In this paper, we present a new Tamil lyrics corpus extracted from Tamil movies captured across a range of 65 years (1954 to 2019). We present a detailed corpus analysis showing the nature of Tamil lyrics with respect to lyricists and the year which it was written. We also present similar- ity score across different lyricists based on their song lyrics. We present experi- mental results based on the SOTA BERT Tamil models to identify the lyricists of a song. Finally, we present future research directions encouraging researchers to pur- sue Tamil NLP research.

This paper presents the Dravidian Offensive Span Identification Dataset (DOSA) for under-resourced Tamil-English and Kannada-English code-mixed text. The dataset addresses the lack of code-mixed datasets with annotated offensive spans by extending annotations of existing code-mixed offensive language identification datasets. It provides span annotations for Tamil-English and Kannada-English code-mixed comments posted by users on YouTube social media. Overall the dataset consists of 4786 Tamil-English comments with 6202 annotated spans and 1097 Kannada-English comments with 1641 annotated spans, each annotated by two different annotators. We further present some of our baseline experimental results on the developed dataset, thereby eliciting research in under-resourced languages, leading to an essential step towards semi-automated content moderation in Dravidian languages. The dataset is available in https://github.com/teamdl-mlsg/DOSA

Offensive speech identification in countries like India poses several challenges due to the usage of code-mixed and romanized variants of multiple languages by the users in their posts on social media. The challenge of offensive language identification on social media for Dravidian languages is harder, considering the low resources available for the same. In this paper, we explored the zero-shot learning and few-shot learning paradigms based on multilingual language models for offensive speech detection in code-mixed and romanized variants of three Dravidian languages - Malayalam, Tamil, and Kannada. We propose a novel and flexible approach of selective translation and transliteration to reap better results from fine-tuning and ensembling multilingual transformer networks like XLMRoBERTa and mBERT. We implemented pretrained, fine-tuned, and ensembled versions of XLM-RoBERTa for offensive speech classification. Further, we experimented with interlanguage, inter-task, and multi-task transfer learning techniques to leverage the rich resources available for offensive speech identification in the English language and to enrich the models with knowledge transfer from related tasks. The proposed models yielded good results and are promising for effective offensive speech identification in low resource settings.

Sentiment analysis tools and models have been developed extensively throughout the years, for European languages. In contrast, similar tools for Indian Languages are scarce. This is because, state-of-the-art pre-processing tools like POS tagger, shallow parsers, etc., are not readily available for Indian languages. Although, such working tools for Indian languages, like Hindi and Bengali, that are spoken by the majority of the population, are available, finding the same for less spoken languages like, Tamil, Telugu, and Malayalam, is difficult. Moreover, due to the advent of social media, the multi-lingual population of India, who are comfortable with both English ad their regional language, prefer to communicate by mixing both languages. This gives rise to massive code-mixed content and automatically annotating them with their respective sentiment labels becomes a challenging task. In this work, we take up a similar challenge of developing a sentiment analysis model that can work with English-Tamil code-mixed data. The proposed work tries to solve this by using bi-directional LSTMs along with language tagging. Other traditional methods, based on classical machine learning algorithms have also been discussed in the literature, and they also act as the baseline systems to which we will compare our Neural Network based model. The performance of the developed algorithm, based on Neural Network architecture, garnered precision, recall, and F1 scores of 0.59, 0.66, and 0.58 respectively.

Hate speech and offensive language recognition in social media platforms have been an active field of research over recent years. In non-native English spoken countries, social media texts are mostly in code mixed or script mixed/switched form. The current study presents extensive experiments using multiple machine learning, deep learning, and transfer learning models to detect offensive content on Twitter. The data set used for this study are in Tanglish (Tamil and English), Manglish (Malayalam and English) code-mixed, and Malayalam script-mixed. The experimental results showed that 1 to 6-gram character TF-IDF features are better for the said task. The best performing models were naive bayes, logistic regression, and vanilla neural network for the dataset Tamil code-mix, Malayalam code-mixed, and Malayalam script-mixed, respectively instead of more popular transfer learning models such as BERT and ULMFiT and hybrid deep models.

This paper presents the methodologies implemented while classifying Dravidian code-mixed comments according to their polarity. With datasets of code-mixed Tamil and Malayalam available, three methods are proposed - a sub-word level model, a word embedding based model and a machine learning based architecture. The sub-word and word embedding based models utilized Long Short Term Memory (LSTM) network along with language-specific preprocessing while the machine learning model used term frequency–inverse document frequency (TF-IDF) vectorization along with a Logistic Regression model. The sub-word level model was submitted to the the track ‘Sentiment Analysis for Dravidian Languages in Code-Mixed Text’ proposed by Forum of Information Retrieval Evaluation in 2020 (FIRE 2020). Although it received a rank of 5 and 12 for the Tamil and Malayalam tasks respectively in the FIRE 2020 track, this paper improves upon the results by a margin to attain final weighted F1-scores of 0.65 for the Tamil task and 0.68 for the Malayalam task. The former score is equivalent to that attained by the highest ranked team of the Tamil track.

