DravidianCodeMix: sentiment analysis and offensive language identification dataset for Dravidian languages in code-mixed text

Abstract

This paper describes the development of a multilingual, manually annotated dataset for three under-resourced Dravidian languages generated from social media comments. The dataset was annotated for sentiment analysis and offensive language identification for a total of more than 60,000 YouTube comments. The dataset consists of around 44,000 comments in Tamil-English, around 7000 comments in Kannada-English, and around 20,000 comments in Malayalam-English. The data was manually annotated by volunteer annotators and has a high inter-annotator agreement in Krippendorff’s alpha. The dataset contains all types of code-mixing phenomena since it comprises user-generated content from a multilingual country. We also present baseline experiments to establish benchmarks on the dataset using machine learning and deep learning methods. The dataset is available on Github and Zenodo.

Adapter-based fine-tuning of pre-trained multilingual language models for code-mixed and code-switched text classification

“…For classifying CMCS data, Machine Learning approaches such as Logistic Regression, Support Vector Machines, Multinomial Naive Bayes, K-Nearest Neighbors, Decision Trees, and Random Forest have been used by early research (Chakravarthi et al, 2022). Later, Deep Learning techniques such as CNN, LSTM, and BiLSTM (Kamble and Joshi, 2018) have been widely used.…”

“…CMCS can appear in various forms, including code-switching, inter-sentential, and intra-sentential code-mixing, and texts written in both Latin and native scripts. CMCS text classification corpora have been mainly created with respect to Indian languages such as Hindi-English (Bohra et al, 2018), Telugu-English (Gundapu and Mamidi, 2018), Tamil-English, Kannada-English, and Malayalam-English (Chakravarthi et al, 2022), while there are some corpora for other CMCS languages such as Sinhala-English (Smith and Thayasivam, 2019), Spanish-English (Vilares et al, 2016) and Arabic-English (Sabty et al, 2019). Except for the dataset created by Chakravarthi et al (2022), others have removed the text written in native script and considered only a limited type of code-mixing levels.…”

“…CMCS text classification corpora have been mainly created with respect to Indian languages such as Hindi-English (Bohra et al, 2018), Telugu-English (Gundapu and Mamidi, 2018), Tamil-English, Kannada-English, and Malayalam-English (Chakravarthi et al, 2022), while there are some corpora for other CMCS languages such as Sinhala-English (Smith and Thayasivam, 2019), Spanish-English (Vilares et al, 2016) and Arabic-English (Sabty et al, 2019). Except for the dataset created by Chakravarthi et al (2022), others have removed the text written in native script and considered only a limited type of code-mixing levels. Chakravarthi et al (2022) have considered all types of codemixing phenomena including inter-sentential and intra-sentential code-mixing, including texts written in both Tamil and Latin scripts.…”

“…Except for the dataset created by Chakravarthi et al (2022), others have removed the text written in native script and considered only a limited type of code-mixing levels. Chakravarthi et al (2022) have considered all types of codemixing phenomena including inter-sentential and intra-sentential code-mixing, including texts written in both Tamil and Latin scripts. However, most of the studies have annotated their datasets for limited types of classification tasks as given in Table 2.…”

Benchmarking Multi-Task Learning for Sentiment Analysis and Offensive Language Identification in Under-Resourced Dravidian Languages

“…We make use of the multi-lingual dataset, Dravidian-CodeMix 2 [57], consisting of over 60,000 manually annotated YouTube comments. The data set comprises sentences from 3 code-mixed Dravidian languages: Kannada [58], Malayalam [14], and Tamil [6,43].…”