Open Access. Powered by Scholars. Published by Universities.®
Articles 2881 - 2910 of 4108
Full-Text Articles in Entire DC Network
Generative Video Models To Create Panning Videos Anchored On A User Image, Lenord Melvix Joseph Stephen Max
Generative Video Models To Create Panning Videos Anchored On A User Image, Lenord Melvix Joseph Stephen Max
Defensive Publications Series
Creating videos with the panning effect is a cumbersome multi-step process that needs advanced video editing expertise and a deep understanding of video editing tools and source images. These are barriers to the use of panning videos in online media content. This disclosure describes the use of cloud-based video generation models to automate the creation of panning videos of user-specified duration and quality based on a user-selected anchor and satellite imagery end point. The generative model automatically identifies a panning path between the end points, identifies images along different points in the panning path, and automatically generates a panning video …
Merchandise Returns Predictability, Deepak Tomar, Veerendra Singh
Merchandise Returns Predictability, Deepak Tomar, Veerendra Singh
Defensive Publications Series
The present disclosure pertains to the field of payment processing and predictive modelling for the Merchandise returns transactions in the context of payment transactions. More specifically, it focuses on developing a new model that offers predictability scores for merchandise returns, allowing merchants and acquirers to make informed decisions about the clearing of the original transactions. This innovative solution leverages historical data and employs linear regression to predict likelihood of a potential merchandise returns, potentially saving huge costs and resources for the merchant and acquirer and all the parties involved in the transaction processing.
Identifying Originators/Speakers Of Statements/Ideas In Virtual Meeting Environments, Hazim Dahir, Zaheer Aziz, Subha Dhesikan
Identifying Originators/Speakers Of Statements/Ideas In Virtual Meeting Environments, Hazim Dahir, Zaheer Aziz, Subha Dhesikan
Defensive Publications Series
In technology, entertainment, business, and/or legal proceedings, especially in scenarios in which significant monetary and/or legal rights are at stake, it can be important to determine when a statement was made and by whom the statement was made. As a majority of business is currently conducted virtually or as a hybrid (e.g., partial virtual presence), it would be desirable to be able to determine, with high certainty, an originator of an "idea", phrase, formula, etc. This proposal provides techniques that may be utilized to enable speaker attribution for virtual meeting environments.
Forecasting Engineering Team Capacity And Work Estimates Using Product Roadmaps, Saif Khalaf, Mailys Levassort, Amisha Singh, Kimberly Reardon
Forecasting Engineering Team Capacity And Work Estimates Using Product Roadmaps, Saif Khalaf, Mailys Levassort, Amisha Singh, Kimberly Reardon
Defensive Publications Series
Despite substantial research, deriving work estimates and team capacity for a given software development project continues to be notoriously difficult to get right. This disclosure describes techniques to automatically and rapidly calculate, in the same unit of measure, engineering team capacity and engineering work estimates for a given set of product features. With use of the described techniques, software development projects can be better organized, and engineering talent can spend time on engineering problems rather than on planning. The automated team capacity and work estimate calculations can be more accurate than typical human-generated estimates, and lead to more accurate product …
In-Conference Tool Virtual Assistant With Real Time Speaking Latency, Dongeek Shin
In-Conference Tool Virtual Assistant With Real Time Speaking Latency, Dongeek Shin
Defensive Publications Series
A framework is proposed for reducing latency for real-time speaking applications, such as virtual assistants. The framework utilizes a hybrid scheme between a large language model (LLM) and a small sound model (SSM). The SSM timestamps the end of a user’s verbal query and notifies the transcription engine that the query has ended. The transcription engine will then abort and send the existing set of transcriptions to the LLM for processing without waiting for the buffer to meet the specific value or time. This minimizes the net response time for real time speech-based models due to the reduced latency from …
Meeting Summarization And Alerts For Video Conference Participants, Shiblee Hasan, Kathleen Bryan
Meeting Summarization And Alerts For Video Conference Participants, Shiblee Hasan, Kathleen Bryan
Defensive Publications Series
A technique is proposed for summarizing meetings to help distracted participants transition back into conversations during video conferences. Processing logic may receive, from a video conferencing platform, a video file including multiple frames and audio. The processing logic may determine a period of time where a participant is distracted based on the participant’s body movement, eye movement, and camera settings. This determination can be made using a machine learning model. The processing logic can also generate a summary of the meeting based on received voice input, chat comments, and participant expressions. This summary may be generated using a generative artificial …
Burn-In Compensation With Foveated Rendering Display
Burn-In Compensation With Foveated Rendering Display
Defensive Publications Series
Systems, methods, and devices for compensating image(s) or video(s) data are provided. A device may divide a display panel including a plurality of pixels into one or more estimated blocks including pixels associated with an original image. The device may also calculate stress data separately for the estimated blocks by calculating a stress value of the pixels associated with the estimated block. The device may also generate an accumulated stress data based on an accumulation of stress data. The device may also degrade data associated with the original image to generate compensation data based on the accumulated stress data.
