Microsoft Technology Licensing, LLC

Aman Grover

Dhruv Arya

Ganesh Venkataraman

Liang Zhang

Siyu You

Benjamin Hoan Le

Qing Duan

Kimberly Faith McManus

Manoj Rameshchandra Thakur

Yongwoo Noh

Krishnaram Kenthapadi

Kun Liu

Songtao Guo

Parul Jain

Saurabh Kataria

Xiaowen Zhang

Wen Pu

Nadia Fawaz

Christopher Matthew Degiere

Fedor Vladimirovich Borisyuk

Varun Mithal

Zhifeng Deng

Guoyin Wang

Girish Kathalagiri Somashekairah

Machine learning based method for multilabel learning with label relationships is provided. This methodology addresses the technical problem of alleviating computational complexity of training a machine learning model that generates multilabel output with constraints, especially in contexts characterized by a large volume of data, by providing a new formulation that encodes probabilistic relationships among the labels as a regularization parameter in the training objective of the underlying model. For example, the training process of the model may be configured to have two objectives. Namely, in addition to the objective of minimizing conventional multilabel loss, there is another training objective, which is to minimize penalty associated with the prediction generated by the model breaking probabilistic relationships among the labels.

Multilabel learning with label relationships

The disclosed embodiments provide a system for processing data. During operation, the system performs processing related to a first set of features for a first entity using a first series of embedding layers, wherein the processing includes applying each embedding layer in the first series of embedding layers to a concatenation of all outputs of one or more layers preceding the embedding layer. Next, the system obtains a first embedding as an output of a first final layer in the first series of embedding layers. The system then outputs the first embedding for use by a machine learning model.

Entity pyramid embeddings

A system and method for inferring service opportunities are provided. In example embodiments, a member event associated with a particular member of a social networking service is detected. In response to detecting the member event, a service request for a particular service is inferred based on the member event. A provider member capable of fulfilling the inferred service request is identified among members of the social networking service. A match score for each of the identified provider members is calculated. The identified provider members are ranked according to the calculated match score. At least a portion of the ranked identified provider members are presented on a user interface.

Inferring service opportunities

A member profile including a vector containing a field for each of a plurality of skills and a rating of one or more of the skills in the vector for a member of a social networking service is obtained. A first distance indicating a vector distance between the vector of the member profile and a vector of a hypothetical member profile representing the perfect job candidate is obtained. A hypothetical member profile for the member is created by combining the vector of the member profile with the indication of how each of the one or more skills is improved through taking the course from course information. A second distance between the member and the hypothetical perfect candidate for the job is obtained, and the difference between the first distance and the second distance is calculated to determine an estimate of how much the course will increase the member's job chances.

Estimating effects of courses

In an example, features in a boosting decision tree model are initialized to zero, the boosting decision tree model located in a GLMM and connected to a deep neural network collaborative filtering model via a prediction layer. While the features in the boosting decision tree model remain zero, the deep neural network collaborative filtering model is trained. One or more trees in the boosting decision tree model are boosted using logits produced by the training of the deep neural network collaborative filtering model as a margin. The prediction layer is trained using features from the deep neural network collaborative filtering model and features from the boosting decision tree model. It is then determined whether a set of convergence criteria is met. If not, then the deep neural network collaborative filtering model is retrained using the features and the process is repeated until the set of convergence criteria is met.

Deep and wide machine learned model for job recommendation

Various embodiments described herein provide for systems and methods for using a machine-learning model to rank job search results based on the similarity of the job title of each job search result and a job search query that produces the job search results. According to some embodiments, the machine-learning model comprises a word-embedding machine-learning model that maps a word to a vector.

Search query and job title proximity computation via word embedding

Instead of a fixed fee for a particular job application, discussed in some examples are methods, systems, and machine readable mediums which provide for a job posting service that utilizes a pay-per-click model. That is, job posters pay a fee for each time the member selects the job posting for further inquiry when that posting is shown to a member (called an impression). The fee that is paid is determined by the job poster. Selecting a job posting may comprise clicking on or otherwise entering an input signifying an intention to view the job.

Electronic job posting marketplace

Methods, systems, and computer programs are presented for expanding a job search that includes an industry by adding other similar industries. A method identifies job titles of members in a social network and performs, utilizing a machine-learning program, semantic analysis of the job titles to identify similarity coefficients among the job titles. The machine-learning program utilizes social network data to identify the similarity coefficients. Further, the method includes an operation for receiving a job search query, from a first member, including a query job title, and for expanding the job search query with job titles that are similar to the query job title. The method further includes operations for executing the expanded job search query to generate a plurality of job results, and for causing presentation on a display of one or more of the top job results.

Determining similarities among job titles to enhance job searching

A system and method for inferring service providers are provided. In example embodiments, member data of members of a social networking service is accessed. From the member data, it is inferred that a particular member among the members of the social networking service is a freelancer member. A service request that the freelancer member is capable of fulfilling is identified based on a service request skill associated with the service request and a freelancer skill of the freelancer member. An option for the freelancer member to fulfill the service request is presented on a user interface of a user device of the freelancer member.

Inferring service providers

Methods, systems, and computer programs are presented for expanding a job search that includes an industry by adding other similar industries. A method accesses, by a social networking server, a plurality of job applications, with each job application being submitted by a member for a job in a company, the member and the job having a respective industry from a plurality of industries. Semantic analysis of the job applications is performed by a machine-learning program to identify similarity coefficients among the plurality of industries. A job search query is received from a first member, the job search query including a query industry, and the job search query is expanded with industries that are similar to the query industry. The social networking server executes the expanded job search query to generate a plurality of job results. Presentation is provided on a display of one or more of the top job results.

Determining similarities among industries to enhance job searching

A trained search system can be configured to retrieve a candidate subset of results, where the trained search system uses data extracted from a machine learning scheme. The machine learning scheme can be trained to identify results that are ranked by a computationally expensive algorithm, such as a ranking algorithm. When a query is received, the trained search system can be used to retrieve results instead of applying the computationally expensive ranking algorithm.

Growing community of inventors

Sunnyvale, CA, United States of America

Aman Grover