Microsoft Technology Licensing, LLC

Vita G Markman

Craig Martell

Yongzheng Zhang

Lutz Thomas Finger

Anmol Bhasin

Jeffrey William Pasternack

Baoshi Yan

Romer E Rosales-Delmoral

Ye Yuan

Mohammad Shafkat Amin

Varun Mithal

Igor Vladimir Yagolnitser

Methods, systems, and computer programs are presented for automatically generating phrase-based talent-screening questions. One method includes analyzing job descriptions to generate ngrams. Each ngram comprises one or more words. Further, the method includes identifying, from the ngrams, an ngram set comprising a predetermined number of bigrams and trigrams according to frequency of appearance in the job descriptions. The method further includes removing, from the ngram set, bigrams and trigrams comprising one or more of stop words, negation words, or requirement words, to obtain first seed phrases. The first seed phrases are filtered based on a frequency of appearance of the seed phrase in the job descriptions to obtain second seed phrases. Further, the second seed phrases are added to the first seed phrases to obtain third seed phrases. Each seed phrase is a sequence of one or more words that is associated with a category of talent-screening questions.

Discovery, extraction, and recommendation of talent-screening questions

Methods, systems, and computer programs are presented for estimating the language used in a user communication. One method includes an operation for utilizing counters to track use of languages by a user of an online service, the counters being updated based on interactions of the user in the online service. Further, the method includes operations for detecting a text entered by the user and obtaining, by a language classifier, an initial prediction having probabilities for the languages that the text is in the language. A language distribution prediction is calculated based on the initial prediction and the user counters, where the language distribution prediction comprises a probability, for each language, that the text is in the language. Further, the method includes operations for selecting a language used in the text based on the language distribution prediction and causing presentation on a display of a message in the selected language.

Language profiling service

Techniques for using online user activity in determining relevance of attributes to improve computer functionality in generating recommendations of online content are disclosed herein. In some embodiments, a computer system calculates a corresponding relevance score for each attribute of a user based on a total number of online postings for which the user has performed at least one of a plurality of online actions within a particular sliding window of time defining a most recent time period, an attribute activity number representing a number of online postings in the plurality of online postings that have the attribute, and an inverse of a frequency value representing how many of a total number of online postings published within the particular sliding window of time have the attribute. In some embodiments, the computer system causes at least one recommendation associated with the user to be displayed based on the calculated relevance scores.

System user attribute relevance based on activity

The disclosed embodiments provide a system for processing data. During operation, the system obtains a set of topics associated with a set of content items. Next, the system obtains a first set of attributes associated with a first topic in the set of topics and a second set of attributes associated with a second topic in the set of topics. Next, the system calculates a similarity between the first and second sets of attributes and applies a threshold to the similarity to identify the first and second topics as synonymous when the similarity exceeds a threshold. The system then merges the first and second topics under a representative topic. Finally, the system displays the representative topic to a user to improve understanding of the set of content items by the user without requiring the user to manually analyze the set of content items.

Detecting and combining synonymous topics

The disclosed embodiments provide a system for processing data. During operation, the system obtains a set of clauses in a first set of content items comprising unstructured data. Next, the system obtains a set of stop words comprising high-frequency words that occur in a second set of content items. The system then automatically extracts a set of topics from the set of clauses by generating a set of n-grams from the set of clauses and excluding a first n-gram in the set of n-grams from the set of topics when the first n-gram contains a word in the set of stop words in a pre-specified position of the first n-gram. Finally, the system displays the set of topics to a user to improve understanding of the first set of content items by the user without requiring the user to manually analyze the first set of content items.

Topic extraction using clause segmentation and high-frequency words

This disclosure provides systems and methods for determining a classification model for a secondary language different from a primary language. A social networking server is configured to obtain primary language content written in a first spoken language and secondary language content written in a second spoken language. The social networking server further obtains a machine translation of the primary language content. The social networking server then determines an initial language model from the machine translation. The social networking further determines a language model perturbation using the initial language model, where the language model perturbation accounts for a difference between the machine translation and the secondary language content. The social networking server also determines a classification model from the initial language model and the language model perturbation, which is then applied to a plurality of comments associated with an item of interest provided by a social networking service.

Transfer learning for bilingual content classification

The disclosed embodiments provide a system for processing data. During operation, the system obtains a set of content items and a set of topics in the set of content items. For each topic in the set of topics, the system automatically extracts a set of attributes that provides a context for the topic from a subset of the content items containing the topic. The system then displays the set of attributes in the context of the topic to improve understanding of the set of content items by the user without requiring the user to manually analyze the set of content items.

Growing community of inventors

San Francisco, CA, United States of America

Vita G Markman