Meitu Internet Technology Salon: AI and Machine Learning Applications in Practice

Artificial Intelligence and Machine Learning represent the most trending and popular topics globally, with international tech giants accelerating their AI industrialization布局. This report covers the 4th Meitu Internet Technology Salon held on March 25, 2017 in Xiamen, focusing on AI and ML applications.

Meipai: Community Text Mining and Practice

As a leading short video and live streaming platform, Meipai generates millions of data entries daily, resulting in massive text data accumulation. The technical team presented their work in three modules: text anti-spam and fraud detection, text mining and exploration, and Meipai video search evolution.

For combating spam attacks, the team adopted LCS (Longest Common Subsequence) and LRU (Least Recently Used) sequence methods for identification and decision-making, implementing a more systematic text anti-spam framework with anomaly detection and hierarchical processing. Through systematic construction using machine learning, the team can now effectively detect and analyze spam and fraudulent content, minimizing user spam impact on the platform.

Additional developments include: a hot topic identification model specifically for Meipai scenarios; sentiment classification for comments to explicitly analyze user sentiment polarity towards videos; and optimized word segmentation with new word discovery.

The team also shared insights on Meipai's in-site search evolution - from early strategic search models to high-conversion models, eventually evolving into a comprehensive community search model with personalized features and environmental community characteristics.

Baidu: Machine Learning Technology in Business Intelligence

ML-based intelligent tools can automatically integrate, mine, and learn multi-source heterogeneous data across different application scenarios, providing scientific and objective decision support. Baidu's senior data scientist introduced the background and opportunities of BI data mining research in the AI era, sharing research achievements in mobile internet user profiling, real estate, sports, and enterprise management.

Business Intelligence is divided into three phases: manual data analysis in the 1970s-80s, commercial reporting software (like IDM), and the current phase using AI, big data, and cloud computing. The current goal is to discover hidden valuable information from big data to support better business decisions.

Current research focuses on AI technology for intelligent enterprise management and human resources management, including automated analysis of company recruitment status, requirements, and themes using machine learning methods.