Microsoft Technology Licensing, LLC

Microsoft Corporation

Kunal Mukerjee

Thomas W Holcomb

Sridhar Srinivasan

Pohsiang Hsu

Bruce Chih-Lung Lin

Shankar Lakshmi Regunathan

Chih-Lung Lin

Sorin Gherman

Surajit J Chaudhuri

Henricus Johannes Maria Meijer

Lev Novik

John Carlton Platt

Efim Hudis

Christian Liensberger

Christopher Alan Hays

Venkatesh Krishnan

R Donald Thompson

Ziv Kasperski

Stéphane Nyombayire

Adam Prout

PalaniRaj Kaliyaperumal

Kuralmani Elango

Santosh Balasubramanian

Kazuhito Koishida

Julian James Odell

Jeffrey Cole

Jian-guang Lou

Weizhu Chen

Haiyuan Cao

Yu Hu

Lily Xinchen Liu

Yi Liu

Rafael Ballesteros

Zeqi Lin

Tom W Holcomb

Brendan Meeder

Naveen Garg

Pohsinag Hsu

Examples described herein generate training data for machine learning (ML) for natural language (NL) processing (such as semantic parsing for translating NL). A formula tree is generated based on sampling both a formula grammar and NL templates. Using the formula tree, an ML training data instance pair is generated comprising a formula example and an NL example. A context example may also be used during instantiation of the formula tree. An ML model is trained with training data including the ML training data instance pair, and ML output is generated from NL input. The ML output includes, for example, a machine-interpretable formula, a database querying language command, or a general programming language instruction. Some examples support context-free grammar, probabilistic context-free grammar, and/or non-context-free production rules.

Generating model training data from a domain specification

The disclosure herein describes generating an application from input content. Input content of a content type is obtained, such as an image file, digital document file, or the like. A content data extractor is selected from a set of content data extractors based on the content type. A set of content entities is extracted from the obtained input content using the selected content data extractor, such as text labels, text boxes, buttons, or the like. The set of content entities are normalized according to a standard interface schema and an application template is generated using the normalized set of content entities, whereby an application can be developed using the generated application template. The disclosure enables application interfaces to be designed using a variety of methods and for those different types of designs to be efficiently converted to a functional application.

Creating applications and templates based on different types of input content

The present invention extends to methods, systems, and computer program products for forming data responsive to a query. Responding to a query can include selecting data views on top of data sources. A data catalog stores data elements from different data sources and stores links between data elements from the different data sources. A view can be selected for a query based on one or more of: classification of data included in the data catalog or prior usage of data included in the data catalog. A selected data view is populated with data elements from the data catalog using the links to determine that data elements spanning multiple data sources satisfy the query. The data catalog can be used to identify relationships between data elements from different data sources that would not otherwise be identified.

Forming data responsive to a query

In one embodiment, datasets are stored in a catalog. The datasets are enriched by establishing relationships among the domains in different datasets. A user searches for relevant datasets by providing examples of the domains of interest. The system identifies datasets corresponding to the user-provided examples. The system them identifies connected subsets of the datasets that are directly linked or indirectly linked through other domains. The user provides known relationship examples to filter the connected subsets and to identify the connected subsets that are most relevant to the user's query. The selected connected subsets may be further analyzed by business intelligence/analytics to create pivot tables or to process the data.

Finding data in connected corpuses using examples

The present invention extends to methods, systems, and computer program products for creating data views. Embodiments of the invention allow automatic creation of data views/projections on top of data sources. A view can be created on one data source (by looking at one or multiple data elements within the data source) or by looking at multiple data sources (of the same or different data types) and combining elements into a view. Data sources can be any kind of data source, including but not limited to: databases, files, and Web services. Views can be picked from a set of predefined views, picked from views of another tenant (either statically or using fuzzy logic), selected based on data element classifications and/or data element relationship classifications, selected based on data usage (e.g., by looking at logs), etc.

Creating data views

Techniques and tools for encoding and decoding motion vector information for video images are described. For example, a video encoder yields an extended motion vector code by jointly coding, for a set of pixels, a switch code, motion vector information, and a terminal symbol indicating whether subsequent data is encoded for the set of pixels. In another aspect, an encoder/decoder selects motion vector predictors for macroblocks. In another aspect, a video encoder/decoder uses hybrid motion vector prediction. In another aspect, a video encoder/decoder signals a motion vector mode for a predicted image. In another aspect, a video decoder decodes a set of pixels by receiving an extended motion vector code, which reflects joint encoding of motion information together with intra/inter-coding information and a terminal symbol. The decoder determines whether subsequent data exists for the set of pixels based on e.g., the terminal symbol.

Coding of motion vector information

Methods, systems, and computer program products are provided for deriving and updating document similarity indices for a plurality of documents. The number of maintained similarities can be controlled to conserve CPU and storage resources.

Deriving document similarity indices

A method includes receiving a concept hierarchy at a computing device. The concept hierarchy identifies concepts associated with a document corpus. An interface based on the concept hierarchy is generated. The interface is operable to navigate, search, and modify the concept hierarchy. The method includes transmitting the interface for display to a display device.

Interface to navigate and search a concept hierarchy

The present invention extends to methods, systems, and computer program products for deriving document similarity indices. Embodiments of the invention include scalable and efficient mechanisms for deriving and updating a document similarity index for a plurality of documents. The number of maintained similarities can be controlled to conserve CPU and storage resources.

The present invention extends to methods, systems, and computer program products for identifying key phrases within documents. Embodiments of the invention include using a tag index to determine what a document primarily relates to. For example, an integrated data flow and extract-transform-load pipeline, crawls, parses and word breaks large corpuses of documents in database tables. Documents can be broken into tuples. The tuples can be sent to a heuristically based algorithm that uses statistical language models and weight+cross-entropy threshold functions to summarize the document into its 'top N' most statistically significant phrases. Accordingly, embodiments of the invention scale efficiently (e.g., linearly) and (potentially large numbers of) documents can be characterized by salient and relevant key phrases (tags).

Identifying key phrases within documents

Embodiments for implementing a speech recognition system that includes a speech classifier ensemble are disclosed. In accordance with one embodiment, the speech recognition system includes a classifier ensemble to convert feature vectors that represent a speech vector into log probability sets. The classifier ensemble includes a plurality of classifiers. The speech recognition system includes a decoder ensemble to transform the log probability sets into output symbol sequences. The speech recognition system further includes a query component to retrieve one or more speech utterances from a speech database using the output symbol sequences.

Noise robust speech classifier ensemble

A clustering tool to generate word clusters. In embodiments described, the clustering tool includes a clustering component that generates word clusters for words or word combinations in input data. In illustrated embodiments, the word clusters are used to modify or update a grammar for a closed vocabulary speech recognition application.

Word clustering for input data

Growing community of inventors

Redmond, WA, United States of America

Kunal Mukerjee