Abstract

This paper presents our solutions for Task4 at SemEval2022: Patronizing and Condescend- ing Language Detection. This shared task con- tains two sub-tasks. The first sub-task is a bi- nary classification task whose goal is to predict whether a given paragraph contains any form of patronising or condescending language(PCL). For the second sub-task, given a paragraph, we have to find which PCL categories express the condescension. Here we have a total of 7 overlapping sub-categories for PCL. Our pro- posed solution uses BERT based ensembled models with hard voting and techniques applied to take care of class imbalances. Our paper de- scribes the system architecture of the submitted solution and other experiments that we con- ducted. Our best performing models achieve an F1 score of 59.4 and 15.7 on sub-tasks 1 and 2 respectively.

Abstract

Social Media Platforms like Facebook, Twitter, Instagram, etc. have large user base all around the world that generates huge amounts of data every second. This includes a lot of posts by fake and spam users, typically used by many organizations around the globe to gain a competitive edge over others. In this work, we aim at detecting such user accounts on Twitter. We show how to distinguish between Genuine and Spam accounts in Twitter using a novel combination of feature engineering, network representation, and natural language processing techniques. Index Terms—social media, fake user, Twitter, spam detection, graph, natural language processing

Abstract

Due to the lack of a large annotated corpus, many resource-poor Indian languages struggle to reap the beneits of recent deep feature representations in Natural Language Processing (NLP). Moreover, adopting existing language models trained on large English corpora for Indian languages is oten limited by data availability, rich morphological variation, syntax, and semantic diferences. In this paper, we explore the traditional to recent eicient representations to overcome the challenges of low resource language, Telugu. In particular, our main objective is to mitigate the low-resource problem for Telugu. Overall, we present several contributions to a resource-poor language viz. Telugu. (i) a large annotated data (35,142 sentences in each task) for multiple NLP tasks such as sentiment analysis, emotion identiication, hate-speech detection, and sarcasm detection, (ii) we create diferent lexicons for sentiment, emotion, and hate-speech for improving the eiciency of the models, (iii) pretrained word and sentence embeddings, and (iv) diferent pretrained language models for Telugu such as ELMo-Te, BERT-Te, RoBERTa-Te, ALBERT-Te, and DistilBERT-Te on a large Telugu corpus consisting of 80,15,588 sentences (16,37,408 sentences from Telugu Wikipedia and 63,78,180 sentences crawled from diferent Telugu websites). Further, we show that these representations signiicantly improve the performance of four NLP tasks and present the benchmark results for Telugu. We argue that our pretrained embeddings are competitive or beter than the existing multilingual pretrained models: mBERT, XLM-R, and IndicBERT. Lastly, the ine-tuning of pretrained models show higher performance than linear probing results on four NLP tasks with following F1-scores: Sentiment (68.72), Emotion (58.04), Hate-Speech (64.27) and Sarcasm (77.93). We also experiment on publicly available Telugu datasets (Named Entity Recognition, Article Genre Classiication, and Sentiment Analysis ), ind that our Telugu pretrained language models (BERT-Te and RoBERTa-Te) outperform the state-of-the-art system except for the sentiment task. We open-source our corpus, four diferent datasets, lexicons, embeddings, and code https://github.com/Cha14ran/DREAM-T. he pretrained Transformer models for Telugu are available at https://huggingface.co/ltrctelugu.

Abstract

Document summarization aims to create a precise and coherent summary of a text document. Many deep learning summarization models are developed mainly for English, often requiring a large training corpus and efficient pre-trained language models and tools. However, English summarization models for low-resource Indian languages are often limited by rich morphological variation, syntax, and semantic differences. In this paper, we propose GAE-ISUMM, an unsupervised Indic summarization model that extracts summaries from text documents. In particular, our proposed model, GAE-ISUMM uses Graph Autoencoder (GAE) to learn text representations and a document summary jointly. We also provide a manually-annotated Telugu summarization dataset TELSUM, to experiment with our model GAE-ISUMM. Further, we experiment with the most publicly available Indian language summarization datasets to investigate the effectiveness of GAE-ISUMM on other Indian languages. Our experiments of GAE-ISUMM in seven languages make the following observations: (i) it is competitive or better than state-ofthe-art results on all datasets, (ii) it reports benchmark results on TELSUM, and (iii) the inclusion of positional and cluster information in the proposed model improved the performance of summaries. We open-source our dataset and code 1 . Index Terms—component, formatting, style, styling, insert

Abstract

Due to the lack of a large annotated corpus, many resource-poor Indian languages struggle to reap the benefits of recent deep feature representations in Natural Language Processing (NLP). Moreover, adopting existing language models trained on large English corpora for Indian languages is often limited by data availability, rich morphological variation, syntax, and semantic differences. In this paper, we explore the traditional to recent efficient representations to overcome the challenges of a low resource language, Telugu. In particular, our main objective is to mitigate the low-resource problem for Telugu. Overall, we present several contribu- tions to a resource-poor language viz. Telugu. (i) a large annotated data (35,142 sentences in each task) for multiple NLP tasks such as sentiment analysis, emotion identification, hate-speech detection, and sarcasm detection, (ii) we create different lexicons for sentiment, emotion, and hate-speech for improving the effi- ciency of the models, (iii) pretrained word and sentence embeddings, and (iv) different pretrained language models for Telugu such as ELMo-Te, BERT-Te, RoBERTa-Te, ALBERT-Te, and DistilBERT-Te on a large Telugu corpus consisting of 8,015,588 sentences (1,637,408 sentences from Telugu Wikipedia and 6,378,180 sentences crawled from different Telugu websites). Further, we show that these representations significantly improve the performance of four NLP tasks and present the benchmark results for Telugu. We argue that our pretrained embeddings are competitive or better than the existing multilingual pretrained models: mBERT, XLM-R, and IndicBERT. Lastly, the fine-tuning of pretrained models show higher performance than linear probing results on four NLP tasks with the following F1-scores: Sentiment (68.72), Emotion (58.04), Hate-Speech (64.27), and Sarcasm (77.93). We also experiment on publicly available Telugu datasets (Named Entity Recognition, Article Genre Classification, and Sentiment Analysis) and find that our Telugu pretrained language models (BERT-Te and RoBERTa-Te) outperform the state-of-the-art system except for the sentiment task. We open-source our corpus, four different datasets, lexicons, embeddings, and code https://github.com/Cha14ran/DREAM-T. The pretrained Transformer models for Telugu are available at https://huggingface.co/ltrctelug

