Industry - Bing Search Relevance and Index Quality

Role: Data Auditor Lead, Project Coordinator

Overview

Bing Search Relevance Data Auditing is a project that its main purpose is to ensure the relevance data quality is good so that it can be useful for machine learning and further improve Bing's relevance. 

KPI Achievements

- I led the NDCG team increased data accuracy for nearly 25%. Overall NDCG score increased almost 20%, reached its peak in Bing. 

- I led the Index Quality team increased data accuracy for 13%. 

What I did?

- Bing China/Japan Relevance NDCG

​I was responsible for the NDCG score of Bing CN and JP market. To make sure the score is good every week, I led a team to review Bing's relevance data on a weekly basis and find out on which scenario (such as freshness, relevance, authority etc) Bing still needs to improve. 

​I was also responsible for optimize the guideline of relevance measurement, communication with data auditors across the globe (mostly UK, US, Brazil) to discuss each market's performance and solution to solve problems we were facing as well as training and managing offsite vendors. 

- Bing China/Japan Index Quality

​I was responsible for the Index data quality of Bing CN and JP market. I led a team to review Bing's index data on a weekly basis and find out new categories and websites of either spam or junk websites within Bing's index. 

To make that happen, I was also responsible for updating the Index Quality guideline for Bing Spam and Junk categories. Once updated majorly, I also need to doing training for PMs and Devs to keep eveyrone on the same page. 

Industry - Bing Data Projects

Role: Data Auditor Lead, Project Coordinator

Overview

During my 2 years of work at Bing. I was also in charge of more than 60 ad hoc data projects. Each data project is usually different from the others and have different goals. Things i need to consider for each project include: Requirement, project goal, project budget, ETA, guideline, vendor resource, vendor training, raw data quality, labeled data quality, review quality and delivery.