Personalization Recommendation & User Understanding
• Core member to build Yahoo! Content Personalization & Understanding platform, focus on machine learning based ranking solutions of personalized recommendation for Yahoo! homepage and multiple channels.
• Continuous contribute to user engagement improvements, including user understanding by behavior modeling and preference modeling, ranking signal collection, ranking model improvement, ranking online evaluation and offline data analysis. Optimizations over the past two years, represented by GBDT ranking model, long-term Sparse Polarity user profile and click feed back based short term signal, bring over significant metrics lift to Yahoo! homepage.
Yahoo Search Technology
Focus on web/local search ranking and runtime serving for Yahoo mobile search. Contribute to the design and implementation of clicksimilarity feature and clicktext feature which are the most important features of yahoo search; Build up local search module and bring significant improvement to explicit/implicit local intent detection; Improving search model interpretation, rule-based model, GBDT feature importance improvement, etc.
Natural Language Understanding in Yahoo Chatbot Platform
I am the core member to build the first version Yahoo chat bot platform, focusing on NLU part. Drive the design and implementation of intent classification, CRF based slot detection and active learning pipeline. Thousands of bots have been built and launched based on this platform in various domains by our customers, with very limited bot developer efforts.
