Cutting‑Edge AI from ByteDance & OPPO: Audio, NLP, and Translation

ByteDance Intelligent Audio Signal Processing Practice

Speaker: Xu Ning, Audio Signal Processing Algorithm Engineer at ByteDance.

With the proliferation of smart hardware and improved on‑device chip performance, the talk explored how intelligent audio processing meets real‑world business needs, achieving high quality, low latency, and low computational cost by combining traditional signal processing with deep learning for high‑quality audio capture, sound‑field reconstruction, and intelligent voice interaction.

NLP in OPPO Recommendation Scenarios

Speaker: Deng Zhengyu, Senior Data Mining Engineer at OPPO.

The presentation described the growing connection between NLP and recommendation, explaining how NLP builds article profiles through explicit and implicit tag extraction, and how common NLP models are transferred to recommendation tasks.

Entity Linking in OPPO Knowledge Graph and XiaoBu Assistant

Speaker: Fan Chengyuan, Senior NLP Algorithm Engineer at OPPO.

Entity linking connects text entities to a knowledge base. In OPPO’s XiaoBu Assistant, about 30% of queries contain entity terms, many novel or ambiguous. The talk showed how entity‑linking technology accurately identifies user‑referenced entities, improving responses using OPPO’s OGraph knowledge graph.

Machine Translation in the Multimedia Era at ByteDance

Speaker: Cheng Yu, Senior Algorithm Engineer at ByteDance.

The session introduced ByteDance’s research on text and multimedia machine translation, focusing on enhancing translation quality and applying the technology to multimedia scenarios.