24 May, 2005 at 21:05 Leave a comment

Yahoo! Research Labs 
  Research Home – Next Home – Help

Research Projects

Yahoo! Research Labs seeks the inventive computer science solutions that will help make our products more and more essential to consumers in their daily lives. We want to make their search for information on the Internet easier and more efficient, find new and better ways for them to connect and communicate with family and friends, decrease the amount of junk mail in their email inbox. For advertisers and customers, we want to help them gain insights on their customers, find and engage their targets, maximize their marketing investments.

What’s Cooking?

Here, you will find a number of in-progress research projects. At times, they will be unavailable. At times, they will produce spurious results. And at times, they will amaze and delight you.

  • Tech Buzz Game
    The Tech Buzz Game is a fantasy prediction market for high-tech products, concepts, and trends. As a player, your goal is to predict how popular various technologies will be in the future. Popularity or buzz is measured by Yahoo! Search frequency over time. Predictions are made by buying virtual stock in the products or technologies you believe will succeed, and selling stock in the technologies you think will flop. In other words, you “put your play money where your mouth is.”
  • Concept Discovery
    What’s on the consumers’ minds today? This week? How does that compare to last week? More importantly, what might it tell us about next week? Grab your favorite printed publication and see how closely the stories align with what users are asking for online.
  • Cluster Graphing
    Anyone who has ever had to complete a what doesn’t belong question on a test has an interest in clustering technology. How close are the terms “slime mold”, “skunk odor removal” and “luxury bathroom” anyway? Zoom in here and find out. (Clue: They are all green).

Success Stories

The most rewarding aspect of our research is seeing theory manifest in application. Here are a few examples of solutions conceived and incubated by our labs that were eventually deployed into products that deliver an improved experience for both consumers and businesses.

Spelling Correction

  • The Problem: When consumers misspell a search query word, they get incorrect or inconclusive results and marketers miss the opportunity to present relevant advertising. Our challenge was to determine how to pick the most appropriate spelling correction for a mistyped query from a number of possible candidates.
  • The Solution: Based on techniques originally developed within the BioInformatics community for use in gene DNA comparisons, our researchers developed a patent-pending method for determining which spelling correction candidates were better than others in a given context. In this way, we are able to match the most appropriate advertiser keywords to misspelled user queries.
  • The Impact: The implementation of the Orthographic Match Driver project has enabled Overture’s advertisers to receive highly qualified, targeted leads in cases where the user has mistyped the intended query term. The net result is a win-win for both the consumer and the advertiser.

Sub-Phrase and Broad Matching

  • The Problem: Pay-for-performance search advertisers who bid on multiple-word search queries (e.g. fresh cut flowers delivery) want to maximize their ability to serve relevant ads to consumers entering related sub-phrases (e.g. fresh flowers). Our challenge was determining how to extend the coverage of advertiser listings while maintaining the relevance of the search results for both users and advertisers.
  • The Solution: Adapting and extending the technologies developed for the Spell Correction project, we created technologies to allow advertiser listings to match user queries where the bidded term is a sub-phrase of the user query. By identifying the optimal sub-phrase matches from a set of candidates, we ensure that the relevance of the results shown to the consumer is maintained.
  • The Impact: By providing both the core matching technologies as well as term selection tools enabling advertisers to identify which “root terms” they should bid on, we are enabling Overture advertisers to extend their reach and consumers to see highly targeted ads where no exact match exists.

Contextual Advertising

  • The Problem: If a consumer simply browses a content site without having conducted a search query, how can you present advertiser listings that are contextually relevant to the specific content being displayed?
  • The Solution: Through the application of traditional information retrieval techniques and state-of-the-art machine learning approaches, we are able to identify the key concepts of a web page and extract the most relevant keywords and phrases to match our advertiser listings to.
  • The Impact: We provided the Overture engineering team with the core algorithms and code for the development of the Overture Content Match product.

Entry filed under: Uncategorized.

Parikrama Foundation

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

Trackback this post  |  Subscribe to the comments via RSS Feed


May 2005
« Apr   Jun »


%d bloggers like this: