ENGINE / 02

Reading the mood of a nation in three languages.

Trilingual NLP across Tamil, Sinhala, and English social streams, mapped to district-level political heat in near real time.

Corpus
Tamil · Sinhala · English

Trilingual Data Ingestion

  • Social media firehose: Twitter/X, Facebook, TikTok, YouTube comments
  • News article scraping across 40+ Sri Lankan publications
  • WhatsApp and Telegram public-channel monitoring (anonymised)
  • Forum and blog sentiment from diaspora communities
NLP Engine
Custom fine-tuned models

Language Understanding

  • Transformer-based sentiment classifiers fine-tuned on Sri Lankan political discourse
  • Named entity recognition for politicians, parties, policies, and regions
  • Code-switching detection: seamless handling of Tamil-English and Sinhala-English mixed text
  • Sarcasm and irony detection calibrated to South Asian rhetorical patterns
Geo-Mapping
Real-time spatial overlay

District-Level Heat Maps

  • Geotagged content mapped to 160+ administrative districts
  • Sentiment velocity: rate of change in public opinion by region
  • Topic clustering: which issues dominate which districts
  • Anomaly detection: sudden sentiment spikes flagged for campaign response
Delivery
Dashboard + alerts

Operator Interface

  • Live dashboard with filterable sentiment timelines by region, language, topic
  • Automated daily briefings for campaign strategists
  • Push alerts on sentiment breakpoints (threshold-triggered)
  • Exportable reports for coalition meetings and press strategy
Back to AAMA Home