Скачать или смотреть Unstructured, Under Control Classification at Scale

הרצאה זו היא חלק ממיטאפ משותף עם חברת Cyera וקהילת MDLI
אתם יכולים לצפות בשאר ההרצאות ובמצגות פה: https://machinelearning.co.il/lp-even...


Enterprises are drowning in unstructured data – documents, logs, emails, code, and more- spread across IaaS, PaaS, SaaS, and on-prem systems. Cyera’s Data Security Posture Management (DSPM) delivers a comprehensive understanding of sensitive data extracted from unstructured files across the entire data estate, providing detection and assessment of data security risks, enabling companies to take proactive actions and eliminate risks. Entity extraction is not a new problem and has been around for decades, yet off-the-shelf models rarely perform well enough on messy, domain-specific enterprise data. In this talk, we unpack how we cracked entity extraction from unstructured data. To this end, we developed a production‐grade, AI‐powered classification pipeline at scale, leveraging traditional pattern‐matching techniques, a fine‐tuned open‐source LLM, and GenAI throughout the pipeline to achieve high precision and recall.