Microsoft Technology Licensing, LLC

International Business Machines Corporation

Linkedin Corporation

Chi-Yi Kuan

Yongzheng Zhang

Shen Huang

Yi Zheng

Rui Zhao

Weidong Zhang

Huan Hoang

Xiaosu Huang

Weiqin Ma

Hu Wang

Zhou Jin

Yan Liu

Yu Wang

Yue Li

Krishna Venkatraman

Rob Parkin

Paritosh Desai

Yurong Shi

Fedor N Dzegilenko

Haipeng Li

Ziliang Lin

Paul Ko

Bin Wu

Haiming Zhao

Vibhu Prakash Saxena

SangHyun Park

Hong Lu

France Savard

Quan Wang

Ho Joon Park

Burcu Baran

John Crowther

David Davtian

Wenqian Li

Sang Hyun Park

Shady Elasra

Sui Yan

Fanbin Bu

Liwen Zhang

Srihari R Duddukuru

Techniques for performing unequal sampling are provided. In one technique, multiple scores generated by a prediction model are identified, each score corresponding to a different entity of multiple entities. Multiple buckets are determined, each bucket corresponding to a different range of scores. Each entity is assigned to a bucket based on the score corresponding to the entity. A probability distribution function is generated based on the scores and a number of scores belonging to each bucket. For each entity, a probability of sampling the entity is determined based on the probability distribution function and a score corresponding to the entity. A subset of the entities are sampled based on the probability determined for each entity.

Unequal probability sampling based on a likelihood model score to evaluate prevalence of inappropriate entities

Techniques are provided for predicting talent flow to and/or from a geographical region. In one technique, multiple entity profiles are stored and analyzed to generate training data that is labeled indicating whether a corresponding entity has moved to or moved from a region. A machine-learned prediction model is generated or trained based on the training data. Using the machine-learned prediction model, a prediction is made whether, for each entity corresponding to another entity profile, that entity will move to or move from a particular geographic region. Based on multiple predictions, a number of entities that are predicted to move to or move from the particular geographic region is determined. Talent flow data that is based on the number of entities is presented on a computer display.

Machine learning techniques to predict geographic talent flow

A machine may be configured to generate an enhanced user interface for displaying social analytics based on viral mentions and threading. For example, the machine accesses a plurality of items of digital content. The machine extracts, for each of the plurality of items of digital content, a title that describes a particular item of digital content. The machine generates a group of items of digital content based on the extracted titles associated with the plurality of items of digital content. The machine identifies, from the group of items of digital content, an original item of digital content and one or more subsequent items of digital content. The machine determines a strength value associated with the original item of digital content. The machine generates and causes a display of an enhanced user interface that displays the title and the strength value associated with the original item of digital content.

Social analytics based on viral mentions and threading

Methods, systems, and computer programs are presented for selecting features for a machine-learning model configured to detect anomalies in the evolution of data over time. One method includes an operation for identifying one or more key fields and value fields from the fields in a relational database. The method also includes grouping data of the value fields based on values of the one or more key fields and calculating one or more statistical values for each group of data of the value fields. The method further includes operations for monitoring an evolution of the one or more statistical values over time, and for selecting, based on the evolution of the one or more statistical values over time, features to be used by a machine-learning model to detect anomalies in content of the relational database over time. The method also includes executing the machine-learning model to detect the anomalies.

Automatic feature generation for machine learning in data-anomaly detection

A system, a machine-readable storage medium storing instructions, and a computer-implemented method are described herein are directed to a Key Feature Engine receives a request for a desired number of key features to be identified from a data set in at least a portion of a database. The Key Feature Engine executes instances of multiple types of machine learning data models on the data set to calculate respective regression coefficients, in each machine learning data model instance, for each feature in a plurality of features defined by a plurality of data categories of a social network service. The Key Feature Engine identifies at least one key feature, of one or more instances of the multiple types of machine learning data models, based on a value of a corresponding regression coefficient.

Understanding business insights and deep-dive using artificial intelligence

The disclosed embodiments provide a system for performing A/B testing for search engine optimization (SEO). During operation, the system divides a set of web pages into a control group, an inbound treatment group, an outbound treatment group, and a full treatment group. Next, the system generates a first set of links from the outbound treatment group to the inbound treatment group and a second set of links within the full treatment group. The system then applies an A/B test to pairs of groups selected from the control group, the inbound treatment group, the outbound treatment group, and the full treatment group. Finally, the system outputs a result of the A/B test for use in assessing an effect of the first and second sets of links on search engine results associated with the set of web pages.

