Abstract

Sexism, a permeate form of oppression, causes profound sufering through various manifestations. Given the increasing number of experiences of sexism shared online, categorizing these recollections automatically can support the battle against sexism, since it can promote successful evaluations by gender studies researchers and government representatives engaged in policy making. In this paper, we examine the fne-grained, multi-label classifcation of accounts (reports) of sexism. To the best of our knowledge, we consider substantially more categories of sexism than any related prior work through our 23-class problem formulation. Moreover, we present the frst semi-supervised work for the multi-label classifcation of accounts describing any type(s) of sexism. We devise self-training-based techniques tailor-made for the multi-label nature of the problem to utilize unlabeled samples for augmenting the labeled set. We identify high textual diversity with respect to the existing labeled set as a desirable quality for candidate unlabeled instances and develop methods for incorporating it into our approach. We also explore ways of infusing class imbalance alleviation for multi-label classifcation into our semisupervised learning, independently and in conjunction with the method involving diversity. In addition to data augmentation methods, we develop a neural model which combines biLSTM and attention with a domain-adapted BERT model in an endto-end trainable manner. Further, we formulate a multi-level training approach in which models are sequentially trained using

Abstract

Witness testimonies are important constituents of a court case description and play a significant role in the final decision. We propose two techniques to identify sentences representing witness testimonies. The first technique employs linguistic rules whereas the second technique applies distant supervision where training set is constructed automatically using the output of the first technique. We then represent the identified witness testimonies in a more meaningful structure – event verb (predicate) along with its arguments corresponding to semantic roles A0 and A1 [1]. We demonstrate effectiveness of such representation in retrieving semantically similar prior relevant cases. To the best of our knowledge, this is the first paper to apply NLP techniques to extract witness information from court judgements and use it for retrieving prior court cases.

Abstract

The pairing of natural language sentences with knowledge graph triples is essential for many downstream tasks like data-to-text generation, facts extraction from sentences (semantic parsing), knowledge graph completion, etc. Most existing methods solve these downstream tasks using neural-based end-to-end approaches that require a large amount of well-aligned training data, which is difficult and expensive to acquire. Recently various unsupervised techniques have been proposed to alleviate this alignment step by automatically pairing the structured data (knowledge graph triples) with textual data. However, these approaches are not well suited for low resource languages that provide two major challenges: (1) unavailability of pair of triples and native text with the same content distribution and (2) limited Natural language Processing (NLP) resources. In this paper, we address the unsupervised pairing of knowledge graph triples with sentences for low resource languages, selecting Hindi as the low resource language. We propose cross-lingual pairing of English triples with Hindi sentences to mitigate the unavailability of content overlap. We propose two novel approaches: NER-based filtering with Semantic Similarity and Key-phrase Extraction with Relevance Ranking. We use our best method to create a collection of 29224 well-aligned English triples and Hindi sentence pairs. Additionally, we have also curated 350 human-annotated golden test datasets for evaluation. We make the code and dataset publicly available.

Abstract

Sexism, a prejudice that causes enormous suffering, mani- fests in blatant as well as subtle ways. As sexist content towards women is increasingly spread on social networks, the automatic detection and categorization of these tweets/posts can help social scientists and policy- makers in research, thereby combating sexism. In this paper, we explore the problem of detecting whether a Twitter/Gab post is sexist or not. We further discriminate the detected sexist post into one of the fine-grained sexism categories. We propose a neural model for this sexism detec- tion and classification that can combine representations obtained using RoBERTa model and linguistic features such as Empath, Hurtlex, and Perspective API by involving recurrent components. We also leverage the unlabeled sexism data to infuse the domain-specific transformer model into our framework. Our proposed framework also features a knowledge module comprised of emoticon and hashtag representations to infuse the external knowledge-specific features into the learning process. Several proposed methods outperform various baselines across several standard metrics

Abstract

Social media platforms have democratized the publication process resulting into easy and viral propagation of information. However, spread of rumors via such media often results into undesired and extremely impactful political, economic, social, psychological and criminal consequences. Several manual as well as automated efforts have been undertaken in the past to solve this critical problem. Existing automated methods are text based, user credibility based or use signals from the tweet propagation tree. We aim at using the text, user, propagation tree and temporal information jointly for rumor detection on Twitter. This involves several challenges like how to handle text variations on Twitter, what signals from user profile could be useful, how to best encode the propagation tree information, and how to incorporate the temporal signal. Our novel architecture, (Text and Temporal Tree Network), leverages deep learning based architectures to encode text, user and tree information in a temporal-aware manner. Our extensive comparisons show that our proposed methods outperform the state-of-the-art techniques by 7 and 6% points respectively on two popular benchmark datasets, and also lead to better early detection results.