A Cost-Effective Method To Prevent Data Exfiltration From Llm Prompt Responses, Assaf Namer, Jim Miller, Hauke Vagts, Brandon Maltzman
A Cost-Effective Method To Prevent Data Exfiltration From Llm Prompt Responses, Assaf Namer, Jim Miller, Hauke Vagts, Brandon Maltzman
Defensive Publications Series
Large language models (LLMs) are susceptible to security risks wherein malicious attackers can manipulate LLMs by poisoning their training data or using malicious text prompts or queries designed to cause the LLM to return output that includes sensitive or confidential information, e.g., that is part of the LLM training dataset. This disclosure describes the use of a data loss prevention (DLP) system to protect LLMs against data exfiltration. The DLP system can be configured to detect specific data types that are to be prevented from being leaked. The LLM output, generated in response to a query from an application or …
A Method And System To Implement Adaptive Training Algorithm: Adam Q, Ka Lok Ng, Panos Chatzigiannis
A Method And System To Implement Adaptive Training Algorithm: Adam Q, Ka Lok Ng, Panos Chatzigiannis
Defensive Publications Series
The present disclosure relates to providing a method and system for training neural networks. It discloses an adaptive training algorithm Adam Q to overcome the challenges associated with its predecessors. It proposes a look-up table to be fused with the existing Adam algorithm such that Adam Q may not need to run all the computational operations while determining the updated weights and instead may directly see the results from the fused look-up table. It further goes on to propose a quantization technique where the received inputs from the previous iterations are first quantized and then taken as an input for …
Infinity Pay, Sahil Arora
Infinity Pay, Sahil Arora
Defensive Publications Series
The present disclosure relates to the field of contactless payment systems and biometric authentication. More specifically, it relates to a novel system and method for enabling secure contactless payments using biometric identification, even in situations where the user's primary device is unavailable. This innovative solution incorporates a combination of machine learning, neural networks, blockchain technology, and interconnected IoT devices to facilitate seamless and secure transactions.