Unsupervised Neural Machine translation (UNMT) is beneficial especially for under-resourced languages such as from the Dravidian family. They learn to translate between the source and target, relying solely on only monolingual corpora. However, UNMT systems fail in scenarios that occur often when dealing with low resource languages. Recent works have achieved state-of-the-art results by adding auxiliary parallel data with similar languages. In this work, we focus on unsupervised translation between English and Kannada by using limited amounts of auxiliary data between English and other Dravidian languages. We show that transliteration is essential in unsupervised translation between Dravidian languages, as they do not share a common writing system. We explore several model architectures that use the auxiliary data in order to maximize knowledge sharing and enable UNMT for dissimilar language pairs. We show from our experiments it is crucial for Kannada and reference languages to be similar. Further, we propose a method to measure language similarity to choose the most beneficial reference languages.

Code-mixing is a frequently observed phenomenon in multilingual communities where a speaker uses multiple languages in an utterance or sentence. Code-mixed texts are abundant, especially in social media, and pose a problem for NLP tools as they are typically trained on monolingual corpora. Recently, finding the sentiment from code-mixed text has been attempted by some researchers in SentiMix SemEval 2020 and Dravidian-CodeMix FIRE 2020 shared tasks. Mostly, the attempts include traditional methods, long short term memory, convolutional neural networks, and transformer models for code-mixed sentiment analysis (CMSA). However, no study has explored graph convolutional neural networks on CMSA. In this paper, we propose the graph convolutional networks (GCN) for sentiment analysis on code-mixed text. We have used the datasets from the Dravidian-CodeMix FIRE 2020. Our experimental results on multiple CMSA datasets demonstrate that the GCN with multi-headed attention model has shown an improvement in classification metrics.

Sentiment analysis in Code-Mixed languages has garnered a lot of attention in recent years. It is an important task for social media monitoring and has many applications, as a large chunk of social media data is Code-Mixed. In this paper, we work on the problem of sentiment analysis for Dravidian Code-Switched languages - Tamil-Engish and Malayalam-English, using three different BERT based models. We leverage task-specific pre-training and cross-lingual transfer to improve on previously reported results, with significant improvement for the Tamil-Engish dataset. We also present a multilingual sentiment classification model that has competitive performance on both Tamil-English and Malayalam-English datasets.

Uvama urubugal in Tamil are used to explain a particular context by citing another equivalent context. This is referred to as “Uvamaiyani” in Tamil Grammar rules as stated in Tholkappiam. The is called as simile in English. Similes bring out many beautiful poetic contexts. Automatic extraction of such similes can help to build better Natural Language Generation applications such as, story generation systems and lyric suggestion systems. This paper attempts to automatically extract the uvama urubugal from Tamil Sangam Literatures. Natrinai and Mullai Pattu have been used for the analysis. There are 12 uvama urupugal in Tamil as per Nanool and this paper has attempted to analyze the usage of these 12 uvama urubugal in Sangam Literatures and compares their usage distribution in the Tamil Film songs data set comprising of 4215 songs. It was found that only two uvama urubugal were used in the current-day Tamil Film songs. This comparison was done to reveal the diminishing usage of these beautiful uvama urubugal by the current generation and the urge to use them again.

Task-oriented dialog systems help a user achieve a particular goal by parsing user requests to execute a particular action. These systems typically require copious amounts of training data to effectively understand the user intent and its corresponding slots. Acquiring large training corpora requires significant manual effort in annotation, rendering its construction infeasible for low-resource languages. In this paper, we present a two-step approach for automatically constructing task-oriented dialogue data in such languages by making use of annotated data from high resource languages. First, we use a machine translation (MT) system to translate the utterance and slot information to the target language. Second, we use token prefix matching and mBERT based semantic matching to align the slot tokens to the corresponding tokens in the utterance. We hand-curate a new test dataset in two low-resource Dravidian languages and show the significance and impact of our training dataset construction using a state-of-the-art mBERT model - achieving a Slot F1 of 81.51 (Kannada) and 78.82 (Tamil) on our test sets.

