Adobe, Inc.

Moumita Sinha

Harvineet Singh

Gaurush Hiranandani

Kandarp Sunil Khandwala

Dharwar Prasanna Kumar Tejas

Franck Dernoncourt

Balaji Vasan Srinivasan

Trung Huu Bui

Niyati Himanshu Chhaya

Vishwa Vinay

Fan Du

Shiv Kumar Saini

Doo Soon Kim

Sana Malik Lee

Trevor Paulsen

Anandhavelu Natarajan

Kushal Chawla

Harsh Jhamtani

Kokil Jaidka

Meghanath Macha Yadagiri

Tathagata Sengupta

Shankar Srinivasan

Frederic Mary

Pari Sawant

Niloy Ganguly

Shriram V S Revankar

Varun Gupta

Faran Ahmad

Vaishnav Pawan Madandas

Aditya Jain

Anthony M Colas

Po Ming Law

This disclosure involves detecting biases in predictive models and the root cause of those biases. For example, a processing device receives test data and training data from a client device. The processing device identifies feature groups from the training data and the test data generates performance metrics and baseline metrics for a feature group. The processing device detects biases through a comparison of the performance metrics and the baseline metrics the feature group. The processing device then isolates a portion of the training data that corresponds to the detected bias. The processing device generates a model correction usable to remove the bias from the predictive model.

Methods and systems for detection and isolation of bias in predictive models

The present disclosure relates to systems, methods, and non-transitory computer-readable media for generating pairs of natural language queries and corresponding query-language representations. For example, the disclosed systems can generate a contextual representation of a prior-generated dialogue sequence to compare with logical-form rules. In some implementations, the logical-form rules comprise trigger conditions and corresponding logical-form actions for constructing a logical-form representation of a subsequent dialogue sequence. Based on the comparison to logical-form rules indicating satisfaction of one or more trigger conditions, the disclosed systems can perform logical-form actions to generate a logical-form representation of a subsequent dialogue sequence. In turn, the disclosed systems can apply a natural-language-to-query-language (NL2QL) template to the logical-form representation to generate a natural language query and a corresponding query-language representation for the subsequent dialogue sequence.

Utilizing logical-form dialogue generation for multi-turn construction of paired natural language queries and query-language representations

This disclosure relates to methods, non-transitory computer readable media, and systems that generate and visualize bias scores within segment-generation-user interfaces prior to executing proposed actions with regard to target segments. For example, the disclosed systems can generate a bias score indicating a measure of bias for a characteristic within a segment of users selected for a proposed action and visualize the bias score and corresponding characteristic in a segment-generation-user interface. In some implementations, the disclosed systems can further integrate detecting and visualizing bias as a bias score with selectable options for a segmentation-bias system to generate and modify segments of users to reduce detected bias.

Generating and visualizing bias scores representing bias in digital segments within segment-generation-user interfaces

The present disclosure relates applying a survival analysis to model when a particular recipient will view an electronic message. For example, one or more embodiments train a survivor function to model the time that will elapse, on a continuous scale, before a recipient will open an electronic message. For example, one or more embodiments involve accessing analytics training data and extracting a first set of features affecting the time that elapsed before past recipients opened an electronic message and a second set of features affecting whether the recipients opened the electronic message at all. The system then generates a mixture model modified survivor function and determines the effect of each feature set on its corresponding outcome to learn parameters for the mixture model modified survivor function.

Modeling time to open of electronic communications

The present disclosure is directed toward systems and methods for generating an un-subscription model and predicting whether a potential customer will un-subscribe from receiving electronic marketing content from a marketing source. For example, systems and methods described herein involve generating a prediction un-subscription model that predicts whether a potential customer is prone to un-subscribe from receiving future communications about a product or merchant in response to receiving a communication for the product or merchant. The systems and methods further involve determining an appropriate action to take with regard to a potential customer based on whether the potential customer is prone to un-subscribe from receiving future communications.

Predicting unsubscription of subscribing users

The present disclosure is directed towards methods and systems for determining a persuasiveness of a content item. The systems and methods receive a content item from a client device and analyze the content item. Analyzing the content item includes analyzing at least one textual element, at least one image element, and at least one layout element of the content item to determine a first persuasion score, a second persuasion score, and a third persuasion score of the elements the content item. The systems and methods also generate a persuasion score of the content item and provide the persuasion score of the content item to the client device.

Determining persuasiveness of user-authored digital content items

Systems and methods are disclosed herein for determining user behavior in an augmented reality environment. An augmented reality application executing on a computing system receives a video depicting a face of a person. The video includes a video frame. The augmented reality application augments the video frame with an image of an item selected via input from a user device associated with a user. The augmented reality application determines, from the video frame, a score representing an action unit. The action unit represents a muscle on the face of the person depicted by the video frame and the score represents an intensity of the action unit. The augmented reality application calculates, from a predictive model and based on the score, an indicator of intent of the person depicted by the video.

Augmented reality predictions using machine learning

Methods and systems for providing targeted marketing include using consumer-centric indices to identify users who are most conversant with marketing communications. In particular, one or more embodiments generate a model that indicates a probability of user interactions based on dynamic data. The dynamic data indicates a time to action for each user interaction with a marketing communication within an observation window. The model fits the dynamic data to a distribution and determines the parameters of the distribution. Using the parameters of the distribution, one or more embodiments calculate interest scores for users who have received marketing communications. One or more embodiments select a set of users as a target audience based on the interest scores and provide marketing communications to target audience.

Generating and utilizing a conversational index for marketing campaigns

Techniques are disclosed for identifying anomalies in small data sets, by identifying anomalies using a Generalized Extreme Student Deviate test (GESD test). In an embodiment, a data set, such as business data or a website metric, is checked for skewness and, if found to be skewed, is transformed to a normal distribution (e.g., by applying a Box-Cox transformation). The data set is checked for presence of trends and, if a trend is found, has the trend removed (e.g., by running a linear regression). In one embodiment, a maximum number of anomalies is estimated for the data set, by applying an adjusted box plot to the data set. The data set and the estimated number of anomalies is run through a GESD test, and the test identifies anomalous data points in the data set, based on the provided estimated number of anomalies. In an embodiment, a confidence interval is generated for the identified anomalies.

Anomaly detection at coarser granularity of data

In embodiments of identifying the end of an on-line cart session, an analytics application captures user click inputs on pages of a Web site, where the user click inputs include adding one or more items for purchase to an on-line cart associated with the Web site. The analytics application then utilizes a predictive model, as well as user and session features of the on-line cart session, to predict whether a previous user click input is the last user click input associated with the on-line cart session, indicating an end of the session. A notification can then be provided that the on-line cart session has ended based on the prediction of the last user click input associated with the on-line cart session. The analytics application or the marketer can then retarget a user associated with the on-line cart session, such as with a message pertaining to the on-line cart session.

Identifying the end of an on-line cart session

Predicting unsubscription of potential customers

Growing community of inventors

Bengaluru, India

Moumita Sinha