Large Language Model Powered Chatbot For Natural Language Query Responses, Priya Mathur, Zhaoying Han
Large Language Model Powered Chatbot For Natural Language Query Responses, Priya Mathur, Zhaoying Han
Defensive Publications Series
While dashboards provide structured information on various metrics, they cannot answer specialized questions. For such questions, data scientists need to develop custom database queries based on natural language questions specified by users, which is costly and time-consuming. This disclosure describes a virtual data assistant that leverages large language models to automatically translate natural language queries to structured database queries and to translate structured responses into natural language answers. The models can be trained on pairs {natural language query, structured query} and {structured answer, natural language answer}. The virtual data assistant can be implemented as a chatbot that enables users to …
System And Method For Dynamic Link Activation In Email Messages For Phishing Prevention, Alok Roy
System And Method For Dynamic Link Activation In Email Messages For Phishing Prevention, Alok Roy
Defensive Publications Series
The present disclosure relates to a system and a method for dynamic link activation in email messages to prevent phishing attacks. The present disclosure suggests receiving incoming emails from a user’s computing device. Thereafter, the received emails are analyzed to detect suspicious emails, and the detected suspicious emails are sent to a link activation system. Upon receiving suspicious emails, the link activation system disables all embedded links in the suspicious email and generates a user interface. The user interface allows users to review information related to disabled links. Subsequently, the user may choose to activate or deactivate the links based …
Enhanced Wearable Devices For Visually Impaired Users, Olga Vladimirovna Barykina-Tassa, Pradip Sairam Pichumani
Enhanced Wearable Devices For Visually Impaired Users, Olga Vladimirovna Barykina-Tassa, Pradip Sairam Pichumani
Defensive Publications Series
Systems and methods are disclosed for enhanced wearable devices for visually impaired users. A system of enhanced wearable devices may comprise a first plurality of wearable devices located on a user, and a second plurality of wearable devices located on a service animal. The first plurality of wearable devices may include at least a pair of smart glasses, such as augmented reality or virtual reality glasses, and one or more sensors, such as an accelerometer, a gyroscope, an image sensor, an acoustics sensor, GPS sensors, or the like. Each wearable device of the second plurality of wearable devices may be …
Overextension Hinge Systems For Glasses Devices, Michael Webb, Jack Zhang, Andriy Pletenetskyy
Overextension Hinge Systems For Glasses Devices, Michael Webb, Jack Zhang, Andriy Pletenetskyy
Defensive Publications Series
Systems, methods, and devices, and computer program products are provided for hinge modules for smart glasses systems, and overextension hinge modules usable with one or more electronic components. In an example, an overextension hinge module may include a hinge base, a paddle connected to the hinge base via a travel pin, and a preloaded spring maintaining a base position between the hinge base and the paddle, allowing movement of the hinge base relative to the paddle along a predefined range. Various examples may include a primary pivot axis defining movement between a frame arm attached to a glasses frame, and …
Automatic Structured Menu Extraction From Menu Photographs, Bo Lin, Jinyang Yu, Jorge Nario, Xinru Yang
Automatic Structured Menu Extraction From Menu Photographs, Bo Lin, Jinyang Yu, Jorge Nario, Xinru Yang
Defensive Publications Series
Restaurant menu images can be utilized to automatically obtain structured data about dish names, prices, etc. However, the raw optical character recognition (OCR) output suffers from low quality and OCR techniques do not have sufficient ability to adapt to the diversity in language and design of restaurant menus. A language model can be used together with OCR to identify dish names and other content through a named entity recognition (NER) process. However, this is not scalable due to the requirement of a large, labeled dataset across languages and countries. This disclosure describes the use of a multimodal large language model …
Automatic Creation Of Map Layers Using Machine Learning, Quinn Tran, Joseph Johnson Jr
Automatic Creation Of Map Layers Using Machine Learning, Quinn Tran, Joseph Johnson Jr
Defensive Publications Series
While map layers with different types of information are available, creating such layers currently requires substantial engineering effort and time. This disclosure describes the use of machine learning based clustering techniques for automatic generation of new layers in digital maps. Layers are automatically generated and can be inspected or corrected by human moderators. The layers include visualizations of aggregate geolocated data. The visualization can utilize generative AI for custom visuals. Layer creation is thus automated and faster, increasing the types of uses for digital maps. If the user permits, the automatically generated custom layers can include commerce-related layers or can …
Predictive Link Bandwidth Contention Optimization For Segment Routing, Praveen Kumar, Navdeep Sood
Predictive Link Bandwidth Contention Optimization For Segment Routing, Praveen Kumar, Navdeep Sood
Defensive Publications Series
Segment Routing (SR) is the most widely adopted backbone routing technology. Within an SR context, techniques are presented herein that support an algorithm under which policy restoration will not lead to any bandwidth contention on target links. The presented techniques ensure that a customer’s backbone traffic will remain free of disruptions by carrying out proactive bandwidth optimization operations across each SR policy and each link in an SR network.
