Literature Survey
Digital marketing has become essential for SMEs because online platforms allow businesses to reach customers, promote products, and measure engagement at lower cost than traditional marketing. However, Sri Lankan SMEs often lack the technical knowledge, marketing expertise, and intelligent decision-support tools required to use digital channels effectively.
Existing research highlights the importance of social media marketing, AI-based decision support, natural language processing, customer engagement automation, campaign prediction, explainable AI, and localized language processing. The literature also shows that Sinhala and Sinhala-English code-mixed communication require special handling because Sinhala is a low-resource language in NLP and Sinhala text rendering has unique Unicode and shaping challenges.
Research Gap
Most existing AI marketing platforms are designed for global English-dominant markets and do not fully support the linguistic, cultural, and operational needs of Sri Lankan SMEs. Available tools usually focus on only one function, such as content generation, chatbot support, campaign analytics, or strategy planning. They rarely provide an integrated solution that combines Sinhala-English content generation, poster creation, campaign performance prediction, RAG-based customer support, personalized strategy generation, explainability, and SME-friendly usability.
There is also a lack of affordable AI platforms that support Sinhala-aware poster rendering, explainable campaign predictions, tenant-specific business support, and personalized strategy maintenance using fresh digital marketing knowledge.
Research Problem
How can an AI-powered digital marketing optimization platform be designed and developed to help Sri Lankan SMEs improve customer engagement, content creation, campaign planning, and marketing strategy generation using localized, affordable, explainable, and easy-to-use AI technologies?
Main Objective
The main objective of this research is to design, develop, and evaluate an AI-powered digital marketing optimization platform for Sri Lankan SMEs that integrates Sinhala-English marketing content generation, customer engagement automation, campaign performance prediction, and personalized marketing strategy recommendation.
Specific Objectives
- To develop a Sinhala and English AI content generation component for SME marketing.
- To create a Sinhala-aware poster generation workflow with correct Unicode rendering.
- To develop a smart customer engagement assistant using RAG and tenant-specific business data.
- To predict social media campaign performance before publication using machine learning and deep learning.
- To provide explainable campaign prediction using SHAP and LIME.
- To generate personalized digital marketing strategies using SME profile data and RAG.
- To provide confidence indicators and drift detection for strategy reliability.
- To support SMEs with an affordable, practical, and localized digital marketing platform.
Methodology
The project follows an applied design science and agile development methodology. The system was divided into four main AI components and developed iteratively. Each component was designed, implemented, tested, and evaluated based on its functional and non-functional requirements.
The methodology includes requirement analysis, literature review, dataset preparation, AI model selection, system architecture design, component implementation, API and frontend integration, functional testing, non-functional testing, evaluation and result analysis, and commercialization analysis.
Technologies Used
HTML
CSS
JavaScript
React
Node.js
Python
FastAPI
MongoDB
PostgreSQL
ChromaDB
pgvector
RAG
LLM
Gemini API
Groq API
Qwen2.5
LLaMA
GPT
Transformer
BiLSTM
BiGRU
SHAP
LIME
n8n
Playwright
Chromium
HarfBuzz
Machine Learning
Deep Learning
Natural Language Processing
