Adobe, Inc.

Palo Alto Research Center Inc.

Microsoft Technology Licensing, LLC

Amazon Technologies, Inc.

Ebay, Incorporated

Tracy Holloway King

Ajinkya Gorakhnath Kale

Jessica Nicola Staddon

Philippe J P Golle

Ji Fang

Franck Dernoncourt

Ryan A Rossi

Baldo Antonio Faieta

Kurt Edward Partridge

Vishwa Vinay

Nicolas B Ducheneaut

Shabnam Ghadar

Richard Chow

Nedim Lipka

John Turner Maxwell, Iii

Saeid Motiian

Pranav Vineet Aggarwal

Ravi Chandra Jammalamadaka

Brian Scott Johnson

Richard Stamford Crouch

Oliver Brdiczka

David D Ahn

Alex Charles Filipkowski

Vamsi Salaka

Valerie Nygaard

Whitney Gale Spathelf

Philippe Jean-Paul Golle

Sourabh Goel

Erick Cantu-Paz

Abhishek Gupta

Ritiz Tambi

James Michael Reinhold

Piyush Chandra

Riccardo Turchetto

Nitin Kumar Garg

Joanna Mun Yee Chan

Andrea Ryerson Burbank

Timothy McDonnell Converse

Feroz Ahmad

Deepika Naryani

Vikas Sagar

Seyedsaed Rezayidemne

Naveen Marri

Christian Biemann

Feng Pan

Valeria De Paiva

John T Maxwell Iii

Systems and methods for image retrieval are described. Embodiments of the present disclosure receive a search query from a user; extract an entity and a color phrase describing the entity from the search query; generate an entity color embedding in a color embedding space from the color phrase using a multi-modal color encoder; identify an image in a database based on metadata for the image including an object label corresponding to the extracted entity and an object color embedding in the color embedding space corresponding to the object label; and provide image information for the image to the user based on the metadata.

Embedding-based color-object retrieval

Systems and methods are disclosed for search query language identification. One method comprises generating a seed dictionary comprising a plurality of labeled dictionary terms and receiving a plurality of unlabeled sample query terms. The plurality of unlabeled sample query terms are compared to the plurality of labeled dictionary terms at a first time, and a first set of labeled sample query terms are generated by labeling at least a subset of the plurality of unlabeled sample query terms based on the first comparison. Remaining unlabeled sample query terms are then compared with the first set of labeled sample query terms at a second time, and a second set of labeled sample query terms are generated by labeling the remaining unlabeled sample query terms based on the second comparison. The first and second sets of labeled sample query terms are provided to a machine learning model configured for query language prediction.

Methods and systems for search query language identification

The present disclosure describes systems and methods for information retrieval. Embodiments of the disclosure provide a retrieval network that leverages external knowledge to provide reformulated search query suggestions, enabling more efficient network searching and information retrieval. For example, a search query from a user (e.g., a query mention of a knowledge graph entity that is included in a search query from a user) may be added to a knowledge graph as a surrogate entity via entity linking. Embedding techniques are then invoked on the updated knowledge graph (e.g., the knowledge graph that includes additional edges between surrogate entities and other entities of the original knowledge graph), and entities neighboring the surrogate entity are retrieved based on the embedding (e.g., based on a computed distance between the surrogate entity and candidate entities in the embedding space). Search results can then be ranked and displayed based on relevance to the neighboring entity.

Knowledge-derived search suggestion

The present disclosure relates to systems for identifying instances of natural language input, determining intent classifications associated with instances of natural language input, and generating responses based on the determined intent classifications. In particular, the disclosed systems intelligently identify and group instances of natural language input based on characteristics of the user input. Additionally, the disclosed systems determine intent classifications for the instances of natural language input based message queuing in order to delay responses to the user input in ways that increase accuracy of the responses, while retaining a conversational aspect of the ongoing chat. Moreover, in one or more embodiments, the disclosed systems generate responses utilizing natural language.

Accuracy of natural language input classification utilizing response delay

Disclosed are various embodiments for a navigation element. An insertion point is identified in a ranking of search results. A navigation element is included in the search results. A selection of the navigation element modifies a constraint used to generate the search result.

Search navigation element

A system and method of query expansion are disclosed. A query expansion source, a query expansion candidate, and feature data for the query expansion source and the query expansion candidate are received. The feature data comprises information for a plurality of features. A determination is made as to whether the query expansion candidate qualifies as an expansion of the query expansion source based on an analysis of the information for the plurality of features. The query expansion candidate is assigned as an expanded query of the query expansion source in a query expansion dictionary in response to a determination that the query expansion candidate qualifies as an expansion of the query expansion source.

Query expansion classifier for E-commerce

Systems, methods, and computer storage media having computer-executable instructions embodied thereon that facilitate generation of snippets. In embodiments, text features within a keyword-sentence window are identified. The text features are utilized to determine break features that indicate favorability of breaking at a particular location of the keyword-sentence window. The break features are used to recognize features of partial snippets such that a snippet score to indicate the strength of the partial snippet can be calculated. Snippet scores associated with partial snippets are compared to select an optimal snippet, that is, the snippet having the highest snippet score.

Generating snippets based on content features

Some embodiments of the present invention provide a system that infers activity-related context information from a message. Upon receiving the message, the system looks for activity-related keywords in the message, wherein the activity-related keywords are from a content database. If one or more activity-related keywords are found in the message, the system infers message-related context information from the one or more activity-related keywords. Next, the system uses the message-related context information to facilitate recommending an activity to a user.

Using a content database to infer context information for activities from messages

One embodiment provides a system that detects sensitive passages. During operation, the system receives a document and disassembles the document into a plurality of passages. For a respective passage, the system performs a search through a non-sensitive-passage database to determine whether the passage is a known non-sensitive passage. If so, the system marks the passage as non-sensitive, and if not, the system determines whether the passage triggers a cut-and-paste attack detection. If so, the system forwards the passage to an administrator and allows the administrator to determine whether the passage is non-sensitive and, further, to add the passage to the non-sensitive-passage database responsive to the administrator determining the passage to be non-sensitive.

Non-sensitive-passage database for cut-and-paste attack detection systems

One embodiment of the present invention provides a system that detects sensitive content in a document. In doing so, the system receives a document, identifies a set of terms in the document that are candidate sensitive terms, and generates a combination of terms based on the identified terms that is associated with a semantic meaning. Next, the system performs searches through a corpus based on the combination of terms and determines hit counts returned for each term in the combination and for the combination. The system then determines whether the combination of terms is sensitive based on the hit count for the combination and the hit counts for the individual terms in the combination, and generates a result that indicates portions of the document which contain sensitive combinations.

Method and apparatus for detecting sensitive content in a document

Techniques are provided for detecting entailment and contradiction. Packed knowledge representations for a premise and conclusion text are determined comprising facts about the relationships between concept and/or context denoting terms. Concept and context alignments are performed based on alignments scores. A union is determined. Terms are marked as to their origin and conclusion text terms replaced with by corresponding terms from the premise text. Subsumption and specificity, instantiability, spatio-temporal and relationship based packed rewrite rules are applied in conjunction with the context denoting facts to remove entailed terms and to mark contradictory facts within the union. Entailment is indicated by a lack of any facts from the packed knowledge representation of the conclusion in the union. Entailment and contradiction markers are then displayed.

Systems and methods for detecting entailment and contradiction

Growing community of inventors

Mountain View, CA, United States of America

Tracy Holloway King