Palo Alto Research Center Inc.

Avaya Technology LLC

Google Inc.

Xerox Corporation

Lichan Hong

Stuart Kent Card

Ed H Chi

Kenneth Charles Cox

Paul Lawrence Richman

Vladimir Nepustil

Ed Huai-Hsin Chi

Mark Jeffrey Stefik

Jock Douglas Mackinlay

Jilin Chen

Huazhong Ning

Jindong Chen

Bongwon Suh

Gregorio Convertino

Tianjiao Zhang

Peng Dai

Richard Carl Gossweiler, Iii

Peter L T Pirolli

Robert C Steiner

David L Hecht

Allison Gyle Woodruff

Aakanksha Chowdhery

Barbara Stefik

Sagar Jain

Hussein Hazimeh

Maheswaran Sathiamoorthy

Peter Lai

Jose F Velez

Ruth E Rosenholtz

Amy Xian Zhang

Zhe Zhao

Yihua Chen

Michael Bernstein

Peter L Pirolli

A system including a main neural network for performing one or more machine learning tasks on a network input to generate one or more network outputs. The main neural network includes a Mixture of Experts (MoE) subnetwork that includes a plurality of expert neural networks and a gating subsystem. The gating subsystem is configured to: apply a softmax function to a set of gating parameters having learned values to generate a respective softmax score for each of one or more of the plurality of expert neural networks; determine a respective weight for each of the one or more of the plurality of expert neural networks; select a proper subset of the plurality of expert neural networks; and combine the respective expert outputs generated by the one or more expert neural networks in the proper subset to generate one or more MoE outputs.

Sparse and differentiable mixture of experts neural networks

Implementations provide an improved system for presenting search results based on entity associations of the search items. An example method includes generating first-level clusters of items responsive to a query, each cluster representing an entity in a knowledge base and including items mapped to the entity, merging the first-level clusters based on entity ontology relationships, applying hierarchical clustering to the merged clusters, producing final clusters, and initiating display of the items according to the final clusters. Another example method includes generating first-level clusters from items responsive to a query, each cluster representing an entity in a knowledge base and including items mapped to the entity, producing final clusters by merging the first-level clusters based on an entity ontology and an embedding space that is generated from an embedding model that uses the mapping, and initiating display of the items responsive to the query according to the final clusters.

Clustering search results

Techniques are described that provide a reading pane for reading one or more content portions of a work and a guide pane that displays a structured overview of additional information from the corpus that is focused by an analysis of information from the currently-visible content portion of the work. The contents of the currently-visible content portion displayed within the reading page are used to derive dynamic representations of the user's immediate interests. At the user's request, a structured persistent context-aware guide is computed and displayed. The persistent context-aware guide provides a framework for accessing other parts of the corpus having the most-related information. The elements in the persistent context-aware guide are other content portions or other organizations such as sections, chapters, or articles in the corpus. The persistent context-aware guide may also incorporate additional closely-related information elements from outside the corpus including advertisements. Advertisements for information on the topics of immediate interest to the user can be incorporated within the structured, context-aware guide so that the reader can purchase them and review that information in the context of the current session. The guide pane and the reading pane support navigation controls so that the user can go to new or previous sub-portions of the content portion, new or previous content portions and/or new or previous persistent context-aware guides.

Systems and methods for persistent context-aware guides

Implementations provide an improved system for presenting search results based on entity associations of the search items. An example method includes, for each of a plurality of crowdsource workers, initiating display of a first randomly selected cluster set from a plurality of cluster sets to the crowdsource worker. Each cluster set represents a different clustering algorithm applied to a set of search items responsive to a query. The method also includes receiving cluster ratings for the first cluster set from the crowdsource worker and calculating a cluster set score for the first cluster set based on the cluster ratings. This is repeated for remaining cluster sets in the plurality of cluster sets. The method also includes storing a cluster set definition for a highest scoring cluster set, associating the cluster set definition with the query, and using the definition to display search items responsive to the query.

Crowdsourced evaluation and refinement of search clusters

A system and method for extracting and reusing metadata to analyze messages is provided. A stream of messages is monitored. Those messages with a predetermined message component pointing to a referent are identified. Words that are related to the referent are extracted from each of the messages. A local similarity of the identified messages is determined by comparing the extracted words of each message. A global similarity of the identified messages is determined by combining the extracted words from all the identified messages and by comparing the combined extracted words with extracted words from all messages that include a different referent. A determination is made as to whether one or more of the extracted words from the identified messages are descriptive of the referent based on the local and global comparisons.

System and method for extracting and reusing metadata to analyze message content

A system and method for triaging of information feeds is provided. A plurality of information feeds are received. At least one topic is identified from each information feed. At least one topic is presented to a user in topic facet including a plurality of identified topics. A selection of one of the plurality of topics is received from the user. The user interface is updated to display only the feeds that contain the selected topic.

