Microsoft Technology Licensing, LLC

Microsoft Corporation

Microsoft Corporaiton

Yun-Cheng Ju

Alejandro Acero

Dong Yu

Ye-Yi Wang

David Gareth Ollason

Siddharth Bhatia

Geoffrey G Zweig

Ashwarya Poddar

Ivan Jelev Tashev

Royi Ronen

Andreas Stolcke

Fileno A Alleva

Oron Nir

Ami Turgman

Edan Hauon

Li Deng

Hsiao-Wuen Hon

Michael Lewis Seltzer

Mei-Yuh Hwang

Yu Shrike Zhang

Xiaolong Li

Peter K L Mau

Stefanie Lyn Tomko

Alex Acero

Lei Lu

Xuedong D Huang

Christopher A Meek

Dilek Zeynep Hakkani-Tur

Gokhan Tur

Andrew Lovitt

Sumit Gulwani

William B Dolan

James G Droppo

Bongshin Lee

Bo Thiesson

Milind V Mahajan

Kris Kuppuswamy Ganjam

Julian James Odell

Kaushik Chakrabarti

Xiao Li

Christopher John Brockett

Li Jiang

Matthai Philipose

Patrick Pantel

Michael J Rozak

Mark Marron

Leonard Alan Collins

Prabhdeep Singh

Daniel Herron

Timothy Seung Yoon Paek

Piali Choudhury

Zhan Ding

Chad R Heinemann

Neal Alan Bernstein

Noelle Mallory Sophy

David Kearney Burton

Svitlana Volkova

Wei-Ting Frank Liu

Seungyeop Han

Frank Liu

Alex Aero

Craig M Fisher

Mark Cecys

Methods for speaker role determination and scrubbing identifying information are performed by systems and devices. In speaker role determination, data from an audio or text file is divided into respective portions related to speaking parties. Characteristics classifying the portions of the data for speaking party roles are identified in the portions to generate data sets from the portions corresponding to the speaking party roles and to assign speaking party roles for the data sets. For scrubbing identifying information in data, audio data for speaking parties is processed using speech recognition to generate a text-based representation. Text associated with identifying information is determined based on a set of key words/phrases, and a portion of the text-based representation that includes a part of the text is identified. A segment of audio data that corresponds to the identified portion is replaced with different audio data, and the portion is replaced with different text.

System and method for speaker role determination and scrubbing identifying information

Solutions for automated labeling of child objects within tagged parents include: receiving a plurality of parent objects, each having a tag and including a plurality of child objects; receiving a machine learning (ML) component operable to rank objects according to relevance to queries; for each parent object: generating a set of restricted objects, wherein each restricted object corresponds to each child object in the plurality of child objects; for each of a plurality of queries, ranking, with the ML component, the restricted objects according to relevance; based at least on the query and an inverse of the rank of the restricted objects, assigning a child object label.

Automated labeling of child objects within tagged parents

Solutions for automated labeling of child objects within tagged parents include: receiving a plurality of parent objects, each having a tag and including a plurality of child objects; receiving a machine learning (ML) component operable to rank the parent objects; for each parent object: generating a set of restricted objects, wherein each restricted object is missing a child object; for each of a plurality of queries, ranking, with the ML component, the restricted objects according to relevance; based at least on the query and an inverse of the rank of the restricted objects, assigning a child object label. In some examples, the object missing from the restricted object having the lowest relevance rank is assigned a label based on the query. Various advantageous uses of the assigned child labels include enhancing search result reporting, automatically generating ML training data, and automatically generating and/or modifying reports.

Described herein are various technologies pertaining to performing an operation relative to tabular data based upon voice input. An ASR system includes a language model that is customized based upon content of the tabular data. The ASR system receives a voice signal that is representative of speech of a user. The ASR system creates a transcription of the voice signal based upon the ASR being customized with the content of the tabular data. The operation relative to the tabular data is performed based upon the transcription of the voice signal.

