Microsoft Technology Licensing, LLC

Christopher Clayton McConnell

Steven Wayne Ickman

Thomas Matthew Laird-McConnell

William L Portnoy

David Raskino

Lili Cheng

Gregory Scott Melander

Shahin Shayandeh

Michael John Novak

David Dawson

Carlos Castro

John Anthony Taylor

Robert Lovejoy Goodwin

Siddhartha Cingh Arora

Weipeng Liu

Matthew Graham Dyor

Christopher Lee Mullins

William Ramsey

Tyler Kien Beam

Scott A Thurlow

Paul Andrew Steckler

Daniel James Driscoll

Tom Laird-McConnell

James H Lewallen

Carlos Brito

Pär Jonas Barklund

Dale Rogerson

Thomas Dale White

Wenlong Dong

Paolo Marcucci

Scott Jeff Gellock

Shannon Scott Hysom

Ilia Sacson

J Shane McRoberts

Jonas Barklund

Igor Rondel

Amit Rai

Kipper York

Scott Whalley

Zhanliang (John) Chen

Tobin D Baker

Interchangeable socket driver

Making an information retrieval process public, so that it can be followed by others, allows capturing of an interest graph that allows people to learn more about shared interests with other people. This also allows items of interest to a trusted resource (such as an expert) to be identified. These items can then be brought to the attention of other users that share the same interest as the expert. In addition, by keeping track of what particular content a user has already seen, the system can bring items of interest to the user's attention, where the user has not yet seen those items.

Identifying unseen content of interest

A method for utilizing shared memory spaces for data lifecycle management in a conversational AI system includes receiving one or more storage requests directing conversational data to multiple nodes of a conversational data object and storing the conversational data that is directed to each different one of the multiple nodes within a different one of multiple discrete memory spaces. Each of the multiple discrete memory spaces is governed by a scope policy defining a lifetime for data stored within the memory space. The method further provides for erasing data of each node of the multiple different nodes of the conversational data object according to the lifetime defined by the scope policy governing the memory space storing the node.

Shared memory space management for conversational artificial intelligence system

Technologies are described herein for denormalizing data instances. Schemas for data instances are embedded with annotations indicating how the denormalization is to be performed. Based on the annotations, one or more sub per object indexes ('sub POIs') can be generated for each data instance and stored. The sub POIs can include a target sub POI containing data from the data instance, and at least one source sub POI containing data from another data instance, if the data instance depends on the other data instance. Data instance updates can be performed by identifying sub POIs that are related to the updated data instance in storage, and updating the related sub POIs according to the update to the data instance. The sub POIs can be sent to an indexing engine to generate an index for a search engine to facilitate searches on the data instances.

Efficient denormalization of data instances

Techniques are provided for building a dialog-state specific contextual language understanding system using subsumption logic. Information establishing conversational rules identifying the conversational dialog is received to present in respective dialog states. Each rule has a Boolean trigger expression of predicates for testing the conversational state together with logical connectives to identify when the rule is applicable. Subsumption logic is used to arrange the rules into a directed acyclic graph (DAG) where more specific rules are preferred to more general rules. During a conversation, the DAG is used to filter the triggered rules to only the most specific triggered rules from which a rule to run is selected. This structure makes it easier to build conversational systems because rules can be added or removed without having to change or reason over other rules. The rules also act as a constraint to help machine learned selection systems converge with less data.

Utilizing rule specificity in conversational AI

A computer-executed conversational system utilizing adaptive dialogs to provide dynamic modification of a conversation structure. The adaptive dialogs facilitate an adaptive and dynamic way to model conversations. A conversational rule may be associated with a trigger received at an adaptive dialog. The conversational rule may populate a plan with steps to provide the conversational structure of the conversational system. Modifying triggers may be received that allow conversational rules to amend the plan including adding steps, removing steps, or modifying steps from the plan. The adaptive dialogs and/or components thereof may be provided as constructible elements in a platform for building a bot such as a developer Software Development Kit (SDK) or the like. In turn, sophisticated conversational modeling may be conveniently facilitated to provide features such as a dialog dispatcher, the ability to handle interruptions, and to build a pluggable, extensible conversational system.

Conversational AI using adaptive dialogs

Conversational event modeling for determining how to process an input in a dialog stack of a computer-executed conversational system. Receipt of an input at an active dialog in a dialog stack may result in generation of an event corresponding to the input. The event may be propagated through the dialog stack. Dialogs in the dialog stack may generate a confidence value in relation to processing the input, and selection of the dialog in the dialog stack for use in processing the input may be based at least in part on the confidence values. In turn, the conversational system may facilitate improved handling of ambiguous or unrelated inputs at dialogs by propagating the event associated with such an input through the dialog stack. The event creation and processing functions may be provided as parameters for dialogs in a modular dialog creation system.

Conversational event modeling

Methods, systems, and computer program products for dispatching an incoming event are described. A scoring stack is accessed, the scoring stack comprising an identity of one or more tasks, each task corresponding to one or more scorable functions. A scorable tree is generated based on the one or more scorable functions corresponding to the one or more tasks of the scoring stack and the incoming event is processed using the scorable tree to generate one or more scores. One or more actions identified in the scorable tree are performed based on the one or more scores.

Dynamic event processing

In an example, a method includes receiving a data package, the data package including a contact identification and a message payload; in response to determining that a key for the contact identification does not exist in a contact table: generating a key for the contact identification; and updating the contact table with the generated key to indicate an association between the key and the contact identification; calculating a hashed identification based on the generated key and contact identification; and associating the message payload with the hashed identification in a message table.

Keyed hash contact table

Methods, systems, and computer program products for dispatching an incoming event are described. A scorable tree is obtained and the incoming event is submitted to a root node of the scorable tree for processing. The incoming event is augmented with scoring data generated by the root node to create a scoring event and the scoring event is iteratively passed from each parent node of the scorable tree to each corresponding child node of the scorable tree for further processing until each leaf node of each branch of the scorable tree is reached, wherein each child node augments an incoming scoring event with additional scoring data and each leaf node generates an intermediate scoring event. Each intermediate scoring result is iteratively passed from each child node of the scorable tree to each corresponding parent node, where each child node produces a scoring result, the scoring result comprising an identification of a path to one or more selected child nodes and final scoring information based on data received from one or more corresponding child nodes.

Event processing using a scorable tree

Various embodiments provide data structures and algorithms for efficiently computing a time ordered stream that is the combination of a time ordered substream for individual persons of interest. In addition, various embodiments provide an approach for identifying how many changes of interest have happened to a particular stream since it was last viewed.

Fast streams and substreams

A topic feed is generated for publication or distribution to a set of followers or recipients who have elected to receive the topic feed from the user or based on the subject matter content of a topic input in the topic feed. The topic feed is distributed and includes the topic input. Experts in subject matter can be identified based on the user's topic input and/or text in the topic feed that the followers interact with.

Identifying and recommending experts using shared posts and interactions

Growing community of inventors

Redmond, WA, United States of America

Christopher Clayton McConnell