This paper presents a transformer-based model for recognizing emotions in Hindi-English code-mixed conversations, adhering to the SemEval task constraints. Leveraging BERT-based transformers, we fine-tune pre-trained models on the dataset, incorporating tokenization and attention mechanisms. Our approach achieves competitive performance (weighted F1-score of 0.4), showcasing the effectiveness of BERT in nuanced emotion analysis tasks within code-mixed conversational contexts.

Semantic Textual Relatedness (STR) will provide insight into the limitations of existing models and support ongoing work on semantic representations. Track A in Shared Task-1, provides pairs of sentences with semantic relatedness scores for 9 languages out of which 7 are low-resources. These languages are from four different language families. We developed models for 8 languages (except for Amharic) in Track A, using Sentence Transformers (SBERT) architecture, and fine-tuned them with multilingual and monolingual pre-trained language models (PLM). Our models for English (eng), Algerian Arabic (arq), andKinyarwanda (kin) languages were ranked 12, 5, and 8 respectively. Our submissions are ranked 5th among 40 submissions in Track A with an average Spearman correlation score of 0.74. However, we observed that the usage of monolingual PLMs did not guarantee better than multilingual PLMs in Marathi (mar), and Telugu (tel) languages in our case.

In this paper, we studied the gender bias in monolingual word embeddings of two Indian languages Hindi and Tamil. Tamil is one of the classical languages of India from the Dravidian language family. In Indian society and culture, instead of racism, a similar type of discrimination called casteism is against the subgroup of peoples representing lower class or Dalits. The word embeddings measurement to evaluate bias using the WEAT score reveals that the embeddings are biased with gender and casteism which is in line with the common stereotypical human biases.

Data in general encodes human biases by default; being aware of this is a good start, and the research around how to handle it is ongoing. The term ‘bias’ is extensively used in various contexts in NLP systems. In our research the focus is specific to biases such as gender, racism, religion, demographic and other intersectional views on biases that prevail in text processing systems responsible for systematically discriminating specific population, which is not ethical in NLP. These biases exacerbate the lack of equality, diversity and inclusion of specific population while utilizing the NLP applications. The tools and technology at the intermediate level utilize biased data, and transfer or amplify this bias to the downstream applications. However, it is not enough to be colourblind, gender-neutral alone when designing a unbiased technology – instead, we should take a conscious effort by designing a unified framework to measure and benchmark the bias. In this paper, we recommend six measures and one augment measure based on the observations of the bias in data, annotations, text representations and debiasing techniques.

Assessing the funniness of edited news headlines task deals with estimating the humorness in the headlines edited with micro-edits. This task has two sub-tasks in which one has to calculate the mean predicted score of humor level and other deals with predicting the best funnier sentence among given two sentences. We have calculated the humorness level using microtc and predicted the funnier sentence using microtc, universal sentence encoder classifier, many other traditional classifiers that use the vectors formed with universal sentence encoder embeddings, sentence embeddings and majority algorithm within these approaches. Among these approaches, microtc with 6 folds, 24 processes and 3 folds, 36 processes achieve the least Root Mean Square Error for development and test set respectively for subtask 1. For subtask 2, Universal sentence encoder classifier achieves the highest accuracy for development set and Multi-Layer Perceptron applied on vectors vectorized using universal sentence encoder embeddings for the test set.

Emotion identification is a process of identifying the emotions automatically from text, speech or images. Emotion identification from textual conversations is a challenging problem due to absence of gestures, vocal intonation and facial expressions. It enables conversational agents, chat bots and messengers to detect and report the emotions to the user instantly for a healthy conversation by avoiding emotional cues and miscommunications. We have adopted a Seq2Seq deep neural network to identify the emotions present in the text sequences. Several layers namely embedding layer, encoding-decoding layer, softmax layer and a loss layer are used to map the sequences from textual conversations to the emotions namely Angry, Happy, Sad and Others. We have evaluated our approach on the EmoContext@SemEval2019 dataset and we have obtained the micro-averaged F1 scores as 0.595 and 0.6568 for the pre-evaluation dataset and final evaluation test set respectively. Our approach improved the base line score by 7% for final evaluation test set.

Offensive language identification (OLI) in user generated text is automatic detection of any profanity, insult, obscenity, racism or vulgarity that degrades an individual or a group. It is helpful for hate speech detection, flame detection and cyber bullying. Due to immense growth of accessibility to social media, OLI helps to avoid abuse and hurts. In this paper, we present deep and traditional machine learning approaches for OLI. In deep learning approach, we have used bi-directional LSTM with different attention mechanisms to build the models and in traditional machine learning, TF-IDF weighting schemes with classifiers namely Multinomial Naive Bayes and Support Vector Machines with Stochastic Gradient Descent optimizer are used for model building. The approaches are evaluated on the OffensEval@SemEval2019 dataset and our team SSN_NLP submitted runs for three tasks of OffensEval shared task. The best runs of SSN_NLP obtained the F1 scores as 0.53, 0.48, 0.3 and the accuracies as 0.63, 0.84 and 0.42 for the tasks A, B and C respectively. Our approaches improved the base line F1 scores by 12%, 26% and 14% for Task A, B and C respectively.