Speech carries not only the semantic content but also the paralinguistic information which captures the speaking style. Speaker traits and emotional states affect how words are being spoken. The research on paralinguistic information is an emerging field in speech and language processing and it has many potential applications including speech recognition, speaker identification and verification, emotion recognition and accent recognition. Among them, there is a significant interest in emotion recognition from speech. A detailed study of paralinguistic information present in speech signal and an overview of research work related to speech emotion for Tamil Language is presented in this paper.

Quality of a product is the degree to which a product meets the customer’s expectation, which must also be valid for the case of lexical semantic resources. Conducting a periodic evaluation of resources is essential to ensure if the resources meet a native speaker’s expectations and free from errors. This paper defines the possible mistakes in a lexical semantic resource and explains the steps applied to quantify Malayalam wordnet quality. Malayalam is one of the classical languages of India. We hope to subset the less quality part of the wordnet and perform crowdsourcing to make it better.

Substantial amount of text data which is increasingly being generated and shared on the internet and social media every second affect the society positively or negatively almost in any aspect of online world and also business and industries. Sentiments/opinions/reviews’ of users posted on social media are the valuable information that have motivated researchers to analyze them to get better insight and feedbacks about any product such as a video in Instagram, a movie in Netflix, or even new brand car introduced by BMW. Sentiments are usually written using a combination of languages such as English which is resource rich and regional languages such as Tamil, Kannada, Malayalam, etc. which are resource poor. However, due to technical constraints, many users prefer to pen their opinions in Roman script. These kinds of texts written in two or more languages using a common language script or different language scripts are called code-mixing texts. Code-mixed texts are increasing day-by-day with the increase in the number of users depending on various online platforms. Analyzing such texts pose a real challenge for the researchers. In view of the challenges posed by the code-mixed texts, this paper describes three proposed models namely, SACo-Ensemble, SACo-Keras, and SACo-ULMFiT using Machine Learning (ML), Deep Learning (DL), and Transfer Learning (TL) approaches respectively for the task of Sentiments Analysis in Tamil-English and Malayalam-English code-mixed texts.

This paper presents an overview of the shared task on machine translation of Dravidian languages. We presented the shared task results at the EACL 2021 workshop on Speech and Language Technologies for Dravidian Languages. This paper describes the datasets used, the methodology used for the evaluation of participants, and the experiments’ overall results. As a part of this shared task, we organized four sub-tasks corresponding to machine translation of the following language pairs: English to Tamil, English to Malayalam, English to Telugu and Tamil to Telugu which are available at https://competitions.codalab.org/competitions/27650. We provided the participants with training and development datasets to perform experiments, and the results were evaluated on unseen test data. In total, 46 research groups participated in the shared task and 7 experimental runs were submitted for evaluation. We used BLEU scores for assessment of the translations.

The internet has facilitated its user-base with a platform to communicate and express their views without any censorship. On the other hand, this freedom of expression or free speech can be abused by its user or a troll to demean an individual or a group. Demeaning people based on their gender, sexual orientation, religious believes or any other characteristics –trolling– could cause great distress in the online community. Hence, the content posted by a troll needs to be identified and dealt with before causing any more damage. Amongst all the forms of troll content, memes are most prevalent due to their popularity and ability to propagate across cultures. A troll uses a meme to demean, attack or offend its targetted audience. In this shared task, we provide a resource (TamilMemes) that could be used to train a system capable of identifying a troll meme in the Tamil language. In our TamilMemes dataset, each meme has been categorized into either a “troll” or a “not_troll” class. Along with the meme images, we also provided the Latin transcripted text from memes. We received 10 system submissions from the participants which were evaluated using the weighted average F1-score. The system with the weighted average F1-score of 0.55 secured the first rank.

Detecting offensive language in social media in local languages is critical for moderating user-generated content. Thus, the field of offensive language identification in under-resourced Tamil, Malayalam and Kannada languages are essential. As the user-generated content is more code-mixed and not well studied for under-resourced languages, it is imperative to create resources and conduct benchmarking studies to encourage research in under-resourced Dravidian languages. We created a shared task on offensive language detection in Dravidian languages. We summarize here the dataset for this challenge which are openly available at https://competitions.codalab.org/competitions/27654, and present an overview of the methods and the results of the competing systems.

In this paper, we describe the GX system in the EACL2021 shared task on machine translation in Dravidian languages. Given the low amount of parallel training data, We adopt two methods to improve the overall performance: (1) multilingual translation, we use a shared encoder-decoder multilingual translation model handling multiple languages simultaneously to facilitate the translation performance of these languages; (2) back-translation, we collected other open-source parallel and monolingual data and apply back-translation to benefit from the monolingual data. The experimental results show that we can achieve satisfactory translation results in these Dravidian languages and rank first in English-Telugu and Tamil-Telugu translation.

