Microsoft Technology Licensing, LLC

Amit Kumar Agarwal

Christopher Hakan Basoglu

Yosseff Levanoni

Weirong Zhu

Zoltan Romocsa

Shuangyu Chang

Sayan Dev Pathak

Yifan Gong

Hosam Adel Khalil

Ziad Al Bawab

Anand U Desai

Xuedong D Huang

Wei Xiong

Eyal Krupka

Adi Diamant

Lijuan Qin

Xiong Xiao

Emilian Y Stoimenov

Chaojun Liu

Shixiong Zhang

Karen Master Ben-Dor

Ido Leichter

Moshe David

Raz Halaly

Lingfeng Wu

Aviv Hurvitz

Yoni Smolin

Ilya Gurvich

Amy Parag Shah

Nathan E Wohlgemuth

Naveen Parihar

Jinyu Li

Lingli Zhang

Shalendra Chhabra

Changliang Liu

Liang Lu

Piyush Behre

Veljko Miljanic

Rishi Girish

Sharman W Tan

Yongkang Zhu

Aadyot Bhatnagar

Ayush Vikram

Andrii Matukhno

Mykola Denysiuk

Tomas Bergl

Skyler Michael Anderson

Mikhail Raer

Halley Weitzman

Michael J Shelton

Harini Kesavamoorthy

Sayan Pathak

Some disclosed embodiments are directed to obtaining a decoded audio data including a spoken language utterance recognized in audio data and identifying a disfluency in the decoded audio data. Upon determining that correcting the disfluency would improve a readability score of the decoded audio data, the system generates a particular correction to correct the disfluency and applies the particular correction to the decoded audio data. Then, an updated decoded audio data is generated which reflects the particular correction. The updated decoded audio data has improved readability over the decoded audio data.

Systems and methods for GPT guided neural punctuation for conversational speech

A data processing system implements obtaining a first textual content, segmenting the first textual content into a plurality of first segments, and providing each segment of the plurality of first segments to a first natural language processing (NLP) model to obtain a set of first readability scores for the plurality of first segments. The first NLP model is configured to analyze a textual input and to output a readability score representing a measurement of readability of the textual input. The system further implements aggregating the set of first segment readability scores to determine a first readability score for the first textual content, and perform at least one of causing the first readability score to be presented to a user or performing one or more actions on the first textual content based on the readability score.

Automated artificial intelligence driven readability scoring techniques

Acoustic features are batched into two different batches. The second batch of the two batches is made in response to a detection of a word hypothesis output by a speech recognition network that received the first batch. The number of acoustic feature frames of the second batch is equal to a second batch size greater than the first batch size. The second batch is also to the speech recognition network for processing.

Adaptive batching to reduce recognition latency

A method for facilitating a remote conference includes receiving a digital video and a computer-readable audio signal. A face recognition machine is operated to recognize a face of a first conference participant in the digital video, and a speech recognition machine is operated to translate the computer-readable audio signal into a first text. An attribution machine attributes the text to the first conference participant. A second computer-readable audio signal is processed similarly, to obtain a second text attributed to a second conference participant. A transcription machine automatically creates a transcript including the first text attributed to the first conference participant and the second text attributed to the second conference participant.

Computerized intelligent assistant for conferences

A system includes acquisition of meeting data associated with a meeting, determination of a plurality of meeting participants based on the acquired meeting data, acquisition of e-mail data associated with each of the plurality of meeting participants, generation of a meeting language model based on the acquired e-mail data and the meeting data, and transcription of audio associated with the meeting based on the meeting language model.

Meeting-adapted language model for speech recognition

A system includes acquisition of a domain grammar, determination of an interpolated grammar based on the domain grammar and a base grammar, determination of a delta domain grammar based on an augmented first grammar and the interpolated grammar, determination of an out-of-vocabulary class based on the domain grammar and the base grammar, insertion of the out-of-vocabulary class into a composed transducer composed of the augmented first grammar and one or more other transducers to generate an updated composed transducer, composition of the delta domain grammar and the updated composed transducer, and application of the composition of the delta domain grammar and the updated composed transducer to an output of an acoustic model.

