Yahoo! Inc.

Microsoft Technology Licensing, LLC

Eren Manavoglu

Kannan Achan

Erick Cantu-Paz

Zhichen Xu

Liang-Yu Tom Chi

Byron Edward Dom

Keng-hao Chang

Nan Duan

Jian Jiao

Rukmini Iyer

Yun Fu

Alexandrin Popescul

Yeyun Gong

Ozgur Cetin

Xingwei He

Pai Chun Lin

Mohammadreza Khalilishoja

Divy Kothiwal

Wen Cui

Cliff Brunk

Dmitry Yurievich Pavlov

Paul Heymann

A technique iteratively updates model weights of a teacher model and a student model. In operation, the teacher model produces noisy original pseudo-labeled training examples from unlabeled training examples. The technique weights the original pseudo-labeled training examples based on validation information. The technique then updates model weights of the student model based on the weighted pseudo-labeled training examples. The validation information, which is used to weight the original pseudo-labeled training examples, is produced by selecting labeled training examples based on an uncertainty-based factor and a similarity-based factor. The uncertainty-based factor describes an extent to which the student model produces uncertain classification results for the set of labeled training examples. The similarity-based factor describes the similarity between the set of labeled training examples and the unlabeled training examples. Overall, the technique is efficient because it eliminates the need to produce a large number labeled training examples.

Resource-efficient training of a sequence-tagging model

Technologies described herein relate to dense retrieval and ranking of search results. A query indicating a computing context or user input is received. An embedding of the query is computed by way of a first encoder, and candidate results selected from a pool of potential results based upon the embedding of the query and embeddings of the potential results. A similarity score for a first of the candidate results is computed by way of a second encoder trained based upon an order metric that defines a ranking over a training set of potential results. The first encoder is trained based upon output of the second encoder prior to computing the embedding of the query. The candidate results are ranked based upon the similarity score of the first candidate result, and results responsive to the query are identified based upon the ranking. The identified results are output to a computing device.

Dense retrieval employing progressive distillation training

A method for combining multiple probability of click models in an online advertising system into a combined predictive model, the method commencing by receiving a feature set slice (e.g. corresponding to demographics or taxonomies or clusters), and using the sliced data for training multiple slice-wise predictive models. The trained slice-wise predictive models are combined by overlaying a weighted distribution model over the trained slice-wise predictive models. The combined predictive model then is used in predicting the probability of a click given a query-advertisement pair in online advertising. The method can flexibly receive slice specifications, and can overlay any one or more of a variety of distribution models, such as a linear combination or a log-linear combination. Using an appropriate weighted distribution model, the combined predictive model reliably yields predictive estimates of occurrence of click events that are at least as good as the best predictive model in the slice-wise predictive model set.

Using linear and log-linear model combinations for estimating probabilities of events

A computer-implemented method and system for selecting a subject advertisement in a sponsored search system based on a user's commercial intent (pertaining to the subject advertisement), using techniques for determining intent-driven clicks from a historical database. The method includes steps for aggregating a training model dataset wherein the training model dataset contains a selected history of clicks. Then, selecting from the training model dataset, a clicked slate (further selection of clicks), the clicked slate comprising a set of clicked ads, and calculating an intent-driven click feedback value for the subject advertisement. The method includes techniques for selecting a clicked slate using features corresponding to clicks received within a particular time period (the time period determined statically or dynamically). A system for implementing the method includes aggregating data from a historical database using selectors such as a position selector, a click feature selector, an impression-advertiser-campaign-creative selector, and a commercial intent selector.

Using clicked slate driven click-through rate estimates in sponsored search

A system is described for targeting data to a site referenced on a page based on a condition. The system may include a processor, a memory, and an interface. The memory may be operatively connected to the processor and the interface and may store a data, a site, a condition, and a page containing content. The interface may be operatively connected to the memory and the processor and may communicate the page to a user. The processor may identify the data, site, condition, and page containing content. The processor may add the data to the page if the content of the page satisfies the condition.

System for targeting data to sites referenced on a page

The present invention is directed towards systems and method for organization of bookmarks. The method according to one embodiment comprises retrieving one or more bookmarks associated with one or more content items, a given bookmark generated by a user of a client device and identifying one or more tags associated with one or uniform resource locators corresponding to the or more bookmarks. A bookmark folder hierarchy is created through use of a clustering algorithm on the basis of the one or more tags associated with the one or more uniform resource locators corresponding to the one or more bookmarks.

System and method for automatically organizing bookmarks through the use of tag data

Growing community of inventors

Menlo Park, CA, United States of America

Eren Manavoglu