Abstract

This paper presents our solutions for Task4 at SemEval2022: Patronizing and Condescending Language Detection. This shared task contains two sub-tasks. The first sub-task is a binary 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 proposed solution uses BERT based ensembled models with hard voting and techniques applied to take care of class imbalances. Our paper describes the system architecture of the submitted solution and other experiments that we conducted. Our best performing models achieve an F1 score of 59.4 and 15.7 on sub-tasks 1 and 2 respectively.

Abstract

CMNEROne at SemEval-2022 Task 11: Code-Mixed Named Entity Recognition by leveraging multilingual data

Abstract

The exponential rise of social media networks has allowed the production, distribution, and consumption of data at a phenomenal rate. Moreover, the social media revolution has brought a unique phenomenon to social media platforms called Internet memes. Internet memes are one of the most popular contents used on social media, and they can be in the form of images with a witty, catchy, or satirical text description. In this paper, we are dealing with propaganda that is often seen in Internet memes in recent times. Propaganda is communication, which frequently includes psychological and rhetorical techniques to manipulate or influence an audience to act or respond as the propagandist wants. To detect propaganda in Internet memes, we propose a multimodal deep learning fusion system that fuses the text and image feature representations and outperforms individual models based solely on either text or image modalities.

Abstract

Named Entity Recognition (NER) is a successful and well-researched problem in English due to the availability of resources. The transformer models, specifically the masked-language models (MLM), have shown remarkable performance in NER in recent times. With growing data in different online platforms, there is a need for NER in other languages too. NER remains underexplored in Indian languages due to the lack of resources and tools. Our contributions in this paper include (i) Two annotated NER datasets for the Telugu language in multiple domains: Newswire Dataset (ND) and Medical Dataset (MD), and we combined ND and MD to form a Combined Dataset (CD) (ii) Comparison of the finetuned Telugu pretrained transformer models (BERT-Te, RoBERTa-Te, and ELECTRA-Te) with other baseline models (CRF, LSTM-CRF, and BiLSTM-CRF) (iii) Further investigation of the performance of Telugu pretrained transformer models against the multilingual models mBERT (Devlin et al., 2018), XLM-R (Conneau et al., 2020), and IndicBERT (Kakwani et al., 2020). We find that pretrained Telugu language models (BERTTe and RoBERTa) outperform the existing pretrained multilingual and baseline models in NER. On a large dataset (CD) of 38,363 sentences, the BERT-Te achieves a high F1-score of 0.80 (entity-level) and 0.75 (token-level). Further, these pretrained Telugu models have shown state-of-the-art performance on various

Abstract

Hateful and offensive content on social media platforms can have negative effects on users and can make online communities more hostile towards certain people and hamper equality, diversity and inclusion. In this paper, we describe our approach to classify homophobia and transphobia in social media comments. We used an ensemble of transformer based models to build our classifier. Our model ranked 2nd for English, 8th for Tamil and 10th for Tamil-English

Abstract

Depression is a common and serious medical illness that negatively affects how you feel, the way you think, and how you act. Detecting depression is essential as it must be treated early to avoid painful consequences. Nowadays, people are broadcasting how they feel via posts and comments. Using social media, we can extract many comments related to depression and use NLP techniques to train and detect depression. This work presents the submission of the DepressionOne team at LT-EDI-2022 for the shared task, detecting signs of depression from social media text. The depression data is small and unbalanced. Thus, we have used oversampling and undersampling methods such as SMOTE and RandomUnderSampler to represent the data. Later, we used machine learning methods to train and detect the signs of depression.

Abstract

Computational Paninian Grammar (CPG) model helps in decoding a natural language expression as a series of modifier-modified relations and therefore facilitates in identifying dependency relations closer to language/context semantics compared to the usual stanford dependency relations. However, the importance of this CPG dependency scheme has not been studied in the context of multi-modal vision and language applications. At IIIT-H, we plan to perform a novel study to explore the potential advantages and disadvantages of CPG framework in a vision-language navigation task setting, a popular and challenging multimodal grounding task.

Abstract

With the increase in user-generated content on social media networks, hate speech and offensive language content are also increasing. From the perspective of computer science, automatic detection of such hate speech and offensive language content is an interesting problem to solve. The natural language community has taken a step to identify such content via automated hate speech and offensive content detection. The hate speech content is generated mostly on social media, and automatic hate speech and offensive language detection face many challenges due to non-standard spelling and grammar variations. Specifically, in a multilingual community, the hate content would be in code-mixed form, making the task further challenging. In this article, we propose a model for code-mixed hate speech detection. This model embeds the knowledge from both user-trained and multilingual pre-trained models. The proposed method also calculates the profanity word list and augments it. Experimental results on code-mixed hate speech and offensive language detection benchmarks show that our method outperforms the existing baselines.

Abstract

The hearing challenged communities all over the world face difficulties to communicate with others. Machine translation has been one of the prominent technologies to facilitate communication with the deaf and hard of hearing community worldwide. We have explored and formulated the fundamental rules of Indian Sign Language(ISL) and implemented them as a translation mechanism of English Text to Indian Sign Language glosses. According to the formulated rules and sub-rules, the source text structure is identified and transferred to the target ISL gloss. This target language is such that it can be easily converted to videos using the Indian Sign Language dictionary. This research work also mentions the intermediate phases of the transfer process and innovations in the process such as Multi-Word Expression detection and synonym substitution to handle the limited vocabulary size of Indian Sign Language while producing semantically accurate translations.

Abstract

This paper presents a learning assistant that tests one’s knowledge and gives feedback that helps a person learn at a faster pace. A learning assistant (based on automated question generation) has extensive uses in education, information websites, self-assessment, FAQs, testing ML agents, research, etc. Multiple researchers, and companies have worked on Virtual Assistance, but majorly in English. We built our learning assistant for Telugu language to help with teaching in the mother tongue, which is the most efficient way of learning. Our system is built primarily based on Question Generation in Telugu. Many experiments were conducted on Question Generation in English in multiple ways. We have built the first hybrid machine learning and rule-based solution in Telugu, which proves efficient for short stories or short passages in children’s books. Our work covers the fundamental question forms with question types: adjective, yes/no, adverb, verb, when, where, whose, quotative, and quantitative (how many/how much). We constructed rules for question generation using Part of Speech (POS) tags and Universal Dependency (UD) tags along with linguistic information of the surrounding relevant context of the word. We used keyword matching, multilingual sentence embedding to evaluate the answer. Our system is primarily built on question generation in Telugu, and is also capable of evaluating the user’s answers to the generated questions.