Digital Commons Network^™

Context Based Automatic Email Responses, Victor Carbune, Daniel Keysers, Thomas Deselaers

Defensive Publications Series

When users are not available to respond to emails, they typically set-up automatic responses to incoming emails using standard messages (e.g., a standard out of office reply) or simple rule based systems. However, such standard responses may not provide adequate context and detail to the recipient of the automatic reply. Techniques described here employ machine learning and heuristic models to dynamically generate contextual responses to emails based on the communication history between the corresponding email sender and recipient(s).

Noise And Howling Mitigation In Audio/Video Conferencing, Christian Schuldt, Robin Lundberg

Defensive Publications Series

Loudspeaker-to-microphone feedback causes echoes in audio/video conferences. A microphone typically has a background-noise level estimator, which provides noise-level estimates for the purpose of setting parameters within acoustic echo cancelers (AEC), noise reducers, comfort noise generators (CNG), etc. Output from a noisy loudspeaker may be picked up by the microphone, adding to background noise already present at the microphone. Loudspeaker noise, when picked up by the microphone, causes a misestimation of background noise, which in turn leads to incorrect parameter setting. This can lead to divergence of adaptive AEC circuitry, audible echoes, howling, etc.

Selective Activation Of Call Ads Based On Advertiser Availability, Varun Soundararajan, Manas Mittal

Defensive Publications Series

Some formats of online advertisements permit users to call the advertiser directly by clicking on the ad. Such ads can result in customer dissatisfaction when the advertiser is busy or unavailable to answer calls placed through the ad. The techniques described herein employ selective activation of call ads to improve the likelihood that calls made by customers to advertisers are answered. Contextual factors such as location, time, activity that an advertiser is engaged in, etc. for which an advertiser provides consent are utilized to predict advertiser availability. The techniques improve customer satisfaction and can reduce wasteful ad expenditure.

Drawing Flowcharts On Touch-Enabled Devices, Wei Li, Erika Lu

Defensive Publications Series

Users benefit from being able to draw flowcharts directly on touch-enabled devices without the hassle of dropdown menus for shape selection or operations to adjust shapes and sizes of drawn objects. This disclosure automatically creates shape elements that match a user’s intention based on the drawing positions, shapes, sizes of strokes, etc. Beautified versions of a user’s drawing strokes are generated and used to replace the strokes. Using the techniques of this disclosure, the touch device intelligently distinguishes between drawing positions, shapes, and handwriting. The techniques distinguish shape from handwriting text and eliminates interruptions to user workflow due to the …

Advanced Dialer Based On Target Contact's Activity, Gaurav Saini, Nishesh Mehra, Sudhanshu Rawal

Defensive Publications Series

An advanced dialer system prioritizes modes of communication based on a target contact’s activity. The system, upon authorization by the user or target contact to access the user’s contacts, receives a selection of a target contact and checks the target contact’s activity status on all possible communication platforms. These communication platforms may include messaging applications using internet connectivity, traditional voice calling applications, video chat applications, etc. The system may further check the target contact’s network connectivity and feasibility of connection. The advanced dialer system calculates feasibility score for different communication platforms to reach the target contact. The system also ranks …

Conversational Bots For Advertisements, Dario Rapisardi, Amit Fatnani, Hemamalini Manickavasagam

Defensive Publications Series

This document describes a chat bot interface that provides answers to user-submitted questions within a chat interface of an advertisement on a web page. The chat bot interface can initially provide answers to some questions based on a data store, while also including live agents to answer questions for which the chat bot interface cannot identify an answer from the data store. The answers provided by the live agents are used to update the chat bot interface and the data store such that when subsequent questions are received within the chat bot interface of the advertisement, the chat bot interface …