Abstract

Pre-training a language model and then fine-tuning it for downstream tasks has demonstrated state-of-the-art results for various NLP tasks. Pre-training is usually independent of the downstream task, and previous works have shown that this pre-training alone might not be sufficient to capture the task-specific nuances. We propose a way to tailor a pre-trained BERT model for the downstream task via task-specific masking before the standard supervised fine-tuning. For this, a word list is first collected specific to the task. For example, if the task is sentiment classification, we collect a small sample of words representing both positive and negative sentiments. Next, a word’s importance for the task, called the word’s task score, is measured using the word list. Each word is then assigned a probability of masking based on its task score. We experiment with different masking functions that assign the probability of masking based on the word’s task score. The BERT model is further trained on MLM objective, where masking is done using the above strategy. Following this standard supervised fine-tuning is done for different downstream tasks. Results on these tasks show that the selective masking strategy outperforms random masking, indicating its effectiveness.

Abstract

Vision-and-language tasks are gaining popularity in the research community, but the focus is still mainly on English. We propose a pipeline that utilizes English-only vision-language models to train a monolingual model for a target language. We propose to extend OSCAR+, a model which leverages object tags as anchor points for learning image-text alignments, to train on visual question answering datasets in different languages. We propose a novel approach to knowledge distillation to train the model in other languages using parallel sentences. Compared to other models that use the target language in the pretraining corpora, we can leverage an existing English model to transfer the knowledge to the target language using significantly lesser resources. We also release a large-scale visual question answering dataset in Japanese and Hindi language. Though we restrict our work to visual question answering, our model can be extended to any sequence-level classification task, and it can be extended to other languages as well. This paper focuses on two languages for the visual question answering task - Japanese and Hindi. Our pipeline outperforms the current state-of-the-art models by a relative increase of 4.4% and 13.4% respectively in accuracy.

Abstract

—Graph Convolutional Networks (GCN) have achieved state-of-art results on single text classification tasks like sentiment analysis, emotion detection, etc. However, the performance is achieved by testing and reporting on resourcerich languages like English. Applying GCN for multi-task text classification is an unexplored area. Moreover, training a GCN or adopting an English GCN for Indian languages is often limited by data availability, rich morphological variation, syntax, and semantic differences. In this paper, we study the use of GCN for the Telugu language in single and multi-task settings for four natural language processing (NLP) tasks, viz. sentiment analysis (SA), emotion identification (EI), hate-speech (HS), and sarcasm detection (SAR). In order to evaluate the performance of GCN with one of the Indian languages, Telugu, we analyze the GCN based models with extensive experiments on four downstream tasks. In addition, we created an annotated Telugu dataset, TEL-NLP, for the four NLP tasks. Further, we propose a supervised graph reconstruction method, Multi-Task Text GCN (MT-Text GCN) on the Telugu that leverages to simultaneously (i) learn the low-dimensional word and sentence graph embeddings from word-sentence graph reconstruction using graph autoencoder (GAE) and (ii) perform multi-task text classification using these latent sentence graph embeddings. We argue that our proposed MT-Text GCN achieves significant improvements on TEL-NLP over existing Telugu pretrained word embeddings [1], and multilingual pretrained Transformer models: mBERT [2], and XLM-R [3]. On TEL-NLP, we achieve a high F1-score for four NLP tasks: SA (0.84), EI (0.55), HS (0.83) and SAR (0.66). Finally, we show our model’s quantitative and qualitative analysis on the four NLP tasks in Telugu. We open-source our TEL-NLP dataset, pretrained models, and code 1 . Index Terms—Graph Convolutional Networks, Text

Abstract

Over the past few years, there has been a growing concern around toxic positivity on social media which is a phenomenon where positivity is used to minimize one’s emotional experience. In this paper, we create a dataset for toxic positivity classification from Twitter and an inspirational quote website. We then perform benchmarking experiments using various text classification models and show the suitability of these models for the task. We achieved a macro F1 score of 0.71 and a weighted F1 score of 0.85 by using an ensemble model. To the best of our knowledge, our dataset is the first such dataset created.

Abstract

In current times, memes have become one of the most popular mediums to share jokes and information with the masses over the internet. Memes can also be used as tools to spread hatred and target women through degrading content disguised as humour. The task, Multimedia Automatic Misogyny Identification (MAMI), is to detect misogyny in these memes. This task is further divided into two sub-tasks: (A) Misogynous meme identification, where a meme should be categorized either as misogynous or not misogynous and (B) Categorizing these misogynous memes into potential overlapping subcategories. In this paper, we propose models leveraging task-specific pretraining with transfer learning on Visual Linguistic models. With our best performing models, we were able to achieve rank 5 th and 10 th on sub-tasks A and B respectively