Abstract

Given the recent progress in language modeling using Transformer-based neural models and an active interest in generating stylized text, we present an approach to leverage the generalization capabilities of a language model to rewrite an input text in a target author’s style. Our proposed approach adapts a pre-trained language model to generate author-stylized text by fine-tuning on the author-specific corpus using a denoising autoencoder (DAE) loss in a cascaded encoder-decoder framework. Optimizing over DAE loss allows our model to learn the nuances of an author’s style without relying on parallel data, which has been a severe limitation of the previous related works in this space. To evaluate the efficacy of our approach, we propose a linguistically motivated framework to quantify stylistic alignment of the generated text to the target author at lexical, syntactic and surface levels. The evaluation framework is both interpretable as it leads to several insights about the model, and self-contained as it does not rely on external classifiers, e.g. sentiment or formality classifiers. Qualitative and quantitative assessment indicates that the proposed approach rewrites the input text with better alignment to the target style while preserving the original content better than state-of-the-art baselines.

Abstract

In this paper, we propose the use of Message Sequence Charts (MSC) as a representation for visualizing narrative text in Hindi. An MSC is a formal representation allowing the depiction of actors and interactions among these actors in a scenario, apart from supporting a rich framework for formal inference. We propose an approach to extract MSC actors and interactions from a Hindi narrative. As a part of the approach, we enrich an existing event annotation scheme where we provide guidelines for annotation of the mood of events (realis vs irrealis) and guidelines for annotation of event arguments. We report performance on multiple evaluation criteria by experimenting with Hindi narratives from Indian History. Though Hindi is the fourth most-spoken first language in the world, from the NLP perspective it has comparatively lesser resources than English. Moreover, there is relatively less work in the context of event processing in Hindi. Hence, we believe that this work is among the initial works for Hindi event processing.

Abstract

Medical concept normalization aims to map a variable length message such as, ‘unable to sleep’ to an entry in a target medical lexicon, such as ‘Insomnia’. Current approaches formulate medical concept normalization as a supervised text classification problem. This formulation has several drawbacks. First, creating training data requires manually mapping medical concept mentions to their corresponding entries in a target lexicon. Second, these models fail to map a mention to the target concepts which were not encountered during the training phase. Lastly, these models have to be retrained from scratch whenever new concepts are added to the target lexicon. In this work we propose a method which overcomes these limitations. We first use various text and graph embedding methods to encode medical concepts into an embedding space. We then train a model which transforms concept mentions into vectors in this target embedding space. Finally, we use cosine similarity to find the nearest medical concept to a given input medical concept mention. Our model scales to millions of target concepts and trivially accommodates growing target lexicon size without incurring significant computational cost. Experimental results show that our model outperforms the previous state-of-the-art by 4.2% and 6.3% classification accuracy across two benchmark datasets. We also present a variety of studies to evaluate the robustness of our model under different training conditions.

Abstract

In this paper, we extend the task of semantic textual similarity to include sentences which contain emojis. Emojis are ubiquitous on social media today, but are often removed in the pre-processing stage of curating datasets for NLP tasks. In this paper, we qualitatively ascertain the amount of semantic information lost by discounting emojis, as well as show a mechanism of accounting for emojis in a semantic task. We create a sentence similarity dataset of 4000 pairs of tweets with emojis, which have been annotated for relatedness. The corpus contains tweets curated based on common topic as well as by replacement of emojis. The latter was done to analyze the difference in semantics associated with different emojis. We aim to provide an understanding of the information lost by removing emojis by providing a qualitative analysis of the dataset. We also aim to present a method of using both emojis and words for downstream NLP tasks beyond sentiment analysis.

Abstract

Maintaining a unified ontology across various languages is expected to result in effective and consistent organization of Wikipedia entities. Such organization of the Wikipedia knowledge base (KB) will in turn improve the effectiveness of various KB oriented multilingual downstream tasks like entity linking, question answering, fact checking, etc. As a first step toward a unified ontology, it is important to classify Wikipedia entities into consistent fine-grained categories across 30 languages. While there is existing work on finegrained entity categorization for rich-resource languages, there is hardly any such work for consistent classification across multiple low-resource languages. Wikipedia webpage format variations, content imbalance per page, imbalance with respect to categories across languages make the problem challenging. We model this problem as a document classification task. We propose a novel architecture, RNN_GNN_XLM-R, which leverages the strengths of various popular deep learning architectures. Across ten participant teams at the NTCIR-15 Shinra 2020-ML Classification Task, our proposed model stands second in the overall evaluation.