Microsoft Technology Licensing, LLC

Guowei Liu

Jian Wang

Bing Jiang Sun

Haidong Zhang

Yantao Li

Min Wang

Wenli Zhu

Shuguang Ye

Qiang Wang

Adnan Azfar Mahmud

Li Deng

Jianfeng Gao

Yingnong Dang

Xiaodong He

Ruofei Zhang

Yelong Shen

Gregory T Buehrer

Jianchang Mao

Yunhong Zhou

Qiang Lou

Adan Azfar Mahmud

Functionality is described herein for transforming first and second symbolic linguistic items into respective first and second continuous-valued concept vectors, using a deep learning model, such as a convolutional latent semantic model. The model is designed to capture both the local and global linguistic contexts of the linguistic items. The functionality then compares the first concept vector with the second concept vector to produce a similarity measure. More specifically, the similarity measure expresses the closeness between the first and second linguistic items in a high-level semantic space. In one case, the first linguistic item corresponds to a query, and the second linguistic item may correspond to a phrase, or a document, or a keyword, or an ad, etc. In one implementation, the convolutional latent semantic model is produced in a training phase based on click-through data.

Convolutional latent semantic models and their applications

Described is a technology by which software instrumentation data collected from user program sessions are analyzed to output an analysis report or the like via example methods and an architecture configured for efficient operation. A client component queries a service for analysis related information. To process the query, the service works with a data manager, and via a high dimensional analysis component may use information processed from the software instrumentation data, such as in the form of one or more inverted indexes and/or raw value files. The service may include a usage analysis component, a feature recognition component that locates features from command sequences, a user recognition component and/or a program reliability component. One or more counterpart components at the client may generate analysis reports or the like based on the query results. The client also may maintain user libraries and feature libraries to facilitate analyses.

Analyzing software usage with instrumentation data

Described is a technology by which high dimensional source data corresponding to rows of records with identifiers, and columns comprising dimensions of data values, are processed into a file model for efficient access. An inverted index corresponding to any dimension is built by mapping data from raw dimension values to mapped values based on mapping entries in a dimension table. The record identifiers are arranged into subgroups based on their mapped value; a count and/or an offset may be maintained for locating each of the subgroups. The raw values for a dimension are maintained within a raw value file. For sparse data, the raw value file may be compressed, e.g., by excluding nulls and associating a record identifier with each non-null. A data manager provides access to data in the data files, such as by offering various functions, using caching for efficiency.

Efficient data infrastructure for high dimensional data analysis

Described is a technology by which high dimensional data may be efficiently analyzed, including by filtering, grouping, aggregating and/or sorting operations to provide an analysis result. For efficiency in the analysis, an inverted index may be built (e.g., as part of filtering), and/or a hash structure (e.g., as part of grouping). Analysis parameters specify dimensions, on which union and/or intersection operations are performed to provide a final dataset. The analysis tool provides a user interface for inputting analysis parameters and outputting information corresponding to an analysis result. The analysis tool may sort the information corresponding to the analysis result, e.g., to output the topmost or bottommost results.

Multidimensional analysis tool for high dimensional data

Described is a technology for analyzing usage of a software program's features. Software instrumentation data is collected during actual user program usage sessions. The collected data is then processed to determine various feature usage counts and other information, cross-feature usage (e.g., among users who use a feature, how many use another feature or program), and characteristics of feature users, e.g., how long, how much, how often and how extensive feature users use a program. Session analysis may be performed to provide information about the number of sessions in which a set of features occur. Feature usage trends over time may also be determined via analysis. A user interface is described for facilitating selection of one or more features to analyze, for facilitating selection of a group of users, and/or for outputting results corresponding to the analysis.

Software feature usage analysis and reporting

Described is a technology by which software program feature usage is located within a sequence of commands collected during program usage sessions. For example, feature generally corresponds to a series of commands, such as copy and paste. A visual modeling component is controlled via drag-and-drop operations to describe a feature model, which is then compiled by a compiler into a finite state machine. Noise models may be used to exclude any command in the sequence that is irrelevant to the feature usage. A recognition process uses the finite state machine to locate program feature usage within the sequence of recorded commands by matching command sub-sequences corresponding to the feature model via the state machine. An analyzer may then use the located matches to provide an analysis report on feature usage.

Software feature modeling and recognition

Growing community of inventors

Beijing, China

Guowei Liu