Performing an operation relative to tabular data based upon voice input

A system is provided that trains a spoken language understanding (SLU) classifier. A corpus of user utterances is received. For each of the user utterances in the corpus, the user utterance is semantically parsed, and the result of this semantic parsing is represented as a rooted semantic parse graph. The parse graphs representing all of the user utterances in the corpus are then combined into a single corpus graph that represents the semantic parses of the entire corpus. The user utterances in the corpus are then clustered into intent-wise homogeneous groups of user utterances, where this clustering includes finding subgraphs in the corpus graph that represent different groups of user utterances, and each of these different groups has a similar user intent. The intent-wise homogeneous groups of user utterances are then used to train the SLU classifier, and the trained SLU classifier is output.

Clustering user utterance intents with semantic parsing

Different advantageous embodiments provide a crowdsourcing method for modeling user intent in conversational interfaces. One or more stimuli are presented to a plurality of describers. One or more sets of describer data are captured from the plurality of describers using a data collection mechanism. The one or more sets of describer data are processed to generate one or more models. Each of the one or more models is associated with a specific stimulus from the one or more stimuli.

Crowdsourced, grounded language for intent modeling in conversational interfaces

Systems and methods are described for adding entries to a custom lexicon used by a speech recognition engine of a speech interface in response to user interaction with the speech interface. In one embodiment, a speech signal is obtained when the user speaks a name of a particular item to be selected from among a finite set of items. If a phonetic description of the speech signal is not recognized by the speech recognition engine, then the user is presented with a means for selecting the particular item from among the finite set of items by providing input in a manner that does not include speaking the name of the item. After the user has selected the particular item via the means for selecting, the phonetic description of the speech signal is stored in association with a text description of the particular item in the custom lexicon.

Pronunciation learning from user correction

The described implementations relate to automated data cleanup. One system includes a language model generated from language model seed text and a dictionary of possible data substitutions. This system also includes a transducer configured to cleanse a corpus utilizing the language model and the dictionary. The transducer can process speech recognition data in some cases by substituting a second word for a first word which shares pronunciation with the first word but is spelled differently. In some cases, this can be accomplished by establishing corresponding probabilities of the first word and second word based on a third word that appears in sequence with the first word.

Automated data cleanup by substitution of words of the same pronunciation and different spelling in speech recognition

A development system is described for facilitating the development of a spoken natural language (SNL) interface. The development system receives seed templates from a developer, each of which provides a command phrasing that can be used to invoke a function, when spoken by an end user. The development system then uses one or more development resources, such as a crowdsourcing system and a paraphrasing system, to provide additional templates. This yields an extended set of templates. A generation system then generates one or more models based on the extended set of templates. A user device may install the model(s) for use in interpreting commands spoken by an end user. When the user device recognizes a command, it may automatically invoke a function associated with that command. Overall, the development system provides an easy-to-use tool for producing an SNL interface.

Facilitating development of a spoken natural language interface

A database having listings rather than long documents is searched using a term frequency-inverse document frequency (Tf/Idf) algorithm.

Searching a database of listings

The subject disclosure is directed towards training a classifier for spoken utterances without relying on human-assistance. The spoken utterances may be related to a voice menu program for which a speech comprehension component interprets the spoken utterances into voice menu options. The speech comprehension component provides confirmations to some of the spoken utterances in order to accurately assign a semantic label. For each spoken utterance with a denied confirmation, the speech comprehension component automatically generates a pseudo-semantic label that is consistent with the denied confirmation and selected from a set of potential semantic labels and updates a classification model associated with the classifier using the pseudo-semantic label.

Spoken utterance classification training for a speech recognition system

A method for managing an interaction of a calling party to a communication partner is provided. The method includes automatically determining if the communication partner expects DTMF input. The method also includes translating speech input to one or more DTMF tones and communicating the one or more DTMF tones to the communication partner, if the communication partner expects DTMF input.

Telephony service interaction management

This document describes word-dependent language models, as well as their creation and use. A word-dependent language model can permit a speech-recognition engine to accurately verify that a speech utterance matches a multi-word phrase. This is useful in many contexts, including those where one or more letters of the expected phrase are known to the speaker.

Word-dependent language model

Growing community of inventors

Bellevue, WA, United States of America

Yun-Cheng Ju