Microsoft Technology Licensing, LLC

Saurabh Kataria

Dhruv Arya

Ada Cheuk Ying Yu

Swanand Wakankar

Ganesh Venkataraman

Lin Guo

Ananth Sankar

Aman Grover

Vipin Gupta

Benjamin Hoan Le

Nadia Fawaz

Ying Xuan

Bharat Kumar Jain

Sakshi Verma

Xue Xia

Nikita Gupta

Xun Luan

Guoyin Wang

Aman Gupta

Bhargavkumar Kanubhai Patel

Sirjan Kafle

Di Wen

Described herein are systems and methods for generating an embedding—a learned representation—for an image. The embedding for the image is derived to capture visual aspects, as well as textual aspects, of the image. An encoder-decoder is trained to generate the visual representation of the image. An optical character recognition (OCR) algorithm is used to identify text/words in the image. From these words, an embedding is derived by performing an average pooling operation on pre-trained embeddings that map to the identified words. Finally, the embedding representing the visual aspects of the image is combined with the embedding representing the textual aspects of the image to generate a final embedding for the image.

Utilizing visual and textual aspects of images with recommendation systems

Disclosed are systems, methods, and non-transitory computer-readable media for expanding search queries. A search system executes a search query based on a search term and the geographic indicator. In response to determining that a number of the search results is less than a threshold number, the search system determines, based on historical search logs from other users in the first geographic region, a likelihood value indicating a likelihood that the other users in the first geographic region expanded the geographic region of their search queries. The search system compares the likelihood value to a threshold likelihood value, and determines, based on the comparison, that the likelihood value meets or exceeds the threshold likelihood value. The search system then executes an expanded search based on the search term and an expanded geographic indicator that encompasses the first geographic region.

Expanding search queries

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

In an example embodiment, a machine learning algorithm is used to train a deep semantic similarity neural network to output a semantic similarity score between a candidate job search query and a candidate job search result. This semantic similarity score can then be used in a ranking phase to rank job search results in response to a first job search query.

Semantic similarity for machine learned job posting result ranking model

Disclosed are systems, methods, and non-transitory computer-readable media for expanding search queries. A search system determines a set of candidate alternate search terms based on historical search logs that include records of previously submitted search terms, corresponding search results that were presented to users, and corresponding search results that were selected by the users. The set of candidate alternate search terms is selected from titles of the corresponding search results that were selected by the users. The search system ranks the set of candidate alternate search terms based on determined probabilities that each of the alternate candidate search terms will be selected if presented to a user, and selects a first candidate alternate search term from the set of candidate alternate search terms based on the ranking. The search system generates an expanded search term based on the first candidate alternate search term.

In an example embodiment, a machine learning algorithm is used to train a query-based deep semantic similarity neural network to output a query context vector in a vector space that includes both query context vectors and document context vectors. Both the query context vectors and document context vectors are clustered using a clustering algorithm. When an input search query is obtained, the input search query is also passed into the query-based deep semantic similarity neural network and its output document context vector assigned to a first cluster based on the clustering algorithm. Documents within the first cluster are then retrieved in response to the input search query.

Semantic clustering based retrieval for candidate set expansion

Method and system to generate multiple entity aware typeahead suggestions is provided. The system is configured to use multiple Finite State Transducers (FSTs) to examine an input string submitted by a user via a search box, and to generate one or more typeahead suggestions based on the results of the examination. Different FSTs are constructed with respect to strings identified as associated with different entity types. At least one of the typeahead suggestions includes a portion associated with one entity type and a portion associated with a different entity type.

Multiple entity aware typeahead in searches

Growing community of inventors

Newark, CA, United States of America

Saurabh Kataria