A/B testing for search engine optimization

Techniques for delaying the transmission of a message to one or more recipients using an alert in order to provide data protection and security with respect to data included in the message are disclosed herein. In some embodiments, a computer-implemented method comprises: receiving a request to transmit a message from a computing device of a user to a recipient, the request comprising content of the message; detecting an issue with the request using at least one classifier to classify the request as having the issue; generating an alert based on the detecting of the issue; and prior to transmitting the message to a destination associated with the recipient, causing the generated alert to be displayed on the computing device of the user, the alert indicating the issue with the message.

Data protection using alerts to delay transmission

A system and method for identifying trending topics in a document corpus are provided. First, multiple topics are identified, some of which topics may be filtered or removed based on co-occurrence. Then, for each remaining topic, a frequency of the topic in the document corpus is determined, one or more frequencies of the topic in one or more other document corpora are determined, a trending score of the topic is generated based on the determined frequencies. Lastly, the remaining topics are ranked based on the generated trending scores.

Scalable mining of trending insights from text

Among other things, embodiments of the present disclosure discussed herein help identify trending topics and generate customized data feeds that present trending topics to a user based on information in the user's social network profile. In some embodiments, users may select topics of interest to the user and the system can identity and present trending articles in the selected topic to the user via the user's newsfeed.

Customized data feeds for online social networks

The disclosed embodiments provide a system for processing data. During operation, the system obtains validated training data containing a first set of content items and a first set of classification tags for the first set of content items. Next, the system uses the validated training data to produce a statistical model for classifying content using a set of dimensions represented by the first set of classification tags. The system then uses the statistical model to generate a second set of classification tags for a second set of content items. Finally, the system outputs one or more groupings of the second set of content items by the second set of classification tags to improve understanding of content related to the set of dimensions without requiring a user to manually analyze the second set of content items.

Model-based classification of content items

Method and system for identifying user expectations in question answering in an on-line social network system are described. The automated support system is configured to address the technical problem of optimization of the processing of user input submitted to a computer in the form of a natural language. The automated support system uses machine learning algorithms to automatically extract, from the user input, information indicative of the user's expectations and obtain data relevant to the input based on said information indicative of the user's expectations.

Identifying user expectations in question answering

A topical representative assessment system implements techniques for determining entities that are ambassadors for one or more topics. The ambassadors are determined based on content items that they have authored or content items that are otherwise attributed to them. An ambassador may be any type of entity such as a person, a company, or an organization. Machine analytics may be used to determine whether a content item corresponds to a specific topic, determine a sentiment for a content item, analyze feedback for a content item, or any combination of these.

Identifying entity representatives for topics reflected in content items using natural language processing

A phony profiles detector for an on-line social network system is described. The phony profiles detector uses characteristics of a profile that was associated with an indication that it represents a malicious user, a so-called seed profile, to identify other profiles that should be flagged as potentially the source of undesirable behavior. Based on the degree of similarity determined for a subject profile with respect to a seed profile, the phony profiles detector generates a malicious user indicator and stores it as associated with the subject profile.

Phony profiles detector

The disclosed embodiments provide a system for processing data. During operation, the system obtains a first configuration for processing a first set of content items from a first data source and a second configuration for processing a second set of content items from a second data source. For each content item in the first set of content items, the system uses mappings from the first configuration to transform original fields from the content item into required fields in a record representing the content item. Next, the system generates, from the required fields, a document key for the content item. The system also performs deduplication of multiple records with the document key and stores a single record with the document key. Finally, the system uses the second configuration to generate, from the second set of content items, a set of records independently of processing the first set of content items.

Scalable processing of heterogeneous user-generated content

In cases where a user of a social network has lost access to one or more email addresses on file with the social network, the social network may confirm an identity of the user before issuing a new password. The social network may confirm the identity of the user by prompting the user to correctly identify pictures of members of the social network that are connected to the user, selecting or receiving a selection of a plurality of trustees (e.g., members of the social network who are connected to the user), prompting the trustees to confirm with the user that the user has lost the access, and receiving notification that at least a specified number of the trustees have confirmed with the user that the user has lost the access. After the social network confirms the identity, the social network may provide a new password to the user.

Growing community of inventors

Fremont, CA, United States of America

Chi-Yi Kuan