Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{206906,
        author = {Rohit Pise},
        title = {Context-Aware Multimodal RAG Frameworks for Regional Real Estate Markets Using Dialect-Specific LLMs},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {2},
        pages = {3488-3494},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=206906},
        abstract = {Regional real estate markets are characterized by fragmented listing sources, heterogeneous documentation (floor plans, brochures, registry excerpts, photographs), and buyer-seller communication that occurs predominantly in local dialects rather than standard national languages. Conventional retrieval-augmented generation (RAG) pipelines, built around monolingual text retrieval and generic embedding models, degrade sharply under these conditions because they cannot align code-mixed or dialectal queries with formally written property metadata, nor can they reason jointly over text and non-text artifacts such as site images, floor plans, and geospatial layers. This paper proposes Context-Aware Multimodal RAG (CAM-RAG), a framework that combines (i) a multimodal retrieval layer that jointly indexes textual listings, structured metadata, images, and geospatial context; (ii) a dialect-adaptation layer that normalizes regional and code-mixed queries prior to retrieval while preserving semantic intent; and (iii) a context-aware fusion and generation module that grounds large language model (LLM) outputs in retrieved evidence to reduce hallucination in price, location, and legal-status claims. We motivate the design against prior work in retrieval-augmented generation, multimodal retrieval, and low-resource/dialectal language modelling, and we present a concrete system architecture, a staged fine-tuning methodology for dialect adaptation, and an evaluation protocol comprising retrieval-quality, generation-faithfulness, and dialect-robustness metrics. We further outline a data governance and bias-auditing procedure appropriate for a regulated domain such as property transactions. The contribution of this paper is architectural and methodological: it defines a reproducible framework and evaluation protocol intended to be instantiated and empirically validated on region-specific real estate corpora in follow-up work, rather than reporting closed empirical results.},
        keywords = {Retrieval-Augmented Generation, Multimodal Retrieval, Dialect-Specific Language Models, Real Estate Information Systems, Context-Aware AI, Low-Resource NLP, PropTech},
        month = {July},
        }