task within the ECA domain named ECSGen (Emotion-Cause Mitigating Suggestion Generation). The task is to generate relevant suggestions to mitigate the cause of a negative emotion expressed in a given text. We propose iZen, a technical framework that leverages large language model (LLM) to perform this task in a zero-shot manner without requiring any new training or fine-tuning steps. We curate two new datasets to evaluate iZen's ability to perform the ECSGen task. Our experiments and analysis demonstrate iZen's promising performance in the ECSGen task.

Emotion Cause Analysis (ECA) has recently garnered substantial attention from the researcher community. In addition to devising various techniques to solve ECA related problems, researchers also introduced different variants of the ECA tasks such as Emotion Cause Extraction (ECE), Emotion Cause Pair Extraction (ECPE), Emotion Cause Span Extraction (ECSE). These are primarily classification tasks where the cause of the emotion and/or type of the emotion expressed in the text are identified. In this paper, we propose a new ECA related task named Emotion Cause Generation (ECG). This is a generative task that aims to generate meaningful cause for an emotion expressed in a given text. We demonstrate the viability of this newly proposed task with promising early observation.

Analyzing public sentiment about an entity or issue can be of interest to governments and businesses alike. There is a growing body of research that attempt to devise new sentiment analysis techniques, especially techniques based on machine learning. These machine learning-based techniques typically require large, labeled training data with a large number of instances for training in order to provide reasonable accuracy in sentiment analysis. However, labelling large volumes of data is tedious and expensive. In this paper, we propose a multi-class sentiment analysis technique, named SG-Elect, utilizing cutting-edge transformer based pre-trained models along with more traditional machine learning based approaches in a semi-supervised setting. Our experiments demonstrate that SG-Elect outperforms a recent state-of-the-art baseline for all three datasets.

Smartphones are becoming larger, mainly because bigger screens offer a better experience for viewing content. One drawback of larger screens is that they make single-hand interactions difficult because of hard to reach touch targets and of the need to re-grip the device, both factors significantly reducing their usability. Flexible smartphones offer an opportunity for addressing this issue. We first set out to determine the use of common single-hand mobile interactions through an online survey. Then, we designed and evaluated one-handed deformable gestures that offer the potential for addressing the finger reach limitation on large smartphones. We identified that the top right up bend and the center squeeze up gestures are the fastest and preferred gestures. We found no hand preference, which indicates that the gestures could be implemented to fit the needs of a wider range of the population, instead of favoring right-handed users. Finally, we discuss the impact on deformable gestures on one-handed interactions issues.