Intelligent Writing: How AI Empowers Media Content Creation

Intelligent writing, powered by AI, is transforming media content creation by improving efficiency, organization, and personalized distribution.

With the rapid development of AI from computational to cognitive intelligence, Baidu's Knowledge Graph team focuses on integrating language understanding, knowledge graphs, and multimodal data to enable both automatic and assisted writing.

The workflow consists of three stages—topic discovery, material organization, and text generation—supported by knowledge‑driven topic mining, material retrieval, and quality control.

Automatic writing generates news, aggregation, and data‑driven articles using macro‑planning (structure), micro‑planning (lexical choices), and surface realization, relying on NLG techniques such as seq2seq, attention, pointer networks, and transformer‑based models, with knowledge graphs providing factual grounding.

Assisted writing offers pre‑writing topic recommendation, material suggestion, real‑time content hints, and post‑writing proofreading, cover generation, and quality assessment to enhance human authors' productivity.

Key challenges include maintaining coherence, factual accuracy, depth, emotional expression, and effective evaluation of assisted tools.

Current results show seconds‑level article generation, a two‑fold increase in medium‑tail content, millions of articles and billions of reads, with ongoing efforts to deepen AI capabilities and broaden industry impact.