Jilin Smart Cloud Media Assets
Jilin Radio & Television Station
One decode, every modality understood.
A smart cloud media-asset platform driven by a one-time-decode multimodal analysis engine — orchestrating vision, speech, and language AI on a domain knowledge graph to make archives instantly searchable and review-ready.
Overview
Jilin Radio & Television Station's Smart Cloud Media Asset System addresses deep media convergence with intelligent methods. Built on a scenario-based compute-scheduling framework and a domain knowledge graph, it automates the extraction and management of media material to raise content usability and security-review efficiency.
Challenge
Turning raw archives into usable, review-ready content at scale requires analyzing every modality efficiently — and orchestrating AI capabilities without long, brittle deployment cycles.
Multimodal at scale
Vision, speech, and language all had to be analyzed efficiently across large volumes of material.
Slow intelligent deployment
Combining AI capabilities into working scenarios traditionally meant long, rigid development cycles.
Finding the right content
Retrieval and recommendation had to understand content deeply enough to serve genuinely relevant results.
Solution
A one-time-decode multimodal approach fuses natural-language, computer-vision, and speech AI for efficient distributed computing. A visual orchestration framework lets teams compose AI atomic capabilities into scenarios quickly, shortening deployment. An integrated search-and-recommendation engine personalizes retrieval through behavior analysis and multimodal understanding, while news knowledge bases and a domain knowledge graph — trained across politics, economy, culture, livelihood, and foreign publicity — enable intelligent thematic aggregation.
Delivered with
The Sobey solutions and products behind this deployment — each built to work together as one chain.
Result
Jilin's archives became a smart, searchable asset platform — one decode driving every modality, faster scenario deployment, and content that surfaces itself for production and review.
AI modalities fused in a single decode — vision, speech, and language.
News knowledge domains modeled — politics, economy, culture, livelihood, foreign publicity.
Decode pass driving all multimodal analysis.