We propose a region-aware hybrid retrieval framework for culturally grounded multilingual question answering. Our system combines BM25-based lexical matching with dense semantic similarity using sentence embeddings, integrating both signals into a unified ranking function. To further prioritize culturally relevant evidence, we introduce a regional weighting heuristic that boosts documents containing explicit region-specific references. The top-ranked evidence passages are incorporated into a structured prompt and processed by a 4-bit quantized Qwen3-14B model. Instead of generating free-form text, the model selects answers deterministically using a logit-based scoring mechanism over the four multiple-choice options. This design enables efficient inference while improving cross-lingual stability, particularly in culturally explicit contexts.

This paper presents our system, Homa, for SemEval-2025 Task 5: Subject Tagging, which focuses on automatically assigning subject labels to technical records from TIBKAT using the Gemeinsame Normdatei (GND) taxonomy. We leverage OntoAligner, a modular ontology alignment toolkit, to address this task by integrating retrieval-augmented generation (RAG) techniques. Our approach formulates the subject tagging problem as an alignment task, where records are matched to GND categories based on semantic similarity. We evaluate OntoAligner’s adaptability for subject indexing and analyze its effectiveness in handling multilingual records. Experimental results demonstrate the strengths and limitations of this method, highlighting the potential of alignment techniques for improving subject tagging in digital libraries.

Toxic Spans Detection(TSD) task is defined as highlighting spans that make a text toxic. Many works have been done to classify a given comment or document as toxic or non-toxic. However, none of those proposed models work at the token level. In this paper, we propose a self-attention-based bidirectional gated recurrent unit(BiGRU) with a multi-embedding representation of the tokens. Our proposed model enriches the representation by a combination of GPT-2, GloVe, and RoBERTa embeddings, which led to promising results. Experimental results show that our proposed approach is very effective in detecting span tokens.