Microsoft Technology Licensing, LLC

Other

International Business Machines Corporation

Royi Ronen

Oron Nir

Ohad Jassin

Roy Eisenstadt

Efim Hudis

Hani Hana Neuvirth

Daniel Alon

Mor Geva Pipek

Abedelkader Asi

Alexander Tsvetkov

Yun-Cheng Ju

Andreas Stolcke

Ashwarya Poddar

Avner Levi

Elad Ziklik

Hanan Shteingart

Tomer Teller

Ami Turgman

Yarin Kuper

Dean Geckt

Abed El Kader Asi

Edan Hauon

Moshe Israel

Elad Yom-Tov

Olivier Nano

Omer Karin

Tomer Koren

Roee Oz

Hani Neuvirth-Telem

Erez Altus

Ika Bar-Menachem

Mathias Scherman

Gilad Michael Elyashar

Ori Ziv

Milan Gada

Neta Haiby-Weiss

Amir Pinchas

Hanan Lavy

Yitzhak Tzahi Weisfeld

Yair Snir

Roy Levin

Anand Ranganathan

Ben Kliger

Chin-Yew Lin

Yuri Gurevich

Philip Newman

Yair Tor

Yuval Eldar

Weizhu Chen

Tom Jurgenson

Shay Kels

Eylon Ami

Shahar Keren

Yi Mao

Gal Lavee

Oran Brill

Corina Feuerstein

Vlad Korsunsky

Song Wang

Daniel Mark Edwards

Yuri Gabaev

Ori Kashi

Tomer Brand

Ran Mizrachi

Hani-Hana Neuvirth

Tomer Rosenthal

Daniel Nurieli

Rona Shaanan

Pooja Bhalla

Atul Agarwal

Greeshma Singh

Michal Braverman-Blumenstyk

Alexander Ostrikov

Olga Goldenberg

Shimrit Rada Bemis

Ashwini Kumar Dalai

Peiheng Hu

Lars Mohr

Shai Baruch Nahum

Oleg Yanovsky

Liron Izhaki-Allerhand

Daniel Alon Hilevich

Roey Vilnai

A computerized method categorizes questions from a transcript and provides those categorized questions to targets. A transcript associated with a meeting is obtained and a question in the obtained transcript is detected by a question detection model. A question vector embedding of the detected question is generated using a question clustering model and a category of the detected question is determined by the question clustering model using the generated question vector embedding and a plurality of category vector embeddings. A target of the detected question is identified using the determined category and the detected question is provided to the identified target via a question provision interface. Further, the question clustering model is used to generate the category vector embeddings from example questions of category question catalogs, such that the set of categories for which the question clustering model is configured can be efficiently modified.

Categorizing questions from a transcript using vector embeddings for information retrieval

A computer-implemented method is provided that detects jailbreak attempts against generative models, which may involve a shift between benign and malicious content. The method includes determining a probability-based metric for each of a plurality of tokens in a target text using a language model, the probability-based metric being based on a probability at least one preceding token. The probability-based metrics are processed to identify a subset of the plurality of tokens having a change in the probability-based metric with respect to others of the plurality of the tokens not within the subset of the plurality of tokens. A jailbreak attempt in the target text is detected in response to identifying the change in the probability-based metric in the subset of the plurality of tokens.

Detecting jailbreak attempts on generative models

Example solutions for reducing the likelihood of hallucinations by language models, such as large language models (LLMs) are disclosed. By injecting a sufficient range and quantity of curated factual data into a prompt, the likelihood of a hallucination by an LLM may be reduced. This enables language models to be used in a wider range of settings, in which fabrication of facts is problematic, while reducing the need for a human to carefully check the generated text for accuracy. Examples include: generating a summary of a transcript using a summarization model; extracting topic-specific data from stored data using a scoring model; dynamically generating a language model prompt using the topic-specific data and the summary; and generating an output text using a language model and the language model prompt.

Dynamic construction of large language model prompts

The disclosure herein describes interpreting attention-based decisions of summarization outputs generated by a deep learning model. A decision interpretation model obtains attention values defining connections between input tokens associated with a source text and output tokens for a selected portion of a summary associated with the source text. The input tokens having the highest attention values indicating the strongest connections between the input tokens of the source text and an output token of the summary are selected as primary tokens. A semantic similarity between the primary tokens for each attention head and an output token is calculated. The model selects the primary tokens having the closest semantic similarity with the summary portion. A visual cue is generated on or within a portion of the source text corresponding to the primary tokens. The visual cue identifies dominant words in the source text used to explain the summary portion.

Interpreting summarization model decisions based on attention

The disclosure herein describes reducing training bias in outputs generated by a generative language model. A communication segment associated with a communication is obtained by at least one processor of a generative language model. An output value associated with the communication segment is generated by the generative language model. The output value is mapped to a set of training bias values associated with the generative language model and based on the mapping of the output value to a training bias value of the set of training bias values, an alternative output value is generated. The alternative output value is used in a generated segment output for the communication segment. The accuracy of segment outputs generated by the generative language model is improved through reducing or eliminating its training biases.