The intensity of online abuse has increased in recent years. Automated tools are being developed to prevent the use of hate speech and offensive content. Most of the technologies use natural language and machine learning tools to identify offensive text. In a multilingual society, where code-mixing is a norm, the hate content would be delivered in a code-mixed form in social media, which makes the offensive content identification, further challenging. In this work, we participated in the EACL task to detect offensive content in the code-mixed social media scenario. The methodology uses a transformer model with transliteration and class balancing loss for offensive content identification. In this task, our model has been ranked 2nd in Malayalam-English and 4th in Tamil-English code-mixed languages.

This article introduces the system for the shared task of Offensive Language Identification in Dravidian Languages-EACL 2021. The world’s information technology develops at a high speed. People are used to expressing their views and opinions on social media. This leads to a lot of offensive language on social media. As people become more dependent on social media, the detection of offensive language becomes more and more necessary. This shared task is in three languages: Tamil, Malayalam, and Kannada. Our team takes part in the Kannada language task. To accomplish the task, we use the XLM-Roberta model for pre-training. But the capabilities of the XLM-Roberta model do not satisfy us in terms of statement information collection. So we made some tweaks to the output of this model. In this paper, we describe the models and experiments for accomplishing the task of the Kannada language.

This paper describes our solution submitted to shared task on Offensive Language Identification in Dravidian Languages. We participated in all three of offensive language identification. In order to address the task, we explored multilingual models based on XLM-RoBERTa and multilingual BERT trained on mixed data of three code-mixed languages. Besides, we solved the class-imbalance problem existed in training data by class combination, class weights and focal loss. Our model achieved weighted average F1 scores of 0.75 (ranked 4th), 0.94 (ranked 4th) and 0.72 (ranked 3rd) in Tamil-English task, Malayalam-English task and Kannada-English task, respectively.

Title: JudithJeyafreedaAndrew@DravidianLangTech-EACL2021:Offensive language detection for Dravidian Code-mixed YouTube comments Author: Judith Jeyafreeda Andrew Messaging online has become one of the major ways of communication. At this level, there are cases of online/digital bullying. These include rants, taunts, and offensive phrases. Thus the identification of offensive language on the internet is a very essential task. In this paper, the task of offensive language detection on YouTube comments from the Dravidian lan- guages of Tamil, Malayalam and Kannada are seen upon as a mutliclass classification prob- lem. After being subjected to language spe- cific pre-processing, several Machine Learn- ing algorithms have been trained for the task at hand. The paper presents the accuracy results on the development datasets for all Machine Learning models that have been used and fi- nally presents the weighted average scores for the test set when using the best performing Ma- chine Learning model.

The submission is being made as a working note as part of the Offensive Language Identification in Dravidian Languages shared task. The proposed model “DrOLIC” uses IndicBERT and a simple 4-layered MLP to do the multiclass classification problem and we achieved an F1 score of 0.85 on the Malayalam dataset.

Tamil is a Dravidian language that is commonly used and spoken in the southern part of Asia. During the 21st century and in the era of social media, memes have been a fun moment during the day to day life of people. Here, we try to analyze the true meaning of Tamil memes by classifying them as troll or non-troll. We present an ingenious model consisting of transformer-transformer architecture that tries to attain state of the art by using attention as its main component. The dataset consists of troll and non-troll images with their captions as texts. The task is a binary classification task. The objective of the model was to pay more and more attention to the extracted features and to ignore the noise in both images and text.

This paper demonstrates our work for the shared task on Offensive Language Identification in Dravidian Languages-EACL 2021. Offensive language detection in the various social media platforms was identified previously. But with the increase in diversity of users, there is a need to identify the offensive language in multilingual posts that are largely code-mixed or written in a non-native script. We approach this challenge with various transfer learning-based models to classify a given post or comment in Dravidian languages (Malayalam, Tamil, and Kannada) into 6 categories. The source codes for our systems are published.

Code-Mixed Offensive contents are used pervasively in social media posts in the last few years. Consequently, gained the significant attraction of the research community for identifying the different forms of such content (e.g., hate speech, and sentiments) and contributed to the creation of datasets. Most of the recent studies deal with high-resource languages (e.g., English) due to many publicly available datasets, and by the lack of dataset in low-resource anguages, those studies are slightly involved in these languages. Therefore, this study has the focus on offensive language identification on code-mixed low-resourced Dravidian languages such as Tamil, Kannada, and Malayalam using the bidirectional approach and fine-tuning strategies. According to the leaderboard, the proposed model got a 0.96 F1-score for Malayalam, 0.73 F1-score for Tamil, and 0.70 F1-score for Kannada in the bench-mark. Moreover, in the view of multilingual models, this modal ranked 3rd and achieved favorable results and confirmed the model as the best among all systems submitted to these shared tasks in these three languages.