Headset With Haptic Indicator, Christian Schuldt, Fredric Lindstrom

Defensive Publications Series

Call capabilities, e.g., using WiFi/LTE, are now available in devices other than phones. For example, a headset can be built with wireless (WiFi/LTE) communication abilities. Such a headset operates in standalone fashion, e.g., without control or short-range communication (e.g., Bluetooth) with a device such as a smartphone or tablet. The headset can be operated in a hands-free manner and can present advantages over the headset-and-smartphone combination. However, a standalone headset operates without the visual cues that are available on the screen of a device such as a phone or tablet. This disclosure describes using a haptic mechanism to convey headset …

Automatic Event Scheduling Based On Weather Conditions, Sara Smollett, Dina Levitan

Defensive Publications Series

Users prefer to engage in certain activities, e.g., swimming, surfing, etc. when the weather conditions are suitable. This disclosure describes techniques to automatically schedule events on a calendar based on activity information and weather information. For time periods when a user is available, weather information is compared with known user preferences and activity-specific weather requirements to schedule activities requested by the user. The techniques can be implemented as part of a calendar application and/or a virtual assistant application.

Prediction Of Incipient Speech In Audio/Video Conferences, Oliver Walter, Fredric Lindstrom, Ivo Creusen

Defensive Publications Series

In conference calls audio/video streams from different sources are received with varying latency. Jitter buffers are used at each receiver to align packets in time. A jitter buffer is long enough to ensure that audio is available to send to the output device, accounting for the variation in packet delay. The jitter buffer is constrained by latency requirements to be short. Jitter buffers can however still delay signals by hundreds of milliseconds, thus reducing audio conferencing quality, e.g., by causing participants to speak simultaneously.

Techniques of this disclosure provide participants with a look-ahead indication of which participant is about to …

Off-Lined Content Temporal Relevance Filtering System, Justin Lewis, Richard Rapp

Defensive Publications Series

The technology disclosed herein relates to a predictive caching mechanism that automatically offlines content based on the temporal relevance of content from a content provided and the frequency in which a user accesses the content provider. For example, the caching mechanism may deprioritize (e.g., filter out) ephemeral content for users that are unlikely to view the content before it expires and prioritize episodic content (e.g., episode in video series) that is not ephemeral.

Intelligent Assistant For Music Practice, Marcos Calvo, Philip Koonce

Defensive Publications Series

Generally, the present disclosure is directed to techniques to automatically provide feedback and suggestions to musicians. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to provide real-time feedback to musicians based on audio and/or video of the musician playing music. The techniques of this disclosure use various input features, e.g., the musician’s practice piece; references from a database of musical scores, data from different sensors, e.g., microphones, cameras, etc. to analyze the musician’s playing and provide real-time feedback or suggestions for corrections to be made, …

Audi Day Makeup Look (Adml), Daniel Hoppe

Defensive Publications Series

AUDI DAY MAKEUP LOOK (ADML)

Platform For Location And Time-Based Content Sharing, Alexander Haubold

Defensive Publications Series

Private collections of images and other media can be of interest to individuals and groups beyond the owner of those collections. This disclosure describes a platform where private collections of media can be shared, with user permission, with others who share a common interest in the location and time associated with contents of these collections. The platform is configured as a private group and functions as a social circle to share images and other media around the spatial and temporal contexts of the contents of the media. The collections can be used, for example, to archive the historical changes at …

Spam Detection For Outgoing And Incoming Calls, Alex Jacobson, Tuna Toksoz

Defensive Publications Series

System and method are disclosed to detect spam in outgoing or incoming calls on a telephone device. The system includes a server running a machine-learning algorithm coupled to a spam detection service running on the device that detects spam during an outgoing or incoming call. When the user places a call; the device listens to the conversation and extracts identifiable characteristics through acoustic characterization. These characteristics constituting a signature are received by the spam detection service to identify spam. If a spam call is detected, the user is warned through an audio cue during the call and ends the call. …

