Social Media Data Capabilities

xFractal extends its Stream Engine beyond on-chain signals into the domain of social media, mainly on X (fka Twitter) and Telegram. Just like Solana’s event horizon, social emissions are constant, noisy, and directionless. The challenge is not just access, but filtration, parsing, and contextualization at scale.

We built a system that ingests and processes raw social data streams in real time, aligns them with on-chain events, and routes them into our agentic intelligence layer.

Data Ingestion

Source Diversity: Twitter/X (via TweetScout, Moni, ElfaAI, Masa Network, Twitter API, Twitter Scraper), Telegram (internal scrapers), Dexscreener metadata, Pumpfun data, and our proprietary, in-house-developed sentiment classification and NLP.
Ingestion: Continuous streaming of posts, replies, reposts, mentions, and media.
Contextual scraping: Mapping wallet-linked users, KOL clusters, and dev accounts across platforms.

Filtering & Parsing

Noise reduction: remove bot activity, spam, irrelevant chatter.
Entity linking: align usernames, wallets, and tokens into unified identifiers.
Sentiment parsing: transformer-based NLP models for token/project-specific polarity.
Topic clustering: density-based methods to detect narratives, trends, and meme propagation.
Language & Encoding Normalization: multilingual posts are translated and tokenized for model-ready inputs.

Core Components

Signal Normalizer: maps heterogeneous payloads into standardized schema.
Engagement Metrics Engine: real-time tracking of virality, diffusion, and narrative strength.
Narrative Detection Layer: clusters semantically related content to surface emerging narratives.