Recent years have witnessed the rise and success of pre-training techniques in visually-rich document understanding. However, most existing methods lack the systematic mining and utilization of layout-centered knowledge, leading to sub-optimal performances. In this paper, we propose ERNIE-Layout, a novel document pre-training solution with layout knowledge enhancement in the whole workflow, to learn better representations that combine the features from text, layout, and image. Specifically, we first rearrange input sequences in the serialization stage, and then present a correlative pre-training task, reading order prediction, to learn the proper reading order of documents. To improve the layout awareness of the model, we integrate a spatial-aware disentangled attention into the multi-modal transformer and a replaced regions prediction task into the pre-training phase. Experimental results show that ERNIE-Layout achieves superior performance on various downstream tasks, setting new state-of-the-art on key information extraction, document image classification, and document question answering datasets. The code and models are publicly available at http://github.com/PaddlePaddle/PaddleNLP/tree/develop/model_zoo/ernie-layout.

Automatically generated emotion arcs -- that capture how an individual or a population feels over time -- are widely used in industry and research. However, there is little work on evaluating the generated arcs. This is in part due to the difficulty of establishing the true (gold) emotion arc. Our work, for the first time, systematically and quantitatively evaluates automatically generated emotion arcs. We also compare two common ways of generating emotion arcs: Machine-Learning (ML) models and Lexicon-Only (LexO) methods. Using a number of diverse datasets, we systematically study the relationship between the quality of an emotion lexicon and the quality of the emotion arc that can be generated with it. We also study the relationship between the quality of an instance-level emotion detection system (say from an ML model) and the quality of emotion arcs that can be generated with it. We show that despite being markedly poor at instance level, LexO methods are highly accurate at generating emotion arcs by aggregating information from hundreds of instances. This has wide-spread implications for commercial development, as well as research in psychology, public health, digital humanities, etc. that values simple interpretable methods and disprefers the need for domain-specific training data, programming expertise, and high-carbon-footprint models.

We present an empirical study on methods for span finding, the selection of consecutive tokens in text for some downstream tasks. We focus on approaches that can be employed in training end-to-end information extraction systems, and find there is no definitive solution without considering task properties, and provide our observations to help with future design choices: 1) a tagging approach often yields higher precision while span enumeration and boundary prediction provide higher recall; 2) span type information can benefit a boundary prediction approach; 3) additional contextualization does not help span finding in most cases.

Current Explainable AI (ExAI) methods, especially in the NLP field, are conducted on various datasets by employing different metrics to evaluate several aspects. The lack of a common evaluation framework is hindering the progress tracking of such methods and their wider adoption. In this work, inspired by offline information retrieval, we propose different metrics and techniques to evaluate the explainability of SA models from two angles. First, we evaluate the strength of the extracted "rationales" in faithfully explaining the predicted outcome. Second, we measure the agreement between ExAI methods and human judgment on a homegrown dataset1 to reflect on the rationales plausibility. Our conducted experiments comprise four dimensions: (1) the underlying architectures of SA models, (2) the approach followed by the ExAI method, (3) the reasoning difficulty, and (4) the homogeneity of the ground-truth rationales. We empirically demonstrate that anchors explanations are more aligned with the human judgment and can be more confident in extracting supporting rationales. As can be foreseen, the reasoning complexity of sentiment is shown to thwart ExAI methods from extracting supporting evidence. Moreover, a remarkable discrepancy is discerned between the results of different explainability methods on the various architectures suggesting the need for consolidation to observe enhanced performance. Predominantly, transformers are shown to exhibit better explainability than convolutional and recurrent architectures. Our work paves the way towards designing more interpretable NLP models and enabling a common evaluation ground for their relative strengths and robustness.

Aspect Based Sentiment Analysis is the most granular form of sentiment analysis that can be performed on the documents / sentences. Besides delivering the most insights at a finer grain, it also poses equally daunting challenges. One of them being the shortage of labelled data. To bring in value right out of the box for the text data being generated at a very fast pace in today's world, unsupervised aspect-based sentiment analysis allows us to generate insights without investing time or money in generating labels. From topic modelling approaches to recent deep learning-based aspect extraction models, this domain has seen a lot of development. One of the models that we improve upon is ABAE that reconstructs the sentences as a linear combination of aspect terms present in it, In this research we explore how we can use information from another unsupervised model to regularize ABAE, leading to better performance. We contrast it with baseline rule based ensemble and show that the ensemble methods work better than the individual models and the regularization based ensemble performs better than the rule-based one.