This paper introduces the system description of the HUB team participating in DravidianLangTech - EACL2021: Offensive Language Identification in Dravidian Languages. The theme of this shared task is the detection of offensive content in social media. Among the known tasks related to offensive speech detection, this is the first task to detect offensive comments posted in social media comments in the Dravidian language. The task organizer team provided us with the code-mixing task data set mainly composed of three different languages: Malayalam, Kannada, and Tamil. The tasks on the code mixed data in these three different languages can be seen as three different comment/post-level classification tasks. The task on the Malayalam data set is a five-category classification task, and the Kannada and Tamil language data sets are two six-category classification tasks. Based on our analysis of the task description and task data set, we chose to use the multilingual BERT model to complete this task. In this paper, we will discuss our fine-tuning methods, models, experiments, and results.

This article describes our system for task DravidianLangTech - EACL2021: Meme classification for Tamil. In recent years, we have witnessed the rapid development of the Internet and social media. Compared with traditional TV and radio media platforms, there are not so many restrictions on the use of online social media for individuals and many functions of online social media platforms are free. Based on this feature of social media, it is difficult for people’s posts/comments on social media to be strictly and effectively controlled like TV and radio content. Therefore, the detection of negative information in social media has attracted attention from academic and industrial fields in recent years. The task of classifying memes is also driven by offensive posts/comments prevalent on social media. The data of the meme classification task is the fusion data of text and image information. To identify the content expressed by the meme, we develop a system that combines BiGRU and CNN. It can fuse visual features and text features to achieve the purpose of using multi-modal information from memetic data. In this article, we discuss our methods, models, experiments, and results.

The development of online media platforms has given users more opportunities to post and comment freely, but the negative impact of offensive language has become increasingly apparent. It is very necessary for the automatic identification system of offensive language. This paper describes our work on the task of Offensive Language Identification in Dravidian language-EACL 2021. To complete this task, we propose a system based on the multilingual model XLM-Roberta and DPCNN. The test results on the official test data set confirm the effectiveness of our system. The weighted average F1-score of Kannada, Malayalam, and Tami language are 0.69, 0.92, and 0.76 respectively, ranked 6th, 6th, and 3rd

This paper describes the IIITK team’s submissions to the offensive language identification, and troll memes classification shared tasks for Dravidian languages at DravidianLangTech 2021 workshop@EACL 2021. Our best configuration for Tamil troll meme classification achieved 0.55 weighted average F1 score, and for offensive language identification, our system achieved weighted F1 scores of 0.75 for Tamil, 0.95 for Malayalam, and 0.71 for Kannada. Our rank on Tamil troll meme classification is 2, and offensive language identification in Tamil, Malayalam and Kannada are 3, 3 and 4 respectively.

This paper introduces the related content of the task “Offensive Language Identification in Dravidian LANGUAGES-EACL 2021”. The task requires us to classify Dravidian languages collected from social media into Not-Offensive, Off-Untargeted, Off-Target-Individual, etc. This data set contains actual annotations in code-mixed text posted by users on Youtube, not from the monolingual text in textbooks. Based on the features of the data set code mixture, we use multilingual BERT and TextCNN for semantic extraction and text classification. In this article, we will show the experiment and result analysis of this task.

Identifying offensive information from tweets is a vital language processing task. This task concentrated more on English and other foreign languages these days. In this shared task on Offensive Language Identification in Dravidian Languages, in the First Workshop of Speech and Language Technologies for Dravidian Languages in EACL 2021, the aim is to identify offensive content from code mixed Dravidian Languages Kannada, Malayalam, and Tamil. Our team used language agnostic BERT (Bidirectional Encoder Representation from Transformers) for sentence embedding and a Softmax classifier. The language-agnostic representation based classification helped obtain good performance for all the three languages, out of which results for the Malayalam language are good enough to obtain a third position among the participating teams.

Social media are an open forum that allows people to share their knowledge, abilities, talents, ideas, or expressions. Simultaneously, it also allows people to post disrespectful, trolling, defamation, or negative content targeting users or the community based on their gender, race, religious beliefs, etc. Such posts are available in the form of text, image, video, and meme. Among them, memes are currently widely used to disseminate offensive material amongst people. It is primarily in the form of pictures and text. In the present paper, troll memes are identified, which is necessary to create a healthy society. To do so, a hybrid deep learning model combining convolutional neural networks and bidirectional long short term memory is proposed to identify trolled memes. The dataset used in the study is a part of the competition EACL 2021: Troll Meme classification in Tamil. The proposed model obtained 10th rank in the competition and reported a precision of 0.52, recall 0.59, and weighted F10.3.

