Microsoft Technology Licensing, LLC

Tidyware, LLC

Raman Chandrasekar

Dean A Slawson

Eric Benjamin Watson

Michael K Forney

Robert L Rounthwaite

Eric David Brill

Matthew Robert Scott

James Charles Finger, Ii

Michael H Tsang

David W Steinkraus

Ray Sun

Simon H Corston

Sally K Salas

Tian Bai

Harr Chen

James C Finger, Ii

Abhishek Gupta

Jing Zhao

Christopher Brian Quirk

Arungunram Chandrasekaran Surendran

Erin L Renshaw

Matthew R Richardson

David Michael Sauntry

Eric I-Chao Chang

Sarthak D Shah

Eric Chang

Daniel J Makoski

Zhanliang Chen

Peter Fintan Leonard

Thomas Dale White

Piali Choudhury

Christopher John Champness Burges

Leon Chih Wen Wong

Brian Hayward

Ana Sultana Bacioiu

James Scott Boyle

Alvin G Solidum

Travis T Okahara

Parikshit Sondhi

William L Richman

Patrick Dean Kennedy

Hao Ma

Philip F Carmichael

Dharmarajan Ratnagiri Ayakkad

Systems and methods for training a personalized Machine Learning (ML) model used to detect fall events are described herein. The methods may be implemented by one or more computing devices and may include obtaining sensor data associated with one or more activities of a user. A processed or unprocessed version of at least a copy of the sensor data having been fed to a personalized ML model associated with the user and that has been determined not to be associated with a fall event; and using the obtained sensor data training the personalized ML model.

Systems and methods for training a personalized machine learning model for fall detection

Topical search engines can add contextual keywords to an input query to bias results toward a particular topic or domain. In one instance, query context models can be constructed to facilitate topical search. Upon receipt of one or more topic-relevant sites, a plurality of topical queries can identified automatically. Contextual keywords can be identified with respect to the plurality of topical queries as a function of lexical generality, among other things. Subsequently, a query context model, comprising the identified topical queries and related contextual keywords, can be employed to restrict query results to a particular topic.

Topical search engines and query context models

A software and/or hardware facility for inferring user context and delivering advertisements, such as coupons, using natural language and/or sentiment analysis is disclosed. The facility may infer context information based on a user's emotional state, attitude, needs, or intent from the user's interaction with or through a mobile device. The facility may then determine whether it is appropriate to deliver an advertisement to the user and select an advertisement for delivery. The facility may also determine an appropriate expiration time and/or discount amount for the advertisement.

Context based online advertising

Human computation games are provided wherein a player is shown a page, such as a web page. The player is then asked to provide one or more terms that are intended to cause a search engine to return the page in response to performing a query using the terms. The terms provided by the player during game play are then collected, stored, and utilized to improve the performance of the search engine.

Using a human computation game to improve search engine performance

A user interface display for software has a user satisfaction portion displayed on each page. The user satisfaction portion includes a user selectable element which allows a user to provide a user satisfaction score (or level) with a single mouse click. In response to receiving the user satisfaction level, the context of the software which the user is using is recorded, and the user satisfaction level is correlated to that context. The captured data can be provided to application designers and developers directly or via computed metrics.

User feedback correlated to specific user interface or application features

Domain specific topics, and optionally uniform resource locators (URLs) can be received from a user, and from those domain specific topics and URLs, domain tuned search definitions are generated for a given domain. The domain tuned search definitions are saved and the user is provided with a definition of a domain tuned search interface that is embedded on a site specified by the user. When someone reviewing the user's web site performs a search using the domain tuned, embedded search interface, a search engine is invoked which performs a search on the user's input query, and then returns domain specific search results. The search engine searches for domain specific search results over web sites in addition to the web site that the user is currently reviewing, so the search is more precise than a general web search but broader than a specific site search.

Creating business value by embedding domain tuned search on web-sites

The claimed subject matter provides systems and/or techniques that effectuate combining domain-tuned search systems. The system can include mechanisms that obtain queries, written descriptions, or illustrative web-pages regarding a particular area of interest, and generate a definition related to the area of interest. The definition contains a list of paths with associated weights employed to identify an pre-established first domain-tuned search system related to the area of interest. The first domain-tuned search system thereafter can be combined with a second domain-tuned search system related to another area of interest and presented to a user for utilization in re-ranking generic search results to be specific to the first and second domains of interest if combined, or to only the first domain if weights for the second domain are logically subtracted.