Hardware-In-Loop Testing Of Power Protection And Control Systems, N/A
Hardware-In-Loop Testing Of Power Protection And Control Systems, N/A
Defensive Publications Series
Protection and control (P&C) schemes provide reliable electric supply by transferring power supply across sources, changing power levels in response to load conditions, etc. Currently, the field-testing of P&C schemes is static and their performance under dynamic supply/load scenarios is not characterized. This disclosure describes a fully dynamic, mobile, hardware-in-loop simulator for real-time field-testing of P&C schemes with live field equipment. A real-time, digital simulator model is built with the component models and the topology of the P&C scheme under test. The model provides analog outputs, which are amplified and injected into the P&C scheme under test, and digital outputs …
A System To Implement A Lineage Auto Capture Plugin For Hive, Gurupraveen V. A, Uday Sawant, Jayachandra Adusumalli
A System To Implement A Lineage Auto Capture Plugin For Hive, Gurupraveen V. A, Uday Sawant, Jayachandra Adusumalli
Defensive Publications Series
The present disclosure provides a system and a method for lineage auto capture plugin for hive. The system receives a query using a user interface and hive clients, process the query using a driver, execute the query via an execution engine. Further, the system implements the functionality of hive lineage connector to perform lineage auto capture for hive. Thereby, managing data efficiently and addressing issues related to data efficiently.
Keyword Extraction And Analysis From Free-Form Text, Prerna Kakkar, Shrey Nagpal, Shubham Sharma, Souvik Paul
Keyword Extraction And Analysis From Free-Form Text, Prerna Kakkar, Shrey Nagpal, Shubham Sharma, Souvik Paul
Defensive Publications Series
Log files and reports describing outages and bugs include a number of free-form text fields that include crucial information about the outages or bugs. However, such information may often be lost because it does not have a defined structure. For such information to become useful, developers often manually label the data to find patterns. This is a labor intensive process. This disclosure describes natural language processing (NLP) techniques to automatically extract and store relevant structured information from free-form text fields. Such information, once in structured format, can be used to analyze the data and identify trends. Free-form information is recast …
Generative Ai Based Adaptive Welcome Messages For Business Chat, Ashish Duggal, Sophie Shen, Jp Suh, Frank Vanzile, Ben Welch, Eunju Tae, Bryant Nguyen, Geoff Cox
Generative Ai Based Adaptive Welcome Messages For Business Chat, Ashish Duggal, Sophie Shen, Jp Suh, Frank Vanzile, Ben Welch, Eunju Tae, Bryant Nguyen, Geoff Cox
Defensive Publications Series
Chats with a business entity typically include a welcome message that introduces the entity and sets the tone for the conversation. A well-written welcome message can improve customer satisfaction and can lead to improved outcomes such as higher sales. However, drafting customized welcome messages on an ongoing basis can be time consuming and is a creative challenge. This disclosure describes techniques to automatically generate suggestions for welcome messages using suitable artificial intelligence techniques, such as generative AI models. The generated suggestions are presented to entities such as businesses/merchants that implement online chat functionality. The suggestions are customized for each entity …
Temporal Routing In Delay Tolerant Networks, Pascal Thubert, Akram Sheriff, Frank Brockners, Patrick Wetterwald
Temporal Routing In Delay Tolerant Networks, Pascal Thubert, Akram Sheriff, Frank Brockners, Patrick Wetterwald
Defensive Publications Series
In conventional network environments, routing implies a stable world in which the vision of a next hop is consistent with the vision a forwarding node, so that a packet can progress, for example, in a greedy fashion that reduces a remaining cost at each hop, to a destination. However, in the world of mobile delay tolerant networking (DTN), nodes can move in any direction, nodes may forward packets when they meet a peer, and may move in between such actions. Thus, the relative position of nodes can change between meeting such that it can become difficult to compute a physical …
One-Time Credit Sharing With Family Members And Friends Through Credit Cards, Alok Roy, Sudip Kumar Manna
One-Time Credit Sharing With Family Members And Friends Through Credit Cards, Alok Roy, Sudip Kumar Manna
Defensive Publications Series
The present disclosure discloses a system and method for one-time credit sharing with family members and friends. The one-time credit sharing method includes receiving a credit sharing request from the first user device. The system includes notifying a second user device about the credit sharing request. The second user may approve/reject the credit sharing request of the first user. Once the second user approves the credit sharing request, the available credit limit of the first user is increased by an amount the credit sharing request was for.