This paper describes the submission of the team Amrita_CEN_NLP to the shared task on Offensive Language Identification in Dravidian Languages at EACL 2021. We implemented three deep neural network architectures such as a hybrid network with a Convolutional layer, a Bidirectional Long Short-Term Memory network (Bi-LSTM) layer and a hidden layer, a network containing a Bi-LSTM and another with a Bidirectional Recurrent Neural Network (Bi-RNN). In addition to that, we incorporated a cost-sensitive learning approach to deal with the problem of class imbalance in the training data. Among the three models, the hybrid network exhibited better training performance, and we submitted the predictions based on the same.

The increasing accessibility of the internet facilitated social media usage and encouraged individuals to express their opinions liberally. Nevertheless, it also creates a place for content polluters to disseminate offensive posts or contents. Most of such offensive posts are written in a cross-lingual manner and can easily evade the online surveillance systems. This paper presents an automated system that can identify offensive text from multilingual code-mixed data. In the task, datasets provided in three languages including Tamil, Malayalam and Kannada code-mixed with English where participants are asked to implement separate models for each language. To accomplish the tasks, we employed two machine learning techniques (LR, SVM), three deep learning (LSTM, LSTM+Attention) techniques and three transformers (m-BERT, Indic-BERT, XLM-R) based methods. Results show that XLM-R outperforms other techniques in Tamil and Malayalam languages while m-BERT achieves the highest score in the Kannada language. The proposed models gained weighted f_1 score of 0.76 (for Tamil), 0.93 (for Malayalam ), and 0.71 (for Kannada) with a rank of 3rd, 5th and 4th respectively.

This paper describes our team’s submission of the EACL DravidianLangTech-2021’s shared task on Machine Translation of Dravidian languages. We submitted our translations for English to Malayalam , Tamil , Telugu and also Tamil-Telugu language pairs. The submissions mainly focus on having adequate amount of data backed up by good preprocessing of it to produce quality translations,which includes some custom made rules to remove unnecessary sentences. We conducted several experiments on these models by tweaking the architecture,Byte Pair Encoding (BPE) and other hyperparameters.

This paper presents the participation of the IRNLPDAIICT team from Information Retrieval and Natural Language Processing lab at DA-IICT, India in DravidianLangTech-EACL2021 Offensive Language identification in Dravidian Languages. The aim of this shared task is to identify Offensive Language from a code-mixed data-set of YouTube comments. The task is to classify comments into Not Offensive (NO), Offensive Untargetede(OU), Offensive Targeted Individual (OTI), Offensive Targeted Group (OTG), Offensive Targeted Others (OTO), Other Language (OL) for three Dravidian languages: Kannada, Malayalam and Tamil. We use TF-IDF character n-grams and pretrained MuRIL embeddings for text representation and Logistic Regression and Linear SVM for classification. Our best approach achieved Ninth, Third and Eighth with weighted F1 score of 0.64, 0.95 and 0.71in Kannada, Malayalam and Tamil on test dataset respectively.

Social media often acts as breeding grounds for different forms of offensive content. For low resource languages like Tamil, the situation is more complex due to the poor performance of multilingual or language-specific models and lack of proper benchmark datasets. Based on this shared task “Offensive Language Identification in Dravidian Languages” at EACL 2021; we present an exhaustive exploration of different transformer models, We also provide a genetic algorithm technique for ensembling different models. Our ensembled models trained separately for each language secured the first position in Tamil, the second position in Kannada, and the first position in Malayalam sub-tasks. The models and codes are provided.

Memes act as a medium to carry one’s feelings, cultural ideas, or practices by means of symbols, imitations, or simply images. Whenever social media is involved, hurting the feelings of others and abusing others are always a problem. Here we are proposing a system, that classifies the memes into abusive/offensive memes and neutral ones. The work involved classifying the images into offensive and non-offensive ones. The system implements resnet-50, a deep residual neural network architecture.

This paper describes the result of team-Maoqin at DravidianLangTech-EACL2021. The provided task consists of three languages(Tamil, Malayalam, and Kannada), I only participate in one of the language task-Malayalam. The goal of this task is to identify offensive language content of the code-mixed dataset of comments/posts in Dravidian Languages (Tamil-English, Malayalam-English, and Kannada-English) collected from social media. This is a classification task at the comment/post level. Given a Youtube comment, systems have to classify it into Not-offensive, Offensive-untargeted, Offensive-targeted-individual, Offensive-targeted-group, Offensive-targeted-other, or Not-in-indented-language. I use the transformer-based language model with BiGRU-Attention to complete this task. To prove the validity of the model, I also use some other neural network models for comparison. And finally, the team ranks 5th in this task with a weighted average F1 score of 0.93 on the private leader board.