Coordinated Smart Alarm

Defensive Publications Series

The present disclosure provides systems and methods that coordinate sleep monitor devices worn to monitor sleep quality and determine the most effective sleep phase to wake a user. The systems and methods of the present disclosure can allow sleeping partners to wake at a mutually optimal time for each. In particular, existing sleep monitors solely account for the wearer of the sleep monitor and although a vibration, sounds, or other waking mechanism can be intended to only wake such wearer, often a sleeping partner is woken as the wearer of the sleep monitor wakes and moves about. Thus, if both …

Identification Of Smart Devices In Proximity To A User, Maxwell Sills, Robert Gordon

Defensive Publications Series

Smart devices continue to increase in popularity, availability, and functionality. However, to make use of a smart device, a user must first discover the smart device and determine how to interact with it. The user could search a local area using an augmented reality or identification system to highlight, determine capabilities of, and present instructions for interacting with nearby devices. Alternatively, the smart device itself could broadcast its presence, functionality, and instructions for interaction in a variety of ways. The device could send potential users an electronic message or respond to an audible command. The device could register with a …

Automatic Execution Of Authentication Actions At High Trust Levels, Thomas Price, Justin Lewis

Defensive Publications Series

Online authentication systems for payment, electronic commerce, personal communication, etc. are increasingly subject to attacks. If breached, such systems can permit fraudulent transactions. As a result, software applications (apps), e.g., e-commerce apps, banking apps, etc. deploy additional levels of identity authentication to verify user identity for sensitive transactions, e.g., money transfer from a new device, ordering goods to a new delivery address, etc. This disclosure proposes queueing actions that require additional levels of authentication such that such actions are executed at a later time once a high level of trust is obtained. High level of trust can be obtained in …

Computer Vision Ring, Nicholas Jonas, Barron Webster

Defensive Publications Series

Visually-impaired people have difficulty reading handwritten or printed documents that are not written in Braille. Totally blind people cannot read handwritten or printed documents that are not in Braille. A large fraction of written language encountered in daily life or travel still does not appear in Braille, thereby denying visually impaired and completely blind people access to such information. The present disclosure describes techniques to scan one’s surroundings using a low-profile camera, to identify objects and/or written language using an object or character recognition system, and to convey scanned information aurally to a user. Machine learning and inference models are …

Hash Comparisons To Provide Warnings Of Phishing Attacks, Pedro Gonnet Anders

Defensive Publications Series

This disclosure describes techniques to warn users when a password for a particular online domain is being submitted by a user device to a domain different than the one for which it was originally registered. The warning alerts the user of possible phishing attacks. Cryptographic hashes of user passwords along with corresponding domains are stored. When a user attempts to send data to a website, e.g., via form submission, hashes of the data are calculated and compared with the stored hashes of password. In the case of a match, a warning is presented to the user. The techniques can be …

Automatic Translation Of User-Entered Text, Matthew Sharifi, Jakob Foerster

Defensive Publications Series

This disclosure describes techniques that enable users to interact seamlessly with foreign language webpages. When a user loads a webpage in a foreign language and provides input in a language different from the webpage language, the user input is automatically translated. Based on user selection, the original user input or the translated user input, is sent to the website. The translation is performed in a manner suitable to the context and is not a literal translation of user-entered text. For example, translation is performed such that proper nouns are not translated, certain classes of entities are translated via known mappings, …

Smart Pen And Related Method For Detecting Ink Applied By A Smart Pen Onto A Writing Surface

Defensive Publications Series

The present disclosure relates to a smart pen and related method for quickly and effectively differentiating between content created by writing or marking on a writing surface with the smart pen and other content contained on the writing surface (e.g., pre-existing written or printed content). For example, the smart pen may be provided with a specialized type of ink that is configured to reflect light when illuminated with a given light source, thereby allowing the ink to be easily identified or detected within images captured via a camera of the smart pen. As a result, when images of the writing …

Determination Of Device Pose Using Text Matching From Captured Images, Tilman Reinhardt, Jeremy Pack, Allen Hutchison, Daniel Filip, Brian Brown

Defensive Publications Series

ABSTRACT:

Both server-based and client-based systems can be used for precisely detecting orientation of a user device. Existing sets of geographically referenced imagery include identifiable text, and the same text may appear within multiple images within the set where the multiple images are taken from different viewpoints and camera angles. Multiple observations of the same physical text in the world are used to triangulate measurements and create a 3-dimensional (3D) location and direction for a specific text. The 3D location for the specific text is used to create a spatially indexed 3D text database, which can be located at a …

Hands-Free User Customizable Chat Settings, Benjamin Azose

Defensive Publications Series

ABSTRACT

A system and method are disclosed that provide customizable hands-free notifications or messages, especially when the recipient is not looking at his mobile phone device (phone). The recipient’s phone may be configured to vibrate based on customized or situation-specific content. The vibration setting for the messages maybe a global setting, a content-specific global setting, a personalized setting, a temporary setting or a metadata based setting. The sender sends a message to a recipient in the format that is understood by the system. The system parses the received message, extracts the required content based on the customized setting and issues …

Assistant Word Choice Training, Google Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to adapt to a user’s word choices when providing output to the user. For a variety of reasons (e.g., regional differences, time progression, cultural differences, etc.) users may select different words and phrases for similar meanings. For instance, some users may say water fountain while other users may say bubbler. While users may understand that multiple terms refer to the same object (e.g., have the same or similar meaning), the use of a different term may feel unnatural and awkward to …

Assistant Tapering Language Patterns, Google Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to become more concise (e.g., reduce, slim down, taper, etc.) when interacting with a user over time. Depending on the extent of the user’s background knowledge, saying too much may be just as bad as saying too little. In fact, speakers whose language is perceived as either too elaborate (information in excess) or too abbreviated (insufficient info) may be perceived to be uncooperative. To avoid such negative perceptions, the assistant may provide users with more elaborate, more detailed, more precise content …

Assistant Text Normalization, Google Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to perform text normalization when converting text to speech (e.g., when synthesizing audio data for output to a user). The assistant may perform text normalization by determining how pronounce a particular set of characters (e.g., word, homonyms, number, date, acronym, abbreviation, etc.) based on the context in-which the particular set of characters is used. For instance, when performing text to speech on the text “1233 St. Andrew St.” (e.g., when reading an address aloud), the assistant may determine that the first …

Assistant Vocalized Output Of Word Combinations, Google Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to properly pronounce word junctures when converting text to speech (e.g., when synthesizing audio data for output to a user). Some example word junctures that the assistant may properly pronounce include, but are not limited to, false geminates, affricates, and other such letter/word combinations. For instance, when performing text to speech on the text “black cat”, the assistant may determine that the consecutive combination of the words “black” and “cat” is a false geminate because the last consonant phoneme in “black” …

Sender Voice Modeling For Audible Playback, Google Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to use a voice modeled on a sender of text (e.g., text message, email, instant message, webpage, etc.) when synthesizing audio data to “read” the text aloud (e.g., narrate, generate text-to-speech (TTS), etc.). For instance, when a first user receives a text message from a second user, a computing device of the first user may perform TTS to read the received text message aloud using a voice that is modeled to sound like the second user. The assistant may model the …

Custom Assistant Personas, Google. Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to have a customizable persona (e.g., attitude, voice, dialect, syntax, etc.). Assistants can have different personas when interacting with different users. For instance, an assistant may instantiate (e.g., use) one persona when interacting with a first user and another persona when interacting with a second user.

Observation-Based Form Assistant, Google Inc.

Defensive Publications Series

A virtual, intelligent, or computational assistant (e.g., also referred to simply as an “assistant”) is described that is configured to selectively aid users when completing forms (e.g., documents, surveys, orders, fields, questionnaires, set of questions, etc.) or other online activities for which aid can be offered. When a user is filling out a form, the assistant may monitor the user’s rate of progress (e.g., filling out) of the form. If the rate decreases, the assistant may offer aid in completing the form. If the user accepts the aid, the assistant may fill in some or all of the form.