Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MixSarc: A Bangla-English Code-Mixed Corpus for Implicit Meaning Identification

Kazi Samin Yasar Alam, Md Tanbir Chowdhury, Tamim Ahmed, Ajwad Abrar, Md Rafid Haque

TL;DR

MixSarc is introduced, the first publicly available Bangla-English code-mixed corpus for implicit meaning identification and provides a foundational resource for culturally aware NLP and supports more reliable multi-label modeling in code-mixed environments.

Abstract

Bangla-English code-mixing is widespread across South Asian social media, yet resources for implicit meaning identification in this setting remain scarce. Existing sentiment and sarcasm models largely focus on monolingual English or high-resource languages and struggle with transliteration variation, cultural references, and intra-sentential language switching. To address this gap, we introduce MixSarc, the first publicly available Bangla-English code-mixed corpus for implicit meaning identification. The dataset contains 9,087 manually annotated sentences labeled for humor, sarcasm, offensiveness, and vulgarity. We construct the corpus through targeted social media collection, systematic filtering, and multi-annotator validation. We benchmark transformer-based models and evaluate zero-shot large language models under structured prompting. Results show strong performance on humor detection but substantial degradation on sarcasm, offense, and vulgarity due to class imbalance and pragmatic complexity. Zero-shot models achieve competitive micro-F1 scores but low exact match accuracy. Further analysis reveals that over 42\% of negative sentiment instances in an external dataset exhibit sarcastic characteristics. MixSarc provides a foundational resource for culturally aware NLP and supports more reliable multi-label modeling in code-mixed environments.

MixSarc: A Bangla-English Code-Mixed Corpus for Implicit Meaning Identification

TL;DR

MixSarc is introduced, the first publicly available Bangla-English code-mixed corpus for implicit meaning identification and provides a foundational resource for culturally aware NLP and supports more reliable multi-label modeling in code-mixed environments.

Abstract

Bangla-English code-mixing is widespread across South Asian social media, yet resources for implicit meaning identification in this setting remain scarce. Existing sentiment and sarcasm models largely focus on monolingual English or high-resource languages and struggle with transliteration variation, cultural references, and intra-sentential language switching. To address this gap, we introduce MixSarc, the first publicly available Bangla-English code-mixed corpus for implicit meaning identification. The dataset contains 9,087 manually annotated sentences labeled for humor, sarcasm, offensiveness, and vulgarity. We construct the corpus through targeted social media collection, systematic filtering, and multi-annotator validation. We benchmark transformer-based models and evaluate zero-shot large language models under structured prompting. Results show strong performance on humor detection but substantial degradation on sarcasm, offense, and vulgarity due to class imbalance and pragmatic complexity. Zero-shot models achieve competitive micro-F1 scores but low exact match accuracy. Further analysis reveals that over 42\% of negative sentiment instances in an external dataset exhibit sarcastic characteristics. MixSarc provides a foundational resource for culturally aware NLP and supports more reliable multi-label modeling in code-mixed environments.
Paper Structure (55 sections, 1 equation, 4 figures, 5 tables, 2 algorithms)

This paper contains 55 sections, 1 equation, 4 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Code-Mixed Language Processing
  4. Sentiment Analysis in Code-Mixed Text
  5. Sarcasm and Humor Detection
  6. Offensive and Vulgar Language Detection
  7. MixSarc Dataset
  8. Data Sourcing
  9. Data Cleaning and Preprocessing
  10. Emoji and Non-Textual Cue Removal
  11. Script-Based Filtering
  12. Code-Mixed Validation via mBERT
  13. Data Annotation
  14. Annotation Scheme
  15. Annotators
  16. ...and 40 more sections

Figures (4)

  • Figure 1: Examples of Humorous, Sarcastic, Offensive, and Vulgar utterances from the MixSarc dataset. Each example is presented in its original Bangla–English code-mixed form followed by an English translation. Red represents English words, blue represents Bengali words written in English alphabets.
  • Figure 2: Overview of the dataset preparation pipeline used in this work.
  • Figure 3: Zero-shot prompt used for LLM-based multi-label classification
  • Figure 4: Distribution of sarcastic vs. genuine negative samples within the negative sentiment class of BnSentMix.