Ai-Based Adaptive Load Balancer For Secure Access To Large Language Models, Assaf Namer, Hector Diaz, Jim Miller, Brandon Maltzman, Hauke Vagts
Ai-Based Adaptive Load Balancer For Secure Access To Large Language Models, Assaf Namer, Hector Diaz, Jim Miller, Brandon Maltzman, Hauke Vagts
Defensive Publications Series
Different large language models (LLMs) specialized to domains such as writing code, engaging in conversations, generating content, etc. are available. A specialized LLM can only reliably answer questions in domains over which it has been trained. Large numbers of types of specialized LLMs can make it difficult for a user, such as an application that generates LLM queries, to choose the right type of LLM. This disclosure describes techniques to automatically route query payloads between large language models specialized for different domains. The techniques utilize a vector database to semantically match an LLM to a user query. The techniques also …
Method And Device For Automated Job Failover, Devesh Gupta, Faiz Ahmed
Method And Device For Automated Job Failover, Devesh Gupta, Faiz Ahmed
Defensive Publications Series
In an embodiment of the present disclosure, there is provided a method of managing multi data cluster failure. The method comprises scheduling one or more applications on a primary cluster and a secondary cluster, monitoring a status of the one or more applications scheduled in one cluster by the other data cluster, identifying the status of the applications of one cluster by the other cluster, updating the status of the one or more applications associated with each cluster on the distributed file system (DFS) of respective cluster, and generating output reports of the scheduled one or more applications. The method …
Airbag Module Optimised For Harmful Gases, Konrad Hirschbeck
Airbag Module Optimised For Harmful Gases, Konrad Hirschbeck
Defensive Publications Series
No abstract provided.
Control Of The Electric Turbo Generator Motor Torque To Maximize System Efficiency, Konrad Hirschbeck
Control Of The Electric Turbo Generator Motor Torque To Maximize System Efficiency, Konrad Hirschbeck
Defensive Publications Series
No abstract provided.
Fuel Efficient Hybrid Powertrain With Excellent Off Road Capability And Torque Vectoring, Konrad Hirschbeck
Fuel Efficient Hybrid Powertrain With Excellent Off Road Capability And Torque Vectoring, Konrad Hirschbeck
Defensive Publications Series
No abstract provided.
A System For Decentralized Customer Service, Sahil Arora
A System For Decentralized Customer Service, Sahil Arora
Defensive Publications Series
The present disclosure provides a system and a method for resolving customer 102 complaints using a decentralized customer service. The disclosure proposes a customer service system 104 which receives complaints from the banking customer and resolves them without any involvement of an intermediary in an automated manner. In particular, the customer service system 104 may resolve the complaint using smart contracts and blockchain and stores them for further analysis or learning.
Data Collection On Vehicle Fleet For Condition Assessment Of The Infrastructure, Konrad Hirschbeck
Data Collection On Vehicle Fleet For Condition Assessment Of The Infrastructure, Konrad Hirschbeck
Defensive Publications Series
No abstract provided.