In this paper, we introduce the system for the task of meme classification for Tamil, submitted by our team. In today’s society, social media has become an important platform for people to communicate. We use social media to share information about ourselves and express our views on things. It has gradually developed a unique form of emotional expression on social media – meme. The meme is an expression that is often ironic. This also gives the meme a unique sense of humor. But it’s not just positive content on social media. There’s also a lot of offensive content. Meme’s unique expression makes it often used by some users to post offensive content. Therefore, it is very urgent to detect the offensive content of the meme. Our team uses the natural language processing method to classify the offensive content of the meme. Our team combines the BERT model with the CNN to improve the model’s ability to collect statement information. Finally, the F1-score of our team in the official test set is 0.49, and our method ranks 5th.

With the advent of social media, we have seen a proliferation of data and public discourse. Unfortunately, this includes offensive content as well. The problem is exacerbated due to the sheer number of languages spoken on these platforms and the multiple other modalities used for sharing offensive content (images, gifs, videos and more). In this paper, we propose a multilingual ensemble-based model that can identify offensive content targeted against an individual (or group) in low resource Dravidian language. Our model is able to handle code-mixed data as well as instances where the script used is mixed (for instance, Tamil and Latin). Our solution ranked number one for the Malayalam dataset and ranked 4th and 5th for Tamil and Kannada, respectively.

In the past few years, the meme has become a new way of communication on the Internet. As memes are in images forms with embedded text, it can quickly spread hate, offence and violence. Classifying memes are very challenging because of their multimodal nature and region-specific interpretation. A shared task is organized to develop models that can identify trolls from multimodal social media memes. This work presents a computational model that we developed as part of our participation in the task. Training data comes in two forms: an image with embedded Tamil code-mixed text and an associated caption. We investigated the visual and textual features using CNN, VGG16, Inception, m-BERT, XLM-R, XLNet algorithms. Multimodal features are extracted by combining image (CNN, ResNet50, Inception) and text (Bi-LSTM) features via early fusion approach. Results indicate that the textual approach with XLNet achieved the highest weighted f_1-score of 0.58, which enable our model to secure 3rd rank in this task.

In this paper we present our submission for the EACL 2021-Shared Task on Offensive Language Identification in Dravidian languages. Our final system is an ensemble of mBERT and XLM-RoBERTa models which leverage task-adaptive pre-training of multilingual BERT models with a masked language modeling objective. Our system was ranked 1st for Kannada, 2nd for Malayalam and 3rd for Tamil.

Social networks made a huge impact in almost all fields in recent years. Text messaging through the Internet or cellular phones has become a major medium of personal and commercial communication. Everyday we have to deal with texts, emails or different types of messages in which there are a variety of attacks and abusive phrases. It is the moderator’s decision which comments to remove from the platform because of violations and which ones to keep but an automatic software for detecting abusive languages would be useful in recent days. In this paper we describe an automatic offensive language identification from Dravidian languages with various machine learning algorithms. This is work is shared task in DravidanLangTech-EACL2021. The goal of this task is to identify offensive language content of the code-mixed dataset of comments/posts in Dravidian Languages ( (Tamil-English, Malayalam-English, and Kannada-English)) collected from social media. This work explains the submissions made by SSNCSE_NLP in DravidanLangTech-EACL2021 Code-mix tasks for Offensive language detection. We achieve F1 scores of 0.95 for Malayalam, 0.7 for Kannada and 0.73 for task2-Tamil on the test-set.

Offensive language identification has been an active area of research in natural language processing. With the emergence of multiple social media platforms offensive language identification has emerged as a need of the hour. Traditional offensive language identification models fail to deliver acceptable results as social media contents are largely in multilingual and are code-mixed in nature. This paper tries to resolve this problem by using IndicBERT and BERT architectures, to facilitate identification of offensive languages for Kannada-English, Malayalam-English, and Tamil-English code-mixed language pairs extracted from social media. The presented approach when evaluated on the test corpus provided precision, recall, and F1 score for language pair Kannada-English as 0.62, 0.71, and 0.66, respectively, for language pair Malayalam-English as 0.77, 0.43, and 0.53, respectively, and for Tamil-English as 0.71, 0.74, and 0.72, respectively.