In this paper, we present a characteristic extraction algorithm and the Multi-domain Image Characteristics Dataset of characteristic-tagged images to simulate the way a human brain classifies cross-domain information and generates insight. The intent was to identify prominent characteristics in data and use this identification mechanism to auto-generate insight from data in other unseen domains. An information extraction algorithm is proposed which is a combination of Variational Autoencoders (VAEs) and Capsule Networks. Capsule Networks are used to decompose images into their individual features and VAEs are used to explore variations on these decomposed features. Thus, making the model robust in recognizing characteristics from variations of the data. A noteworthy point is that the algorithm uses efficient hierarchical decoding of data which helps in richer output interpretation. Noticing a dearth in the number of datasets that contain visible characteristics in images belonging to various domains, the Multi-domain Image Characteristics Dataset was created and made publicly available. It consists of thousands of images across three domains. This dataset was created with the intent of introducing a new benchmark for fine-grained characteristic recognition tasks in the future.

Emotion analysis in texts suffers from two major limitations: annotated gold-standard corpora are mostly small and homogeneous, and emotion identification is often simplified as a sentence-level classification problem. To address these issues, we introduce a new annotation scheme for exploring emotions and their causes, along with a new French dataset composed of autobiographical accounts of an emotional scene. The texts were collected by applying the Cognitive Analysis of Emotions developed by A. Finkel to help people improve on their emotion management. The method requires the manual analysis of an emotional event by a coach trained in Cognitive Analysis. We present a rule-based approach to automatically annotate emotions and their semantic roles (e.g. emotion causes) to facilitate the identification of relevant aspects by the coach. We investigate future directions for emotion analysis using graph structures.

Most existing methods focus on sentiment analysis of textual data. However, recently there has been a massive use of images and videos on social platforms, motivating sentiment analysis from other modalities. Current studies show that exploring other modalities (e.g., images) increases sentiment analysis performance. State-of-the-art multimodal models, such as CLIP and VisualBERT, are pre-trained on datasets with the text paired with images. Although the results obtained by these models are promising, pre-training and sentiment analysis fine-tuning tasks of these models are computationally expensive. This paper introduces a transfer learning approach using joint fine-tuning for sentiment analysis. Our proposal achieved competitive results using a more straightforward alternative fine-tuning strategy that leverages different pre-trained unimodal models and efficiently combines them in a multimodal space. Moreover, our proposal allows flexibility when incorporating any pre-trained model for texts and images during the joint fine-tuning stage, being especially interesting for sentiment classification in low-resource scenarios.

Another classical error at this point is assuming a problem is one kind of machine learning when it may be a multi-step, ensemble approach. Again, returning to the sentiment analysis example, suppose we need to turn a pile of tweets into a prediction of what kind of tweets earn the most engagement. We think we're solving for a prediction, and that may be the last step in the problem, but before we can solve for what makes a tweet engaging, we have to solve for turning text into numbers.

The most prominent tasks in emotion analysis are to assign emotions to texts and to understand how emotions manifest in language. An observation for NLP is that emotions can be communicated implicitly by referring to events, appealing to an empathetic, intersubjective understanding of events, even without explicitly mentioning an emotion name. In psychology, the class of emotion theories known as appraisal theories aims at explaining the link between events and emotions. Appraisals can be formalized as variables that measure a cognitive evaluation by people living through an event that they consider relevant. They include the assessment if an event is novel, if the person considers themselves to be responsible, if it is in line with the own goals, and many others. Such appraisals explain which emotions are developed based on an event, e.g., that a novel situation can induce surprise or one with uncertain consequences could evoke fear. We analyze the suitability of appraisal theories for emotion analysis in text with the goal of understanding if appraisal concepts can reliably be reconstructed by annotators, if they can be predicted by text classifiers, and if appraisal concepts help to identify emotion categories. To achieve that, we compile a corpus by asking people to textually describe events that triggered particular emotions and to disclose their appraisals. Then, we ask readers to reconstruct emotions and appraisals from the text. This setup allows us to measure if emotions and appraisals can be recovered purely from text and provides a human baseline. Our comparison of text classification methods to human annotators shows that both can reliably detect emotions and appraisals with similar performance. Therefore, appraisals constitute an alternative computational emotion analysis paradigm and further improve the categorization of emotions in text with joint models.