@inproceedings{s-etal-2026-wise,
    title = "Wise@{D}ravidian{L}ang{T}ech 2026: Dialect-Aware {T}amil Speech Classification and Recognition via Cross-Pipeline Embedding Transfer",
    author = "S, Ganesh Sundhar  and
      N, Hari Krishnan  and
      G, Gnanasabesan  and
      KP, Suriya  and
      G, Jyothish Lal",
    editor = "Chakravarthi, Bharathi Raja  and
      Priyadharshini, Ruba  and
      Madasamy, Anand Kumar  and
      Thavareesan, Sajeetha  and
      Rajiakodi, Saranya  and
      Navaneethakrishnan, Subalalitha  and
      Chinnappa, Dhivya  and
      Palani, Balasubramanian  and
      Subramanian, Malliga  and
      Shanmugavadivel, Kogilavani  and
      Rajalakshmi, Ratnavel",
    booktitle = "Proceedings of the Sixth Workshop on Speech, Vision, and Language Technologies for {D}ravidian Languages",
    month = jul,
    year = "2026",
    address = "Underline (Virtual)",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.dravidianlangtech-1.71/",
    pages = "447--452",
    ISBN = "979-8-89176-401-9",
    abstract = "This paper presents the **Wise** system for the shared task on dialect-based speech processing in Tamil, addressing two subtasks: **(1) four-way dialect region classification** (Northern, Southern, Western, Central), and **(2) dialectal Tamil ASR**. All audio is preprocessed using loudness normalization followed by neural denoising to ensure consistent audio quality for downstream models. For classification, we experiment with different model variants combining multilingual and Tamil-pretrained **Wav2Vec2** backbones with five temporal pooling strategies under frozen and partial fine-tuning settings. Our best configuration, i.e., learned attentive pooling with partial fine-tuning and a differentially trained MLP head, achieves a macro F1 of **0.79**, securing **1st place** with a margin of **0.26** points. For ASR, we propose two novel **dialect-conditioned Whisper** architectures{---}residual injection and cross-attention{---}that inject dialect embeddings from the trained classifier into the ASR pipeline. In addition, we evaluate a vanilla Whisper-Tamil fine-tuned baseline. The best model achieved a **WER of 0.90**, securing **8th place** in the shared task."
}