This paper describes the models submitted by the team MUCS for Offensive Language Identification in Dravidian Languages-EACL 2021 shared task that aims at identifying and classifying code-mixed texts of three language pairs namely, Kannada-English (Kn-En), Malayalam-English (Ma-En), and Tamil-English (Ta-En) into six predefined categories (5 categories in Ma-En language pair). Two models, namely, COOLI-Ensemble and COOLI-Keras are trained with the char sequences extracted from the sentences combined with words as features. Out of the two proposed models, COOLI-Ensemble model (best among our models) obtained first rank for Ma-En language pair with 0.97 weighted F1-score and fourth and sixth ranks with 0.75 and 0.69 weighted F1-score for Ta-En and Kn-En language pairs respectively.

The paper aims to classify different offensive content types in 3 code-mixed Dravidian language datasets. The work leverages existing state of the art approaches in text classification by incorporating additional data and transfer learning on pre-trained models. Our final submission is an ensemble of an AWD-LSTM based model along with 2 different transformer model architectures based on BERT and RoBERTa. We achieved weighted-average F1 scores of 0.97, 0.77, and 0.72 in the Malayalam-English, Tamil-English, and Kannada-English datasets ranking 1st, 2nd, and 3rd on the respective shared-task leaderboards.

Social media are interactive platforms that facilitate the creation or sharing of information, ideas or other forms of expression among people. This exchange is not free from offensive, trolling or malicious contents targeting users or communities. One way of trolling is by making memes. A meme is an image or video that represents the thoughts and feelings of a specific audience. The challenge of dealing with memes is that they are region-specific and their meaning is often obscured in humour or sarcasm. A meme is a form of media that spreads an idea or emotion across the internet. The multi modal nature of memes, postings of hateful memes or related events like trolling, cyberbullying are increasing day by day. Memes make it even more challenging since they express humour and sarcasm in an implicit way, because of which the meme may not be offensive if we only consider the text or the image. In this paper we proposed a approach for meme classification for Tamil language that considers only the text present in the meme. This work explains the submissions made by SSNCSE NLP in DravidanLangTechEACL2021 task for meme classification in Tamil language. We achieve F1 scores of 0.50 using the proposed approach using the test-set.

Dravidian language family is one of the largest language families in the world. In spite of its uniqueness, Dravidian languages have gained very less attention due to scarcity of resources to conduct language technology tasks such as translation, Parts-of-Speech tagging, Word Sense Disambiguation etc,. In this paper, we, team MUCS, describe sequence-to-sequence stacked Long Short Term Memory (LSTM) based Neural Machine Translation (NMT) model submitted to “Machine Translation in Dravidian languages”, a shared task organized by EACL-2021. The NMT model was applied on translation using English-Tamil, EnglishTelugu, English-Malayalam and Tamil-Telugu corpora provided by the organizers. Standard evaluation metrics namely Bilingual Evaluation Understudy (BLEU) and human evaluations are used to evaluate the model. Our models exhibited good accuracy for all the language pairs and obtained 2nd rank for TamilTelugu language pair.

In the last few decades, Code-Mixed Offensive texts are used penetratingly in social media posts. Social media platforms and online communities showed much interest on offensive text identification in recent years. Consequently, research community is also interested in identifying such content and also contributed to the development of corpora. Many publicly available corpora are there for research on identifying offensive text written in English language but rare for low resourced languages like Tamil. The first code-mixed offensive text for Dravidian languages are developed by shared task organizers which is used for this study. This study focused on offensive language identification on code-mixed low-resourced Dravidian language Tamil using four classifiers (Support Vector Machine, random forest, k- Nearest Neighbour and Naive Bayes) using chiˆ2 feature selection technique along with BoW and TF-IDF feature representation techniques using different combinations of n-grams. This proposed model achieved an accuracy of 76.96% while using linear SVM with TF-IDF feature representation technique.

This paper describes our submission to shared task on Meme Classification for Tamil Language. To address this task, we explore a multimodal transformer for meme classification in Tamil language. According to the characteristics of the image and text, we use different pretrained models to encode the image and text so as to get better representations of the image and text respectively. Besides, we design a multimodal attention layer to make the text and corresponding image interact fully with each other based on cross attention. Our model achieved 0.55 weighted average F1 score and ranked first in this task.

Internet advancements have made a huge impact on the communication pattern of people and their life style. People express their opinion on products, politics, movies etc. in social media. Even though, English is predominantly used, nowadays many people prefer to tweet in their native language and some- times by combining it with English. Sentiment analysis on such code-mixed tweets is challenging, due to large vocabulary, grammar and colloquial usage of many words. In this paper, the transformer based language model is applied to analyse the sentiment on Tanglish tweets, which is a combination of Tamil and English. This work has been submitted to the the shared task on DravidianLangTech- EACL2021. From the experimental results, it is shown that an F 1 score of 64% was achieved in detecting the hate speech in code-mixed Tamil-English tweets using bidirectional trans- former model.