Reducing biases of generative language models

Aspects of the technology described herein improve an object recognition system by specifying a type of picture that would improve the accuracy of the object recognition system if used to retrain the object recognition system. The technology described herein can take the form of an improvement model that improves an object recognition model by suggesting the types of training images that would improve the object recognition model's performance. For example, the improvement model could suggest that a picture of a person smiling be used to retrain the object recognition system. Once trained, the improvement model can be used to estimate a performance score for an image recognition model given the set characteristics of a set of training of images. The improvement model can then select a feature of an image, which if added to the training set, would cause a meaningful increase in the recognition system's performance.

Training set sufficiency for image analysis

A method of generating keyword-based dialogue summaries is provided. The method includes inputting a transcript of an audio conversation and a keyword into a machine learning model trained based on encodings representing the keyword and the transcript, generating computer-generated text different from and semantically descriptive of the transcript and semantically associated with the keyword, and outputting the computer-generated text in association with a selectable item selectable for inclusion of the computer-generated text in displayed text representing the transcript, the selectable item associated with the keyword.

Keyword-based dialogue summarizer

A next operation in a workflow corresponding to a subject matter is automatically generated. A subject matter data record is received, including one or more historical operations previously executed in the workflow and a workflow template including one or more dynamic workflow nodes. At least one next operation is programmatically generated from the one or more dynamic workflow nodes based on the one or more historical operations previously executed in the workflow, wherein the at least one next operation is parameterized based on content in the subject matter data record. The at least one programmatically generated next operation is output into the workflow for execution.

Dynamic next operation determination for workflows

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

The disclosure herein describes determining topics of communication transcripts using trained summarization models. A first communication transcript associated with a first communication is obtained and divided into a first set of communication segments. A first set of topic descriptions is generated based on the first set of communication segments by analyzing each communication segment of the first set of communication segments with a generative language model. A summarization model is trained using the first set of communication segments and associated first set of topic descriptions as training data. The trained summarization model is then applied to a second communication transcript and, based on applying the trained summarization model to the second communication transcript, a second set of topic descriptions of the second communication transcript is generated. By training the summarization model based on output of the generative language model, it enables efficient, accurate generation of topic descriptions from communication transcripts.

Determining topic labels for communication transcripts based on a trained generative summarization model

A recommendation system for recommending a target feature value for a target feature for a target deployment is provided. The recommendation system, for each of a plurality of deployments, collects feature values for the features of that deployment. The recommendation system then generates a model for recommending a target feature value for the target feature based on the collected feature values of the features for the deployments. The recommendation system applies the model to the features of the target deployment to identify a target feature value for the target feature. The recommendation system then provides the identified target feature value as a recommendation for the target feature for the target deployment.

Automatic recommendations for deployments in a data center

In various embodiments, methods and systems for implementing a media management system, for video data processing and adaptation data generation, are provided. At a high level, a video data processing engine relies on different types of video data properties and additional auxiliary data resources to perform video optical character recognition operations for recognizing characters in video data. In operation, video data is accessed to identify recognized characters. A video OCR operation to perform on the video data for character recognition is determined from video character processing and video auxiliary data processing. Video auxiliary data processing includes processing an auxiliary reference object; the auxiliary reference object is an indirect reference object that is a derived input element used as a factor in determining the recognized characters. The video data is processed based on the video OCR operation and based on processing the video data, at least one recognized character is communicated.

Media management system for video data processing and adaptation data generation

Methods, systems, and media are shown for reducing the vulnerability of user accounts to attack that involve creating a rule for a user account that includes a permitted parameter corresponding to a user account activity property, monitoring the account activity of the user account. If it is determined that account activity property is inconsistent with the permitted parameter, then the user account is disabled. An example of a permitted parameter is a permitted time period, such as a start time, an end time, a recurrence definition, a days of the week definition, a start date, an end date, and a number of occurrences definition. Other examples are a physical parameter, such as a permitted geographic location, device, or network, or a permitted usage parameter, such as a permitted application, data access, or domain.

Account management using account activity usage restrictions

Data from social networking applications and other applications that can be used to communicate are combined for a user to generate a graph of the various relationships that the user has with other users in the social networking applications and other applications. In addition, the behaviors of each user with respect to communicating through the various social networking applications and other applications are monitored to generate task data that describes user preferences for communicating using each social networking application or other application for different tasks. At a later time, when a user is looking to connect with another user for an indicated task such as networking, the graph can be used to recommend paths to other users in the various social networking applications and other applications, and the generated task data can be used to rank the recommended paths based on the indicated task.

Identifying paths using social networking data and application data

Growing community of inventors

Tel Aviv, Israel

Royi Ronen