Combining domain-tuned search systems

An information dissemination system ranks the search results based on a temporal weight assigned to each search result. The temporal weight is an indication of the importance to a user that varies with time. For each search result, the information dissemination system calculates a temporal weight that is based on the temporal proximity of the event that is related to the search result. The temporal weight may be used to re-rank the search results.

Temporal ranking of search results

A search and a browse on a single user query is performed. A refined query is selected from the results of the first user query. Thereafter, a list of concepts from a first directory associated with the refined query is obtained. The concepts are defined in a hierarchical relationship with concepts having broader scope being higher in the hierarchy and concepts having a narrower scope being lower in the hierarchy. Additionally, a list of web sites associated with the search concept is obtained from a second directory.

System and method for performing a search and a browse on a query

An information dissemination system provides users information related to topics extracted from event classes. Event classes can generally be thought of as scripts that specify topics and actions for obtaining information related to the topics. The event classes are executed to obtain information related to the particular event class. The information dissemination system identifies a user event that is of interest to a user. The information dissemination system then identifies an event class that corresponds to the user event. The information dissemination system then performs the actions specified for the event class, obtains the action results, and renders the action results to the user.

Presenting information related to topics extracted from event classes

An information dissemination system identifies the interests of a user from information derived from various sources, such as, by way of example, the user's calendar entries, emails, World Wide Web (web) usage information, documents, enterprise resource planning (ERP) data, and the like. The information dissemination system then creates a profile for the user, and includes in the profile the details about the user's interests including the information sources to query, the user events which were analyzed to determine the user's interest, the topics derived from the analyzed events, the queries generated for the analyzed events, a language model of terms that are relevant to the user, and other information regarding the user. The information dissemination system then periodically executes the queries against the appropriate information sources, ranks the search results obtained from executing the queries, and renders the ranked search results to the user.

Adaptive dissemination of personalized and contextually relevant information

Text classification has become an important aspect of information technology. Present text classification techniques range from simple text matching to more complex clustering methods. Clustering describes a process of discovering structure in a collection of characters. The invention automatically analyzes a text string and either updates an existing cluster or creates a new cluster. To that end, the invention may use a character n-gram matching process in addition to other heuristic-based clustering techniques. In the character n-gram matching process, each text string is first normalized using several heuristics. It is then divided into a set of overlapping character n-grams, where n is the number of adjacent characters. If the commonality between the text string and the existing cluster members satisfies a pre-defined threshold, the text string is added to the cluster. If, on the other hand, the commonality does not satisfy the pre-defined threshold, a new cluster may be created. Each cluster may have a selected topic name. The topic name allows whole clusters to be compared in a similar way to the individual clusters, and merged when a predetermined level of commonality exists between the subject clusters. The topic name also may be used as a suggested alternative to the text string. In this instance, the topic name of the cluster to which the text string was added may be outputted as an alternative to the text string.

Method and system for performing phrase/word clustering and cluster merging

Refining a user query is disclosed. In one method, a query is received from a user, and then mapped to one or more search concepts. A list of search concepts associated associated with the query is then displayed. Alternatively or additionally, the search concepts associated with the query are used to provide a set of improved search results. In another method, a number of queries from a number of users are analyzed to identify two or more search concepts, and a popularity value is assigned to them based on the queries. Thus, the relative popularity of the respective search concepts can be determined. Alternatively or additionally, a preferred search query for the search concepts can be determined. The popularity and preferred queries can be used to allow automatic or user-initiated refinement.

System and method for query refinement to enable improved searching based on identifying and utilizing popular concepts related to users' queries

A human assisted method of debugging training data used to train a machine learning classifier is provided. The method includes obtaining a classifier training data set. The training data set is then debugged using an integrated debugging tool configured to implement a debugging loop to obtain a debugged data set. The debugging tool can be configured to perform an estimation and simplification step to reduce data noise in the training data set prior to further analysis. The debugging tool also runs a panel of prediction-centric diagnostic metrics on the training data set, and provides the user prediction based listings of the results of the panel of prediction-centric diagnostic metrics.

Method and apparatus for analysis and decomposition of classifier data anomalies

Growing community of inventors

Seattle, WA, United States of America

Raman Chandrasekar