Yahoo!

About: Yahoo! is a company organization based out in London, United Kingdom. It is known for research contribution in the topics: Population & Web search query. The organization has 26749 authors who have published 29915 publications receiving 732583 citations. The organization is also known as: Yahoo! Inc. & Maudwen-Yahoo! Inc.

Gallery application for content viewing

Abstract: A gallery software application enables a user to browse, view, and interact with various content items, such as still images and videos. The gallery includes a graphical user interface that displays multiple images in the foreground and one image in the background. The foreground images represent content items. The background image is generated based on one of the foreground images. As the foreground images are scrolled, the background image changes.

Systems and methods for matching participants to a conversation

Abstract: Systems and methods for connecting two or more individuals to an Internet conversation based on their mutual interests, the current content they may be viewing and what they want to talk about at that time. The techniques of the present invention allow an individual who starts a conversation to maintain full control over who is able to join that conversation as well as how many are able to join at any one time. A user who desires to start or join a conversation about a particular topic or story selects an indicator, such as an icon, associated with the specific topic or story. The user is presented with an option to start or join a conversation. If the user opts to start a conversation, the user is presented with a comment page, and the user enters a comment, or comments, that preferably is intended to spark an interest in other users. The comment is then presented to other users. Those users who may desire to join in a conversation with the conversation starter respond with their own comment, which is then sent to the conversation starter. The conversation starter reviews the comment, and if the comment is subjectively acceptable, the conversation starter brings the responding user into a conversation. The responding user and conversation starter are then connected in a messaging session, such as an instant messaging session. Multiple users may be connected in a single messaging session by the conversation starter in this manner.

Implementing Sponsored Search in Web Search Engines: Computational Evaluation of Alternative Mechanisms

Abstract: The practice of sponsored search advertising---where advertisers pay a fee to appear alongside particular Web search results---is now one of the largest and fastest growing source of revenue for Web search engines. We model and compare several mechanisms for allocating sponsored slots, including stylized versions of those used by Overture and Google, the two biggest brokers of sponsored search. The performance of these mechanisms depends on the degree of correlation between providers' willingness to pay and their relevance to the search term. Ranking providers based on the product of relevance and bid price performs well and is robust across varying degrees of correlation. Ranking purely based on bid price fares nearly as well when bids and relevance are positively correlated (the expected regime), and is further enhanced by adding an editorial filter. Regardless of the allocation mechanism, sponsored search revenues are lower when users' attention decays quickly at lower ranks, emphasizing the need to develop better user interfaces and control features. The search engine can address initial inscience of relevance scores by modifying rank allocations over time as it observes clickthroughs at each rank. We propose a rank-revision strategy that weights clicks on lower ranked items more than clicks on higher ranked items. This method is shown to converge to the optimal (maximum revenue) ordering faster and more consistently than other methods.

Hilbert Space Embeddings of Hidden Markov Models

Abstract: Hidden Markov Models (HMMs) are important tools for modeling sequence data. However, they are restricted to discrete latent states, and are largely restricted to Gaussian and discrete observations. And, learning algorithms for HMMs have predominantly relied on local search heuristics, with the exception of spectral methods such as those described below. We propose a nonparametric HMM that extends traditional HMMs to structured and non-Gaussian continuous distributions. Furthermore, we derive a local-minimum-free kernel spectral algorithm for learning these HMMs. We apply our method to robot vision data, slot car inertial sensor data and audio event classification data, and show that in these applications, embedded HMMs exceed the previous state-of-the-art performance.

Jointly Attentive Spatial-Temporal Pooling Networks for Video-Based Person Re-identification

Abstract: Person Re-Identification (person re-id) is a crucial task as its applications in visual surveillance and human-computer interaction. In this work, we present a novel joint Spatial and Temporal Attention Pooling Network (ASTPN) for video-based person re-identification, which enables the feature extractor to be aware of the current input video sequences, in a way that interdependency from the matching items can directly influence the computation of each other's representation. Specifically, the spatial pooling layer is able to select regions from each frame, while the attention temporal pooling performed can select informative frames over the sequence, both pooling guided by the information from distance matching. Experiments are conduced on the iLIDS-VID, PRID-2011 and MARS datasets and the results demonstrate that this approach outperforms existing state-of-art methods. We also analyze how the joint pooling in both dimensions can boost the person re-id performance more effectively than using either of them separately 1.