Online language model interpolation for automatic speech recognition

Streaming machine learning unidirectional models is facilitated by the use of embedding vectors. Processing blocks in the models apply embedding vectors as input. The embedding vectors utilize context of future data (e.g., data that is temporally offset into the future within a data stream) to improve the accuracy of the outputs generated by the processing blocks. The embedding vectors cause a temporal shift between the outputs of the processing blocks and the inputs to which the outputs correspond. This temporal shift enables the processing blocks to apply the embedding vector inputs from processing blocks that are associated with future data.

Streaming contextual unidirectional models

Embodiments may include collection of a first batch of acoustic feature frames of an audio signal, the number of acoustic feature frames of the first batch equal to a first batch size, input of the first batch to a speech recognition network, collection, in response to detection of a word hypothesis output by the speech recognition network, of a second batch of acoustic feature frames of the audio signal, the number of acoustic feature frames of the second batch equal to a second batch size greater than the first batch size, and input of the second batch to the speech recognition network.

A context-aware transcription system includes a language model preparation service that retrieves meeting-specific data prior to or during a meeting. The language model preparation service utilizes the meeting-specific data to generate a meeting-specific statistical language model. A speech transcription service can utilize the meeting-specific statistical language model to generate a transcription of audio generated by attendees of a meeting. The system can transmit the transcription to computing devices associated with meeting attendees during the meeting for presentation in a user interface in real time. The language model preparation service can generate the meeting-specific statistical language model in response to receiving a pre-meeting signal. The pre-meeting signal can be generated a predetermined time prior to meetings according to a schedule. The pre-meeting signal might also, or alternately, be generated in response to the receipt of meeting change event notifications before or during meetings that describe changes to the meetings.

Context-aware real-time meeting audio transcription

Embodiments are directed to optimizing data transfers between heterogeneous memory arenas. In one scenario, a computer system receives an indication that a data chunk is to be transferred from a first memory arena to a third memory arena, and then determines that for the data chunk to be transferred from the first memory arena to the third arena, the data chunk is to be transferred from the first memory arena to a second memory arena, and from the second memory arena to the third memory arena. The computer system divides the data chunk into smaller data portions and copies a first data portion from the first memory arena to the second memory arena. The computer system then copies the first data portion from the second memory arena to the third memory arena and copies a second data portion from the first memory arena to the second memory arena in parallel.

Optimizing data transfers between heterogeneous memory arenas

The present invention extends to methods, systems, and computer program products for aliasing buffers. Embodiment of the inventions supporting buffer aliasing through introduction of a level of indirection between a source program's buffer accesses and the target executable physical buffers, and binding the logical buffer accesses to actual physical buffer accesses at runtime. A variety of techniques for can be used supporting runtime aliasing of buffers, in a system which otherwise disallows such runtime aliasing between separately defined buffers in the target executable code. Binding of logical buffer accesses in the source program to the actual physical buffers defined in the target executable code is delayed until runtime.

Aliasing buffers

The present invention extends to methods, systems, and computer program products for debugging in a multiple address space environment. Embodiments of the invention include techniques for recording debug information used for translating between an abstract unified address space and multiple address spaces at a target system (e.g., a co-processor, such as, a GPU or other accelerator). A table is stored in the recorded debug information. The table includes one or more entries mapping compiler assigned IDs to address spaces. During debugging within a symbolic debugger, the recorded debug information can be used for viewing program data across multiple address spaces in a live debugging session.

Debugging in a multiple address space environment

The present invention extends to methods, systems, and computer program products for optimizing execution of kernels. Embodiments of the invention include an optimization framework for optimizing runtime execution of kernels. During compilation, information about the execution properties of a kernel are identified and stored alongside the executable code for the kernel. At runtime, calling contexts access the information. The calling contexts interpret the information and optimize kernel execution based on the interpretation.

Optimizing execution of kernels

Growing community of inventors

Redmond, WA, United States of America

Amit Kumar Agarwal