System and method for triaging of information feeds

A system and method for identifying topics for short text communications is provided. Tokens are extracted from a short text communication. A query is generated using the extracted tokens. The query is applied to a set of documents. Those documents in the set that match the query are identified as search results. Salient terms associated with each of the search results are identified. A threshold is applied to the identified salient terms. The salient terms that satisfy the threshold are selected as topics for the short text communication.

System and method for identifying topics for short text communications

Methods, apparatus and articles of manufacture therefor, are disclosed for annotating documents in a viewer. In one embodiment, the method includes: associating the document with a user; identifying one or more sub-document elements in the document; augmenting the document by making words within the document selectable; displaying the augmented document in the viewer; and responsive to each user input selecting a word in the document displayed in the viewer, further augmenting the document displayed in the viewer by inserting the selected word following the sub-document element from which the word was selected.

System and method for annotating documents using a viewer

Methods, apparatus and articles of manufacture therefor, are disclosed sharing and searching annotated document collections. In one embodiment, the method may be performed by: transforming a plurality of document identifiers and their associated set of tags into a set of URL/TAG pairs; constructing a bigraph using the set of URL/TAG pairs; using the bigraph to compute tag profiles and document profiles using spreading activation for document identifiers and tags in the set of URL/TAG pairs; for a selected tag or document, identifying a tag profile or a document profile associated with the selected tag or document, respectively; arranging by relevance those tags or documents in the identified tag profile or document profile associated with the selected tag or document, respectively.

System and method for searching annotated document collections

A system and method for providing a topic-directed search is provided, which advantageously harnesses user-provided topical indexes and an ability to characterize indexes according to how articles fall under their topical organizations. A corpus of articles and an index that includes topics from the articles is maintained. For each topic, a coarse-grained topic model is built, which includes the characteristic words included in the articles relating to the topic and scores assigned to the characteristic words. A search query is executed against the index. The topics that match the search terms are chosen by their scores. The topics that match the coarse-grained topic models and the articles corresponding to the search query are presented. In contrast to conventional search engines, search results are organized according to topic and search results can be offered across multiple indexes, where part of returned results are selected from most-relevant indexes with their most-relevant topics.

System and method for providing a topic-directed search

Page turning of a three-dimensional electronic book is modeled as a dynamic animation. The page to be turned may be rotated by rotating it about a base line, and deformed by curling points of the front face of the turning page around the outer side of an imaginary geometric shape.

Systems and methods for turning pages in a three-dimensional electronic document

Techniques provide for the dynamic display of a page-oriented information asset using an audio output mode. Pluralities of elements in the page-oriented information asset are determined based on chapters, sections, paragraphs, sentences, words and the like. The elements are rendered to an audio output mode. Optionally selectable representations of the elements are also determined and output using a 3-dimension-like rendering to a visual output mode. The 3-dimension-like rendering of the visual representations of the elements shows how the current element relates to the other elements and its location within the page-oriented information asset. The 3-dimension-like rendering provides visual orientation or context within the page-oriented information asset. The visual representations of the elements are optionally selectable as spatial context indicators providing direct navigation control to support reading, browsing and information retrieval tasks within the page-oriented information asset. Speech input is also optionally recognized as a direct navigation control.

Systems and methods for navigating page-oriented information assets

User's interests can be explicitly determined via keywords that the user specifies, and/or are implicitly constructed from user browsing and reading activity. User interests may be expressed as an interest profile. Conceptual keywords related to user interests are selected by combining spreading activation and word co-occurrence, by latent semantic analysis, or other methods. One embodiment automatically highlights sentences and other information that contain conceptual keywords related to user interests. Highlights can be activated when the user directly performs a keyword search or index search, or the invention can generate information reflecting user interests, apply it to the text, and generate and display highlights. One embodiment includes an algorithm for computing a conceptual keyword vector through an iterative spreading activation process also employing word co-occurrence. A conceptual index of the text may be created and then combined with conceptual highlighting.

Method for automatically performing conceptual highlighting in electronic text

A system and method for turning pages of a three-dimensional electronic book. A closed three-dimensional electronic book is modeled as at least one three-dimensional object having length, width and thickness dimensions. An opened three-dimensional book electronic is modeled as at least two three-dimensional objects having length, width and thickness dimensions, representing the left side of the opened book and the right side of the opened book. The page(s) to be turned is modeled as at least a three-dimensional object having length, width and thickness dimensions. The cross-section parallelogram having a width dimension and a thickness dimension based on the width and thickness dimensions of the closed three-dimensional electronic book, and a user-selected number of page(s) to be turned. Page turning of a three-dimensional electronic book is modeled as a dynamic animation showing the concurrent movement of three three-dimensional objects representing the left side pages block, the right side pages block, and the pages(s) to be turned.

Growing community of inventors

Mountain View, CA, United States of America

Lichan Hong