Audience Sciences is creating the scientific foundation for building great audience-facing properties at Yahoo!. Our mission is to understand our users by building profiles pertaining to their activities, interests, contributions, expertise, and reputation, so that we can optimize our users’ experiences and engagement.

We build user profiles to power the recommendation engine for personalized content consumption on Yahoo! Today, Yahoo! News, and user-generated content comments. We also build abuser profiles to prevent, detect, and monitor abusers, abusive content, and activities across every Yahoo! property. Our goal is to understand our users by developing new user-behavior modeling technology in data mining, machine learning, and statistical modeling to very large scale datasets.

News

WSDM 2011Best Paper Award

"Unbiased Offline Evaluation of Contextual-bandit-based News Article Recommendation Algorithms" Lihong Li; Wei Chu; John Langford; Xuanhui Huang


Team

Projects

Knowledge Acquisition and ManagementKnowledge Acquisition and Management

Web of Objects (WOO) offers unified access to structured, learned knowledge for Yahoo! and provides online indexing/retrieval, query processing, and serving.

Read more Read more
Multimedia SearchMultimedia Search

Multimedia search on Yahoo! helps people find what they are looking for through Web, image, or video search. One of the most interesting challenges is how to match queries to images beyond plain text so that users can more easily find the image they want.

Read more Read more
Traffic ShapingTraffic Shaping

Traffic shaping jointly optimizes CTR and post-click downstream utilities for recommending links. To achieve this goal, multiple objectives are modeled in a constrained optimization framework, aiming to taking revenue as an objective for content optimization.

Read more Read more
CAPTCHAsCAPTCHAs

CAPTCHAs ensure that Yahoo! properties are utilized by humans for the intended purposes to prevent email spam and phishing messages.

Read more Read more
User-Generated Content Abuse DetectionUser-Generated Content Abuse Detection

The Yahoo! Labs Audience Sciences team collaborates with the Yahoo! abuse engineering team to build a standard moderation platform that automatically detects abusive user-generated content (UGC) in the major media properties of Yahoo!

Read more Read more
Auction Anti-FraudAuction Anti-Fraud

The Yahoo! Labs Audience Sciences team builds machine learning models into the fraud detection system to prevent fraud in Yahoo! Auctions.

Read more Read more
User ReputationUser Reputation

Reputable users are valuable assets to Yahoo!. The Yahoo! Labs Audiences Sciences team examines user reputation in a comment rating environment, where users make comments about content and rate the comments of one another

Read more Read more
User UnderstandingUser Understanding

The Yahoo! Labs Audience Science team works on user understanding to collect and leverage data about Yahoo! users to provide more relevant content and search experiences.

Read more Read more
Mail Anti-Spam ProjectMail Anti-Spam Project

Yahoo! Labs Audience Sciences has partnered with the Yahoo! Mail team to cut down on the spam that reaches users' inboxes.

Read more Read more
Data Quality and IntegrityData Quality and Integrity

Working with the Abuse, Data Quality and Mail teams, the Yahoo! Labs Audience Sciences team is actively working to improve data quality at Yahoo!.

Read more Read more