Compared with single-modal content, multimodal data can express users' feelings and sentiments more vividly interestingly. Therefore, sentiment analysis has become a popular research topic. However, most existing methods either learn modal feature independently, without considering their correlations, or they simply integrate features. In addition, publicly available datasets are labeled by polarities, while the emotions expressed users specific. Based on this observation, in paper, we build...

Emotion cause analysis has been a key topic in natural language processing. Existing methods ignore the contexts around emotion word which can provide an clue. Meanwhile, clauses document play different roles on stimulating certain emotion, depending their content relevance. Therefore, we propose co-attention neural network model for with emotional context awareness. The method encodes based bi-directional long short-term memory into high-level input representations, are further fed...

Xiaocui Yang, Shi Feng, Yifei Zhang, Daling Wang. Proceedings of the 59th Annual Meeting Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

In the realm of time series forecasting (TSF), it is imperative for models to adeptly discern and distill hidden patterns within historical data forecast future states. Transformer-based exhibit formidable efficacy in TSF, primarily attributed their advantage apprehending these patterns. However, quadratic complexity Transformer leads low computational efficiency high costs, which somewhat hinders deployment TSF model real-world scenarios. Recently, Mamba, a selective state space model, has...

The incompleteness of knowledge graphs triggers considerable research interest in relation prediction. As the key to predicting relations among entities, many efforts have been devoted learning embeddings entities and by incorporating a variety neighbors' information which includes not only from direct outgoing incoming neighbors but also ones indirect on multihop paths. However, previous models usually consider entity paths limited length or ignore sequential Either simplification will make...

Existing visual perception systems focus on region-level segmentation in single-turn dialogues, relying complex and explicit query instructions. Such cannot reason at the pixel level comprehend dynamic user intent that changes over interaction. Our work tackles this issue by introducing a novel task, Pixel-level Reasoning Segmentation (Pixel-level RS) based multi-turn conversations, tracking evolving via interactions for fine-grained segmentation. To establish benchmark we build ReasonIng...

Multimodal sentiment analysis has gained significant attention due to the proliferation of multimodal content on social media. However, existing studies in this area rely heavily large-scale supervised data, which is time-consuming and labor-intensive collect. Thus, there a need address challenge few-shot analysis. To tackle problem, we propose novel method called Probabilistic Fusion Prompts (MultiPoint) that leverages diverse cues from different modalities for detection scenario....

We have witnessed the rapid proliferation of multimodal data on numerous social media platforms. Conventional studies typically require massive labeled to train models for Multimodal Aspect-Based Sentiment Analysis (MABSA). However, collecting and annotating fine-grained MABSA is tough. To alleviate above issue, we perform three MABSA-related tasks with quite a small number samples. first build diverse comprehensive few-shot datasets according distribution. capture specific prompt each...

Extensive experiments have validated the effectiveness of corpus-based method for classifying word’s sentiment polarity. However, no work is done comparing different corpora in polarity classification task. Nowadays, Twitter has aggregated huge amount data that are full people’s sentiments. In this paper, we empirically evaluate performance similarity measurement, which fundamental task word classification. Experiment results show can achieve a much better than Google, Web1T and Wikipedia...

Review rating prediction is an important research topic. The problem was approached from either the perspective of recommender systems (RS) or that sentiment analysis (SA). Recent SA using deep neural networks (DNNs) has realized importance user and product interaction for better interpreting reviews. However, complexity DNN models in terms scale parameters very high, performance not always satisfying especially when user-product sparse. In this paper, we propose a simple, extensible...

Sentiment expression in microblog posts can be affected by user's personal character, opinion bias, political stance and so on. Most of existing personalized sentiment classification methods suffer from the insufficiency discriminative tweets for personalization learning. We observed that users have consistent individuality bias different languages. Based on this observation, paper we propose a novel user-attention-based Convolutional Neural Network (CNN) model with adversarial cross-lingual...

Emotion cause analysis (ECA) has been an emerging topic in natural language processing, which aims to identify the reasons behind a certain emotion expressed text.Most ECA methods intend clause contains of given emotion, but such clause-level (CECA) can be ambiguous and imprecise.In this paper, we aim at span-level (SECA) by detecting precise boundaries text spans conveying accurate causes from context.We formulate task as sequence labeling position identification problems design two neural...

With the development of social networks and intelligent terminals, it is becoming more convenient to share acquire images. The massive growth number images makes people have higher demands for automatic image processing, especially in aesthetic emotional perspective. Both aesthetics assessment emotion recognition require a ability computer simulate high-level visual perception understanding, which belongs field processing pattern recognition. However, existing methods often ignore prior...

As an indispensable resource for emotion analysis, lexicons have attracted increasing attention in recent years. Most existing methods focus on capturing the single emotional effect of words rather than distributions which are helpful to model multiple complex emotions a subjective text. Meanwhile, automatic lexicon building overly dependent seed but neglect emoticons natural graphical labels fine-grained emotion. In this paper, we propose novel framework that leverages both and...