International Business Machines Corporation

John J Thomas

Barry Edward Willner

Avijit Chatterjee

Rajiv V Joshi

Swaminathan Balasubramanian

Aleksandr Evgenyevich Petrov

Patrick Joseph O'Sullivan

Howard Neil Anglin

Terri A Menendez

Wanting Wang

Maxime Allard

Vinay R Dandin

Edith H Stern

Stephen Arthur Boxwell

Robert Cameron Weir

Kristen Maria Summers

Andrew J Bradfield

Ricardo Balduino

Justo M Guizan

An embodiment includes tokenizing an input passage into an n-gram sequence of tokens. The embodiment also includes evaluating the input passage using a trained classification model that generates an output indicative of a classification of the input passage. The embodiment also includes generating a first token vector for a first token of the sequence of tokens and projecting the first token vector to a higher dimensional space, resulting in a first projected token vector. The embodiment also includes generating a first similarity score for the first projected token vector based on comparisons of the first projected token vector to each of a plurality of class representations. The embodiment also includes generating a ranked list of the tokens, wherein the generating of the ranked list includes ranking the first token among others of the tokens based on the first similarity score.

Explainable passage classification

In one embodiment, a system includes at least one processor and logic configured to receive a request to process data stored in one or more index records in a keyed index of a database, determine an active key range for the one or more index records, and group the one or more index records into a number of groups for parallel processing thereof. In another embodiment, a method for utilizing a keyed index includes receiving a request to process data stored in index records in a keyed index of a database, determining an active key range for the index records, grouping the index records into a number of groups without prior knowledge of a key distribution for dividing the one or more index records into key ranges for parallel processing thereof; and processing the groups of index records in parallel.

Parallel processing of a keyed index file system

A question-and-answer system directed to a specific domain optimally utilizes reference documents that are semantically complete for that domain. Semantic completeness of a document is assessed using quality control questions (provided by subject matter experts) applied to the Q&A system followed by analysis of the proposed answers. That analysis is carried out using a cogency module having a feedforward neural network which receives metadata features of the document such as document ownership, document priority, and document type. A domain-optimized corpus for the Q&A system is built by so assessing multiple documents in a document collection, and adding each reference document that is reported as being semantically complete to the domain-optimized corpus. Thereafter, the deep learning question-and-answer system can receive a natural language query from a user, find a responsive answer in the documents while applying the domain-optimized corpus, and provide that answer to the user.

Providing semantic completeness assessment with minimal domain-specific data

A method of training a neural network includes receiving a text corpus containing a labeled portion and an unlabeled portion, extracting local n-gram features and a sequence of the local n-gram features from the text corpus, processing the text corpus, using convolutional layers, according to the local n-gram features to determine capsule parameters of capsules configured to preserve the sequence of the local n-gram features, performing a forward-oriented dynamic routing between the capsules using the capsule parameters to extract global characteristics of the text corpus, and processing the text corpus according to the global characteristics using a long short-term memory layer to extract global sequential text dependencies from the text corpus, wherein parameters of the neural network are updated according to the local n-gram features, the capsule parameters, global characteristics, and global sequential text dependencies.

Classification of sparsely labeled text documents while preserving semantics

Method and apparatus for generating profiles using machine learning and influencing online interactions are provided. The methods include generating a user profile specifying a plurality of attribute values for a plurality of principle attributes, by processing a corpus of electronic documents using a first trained machine learning model. In an embodiment, the method further comprises generating a provider profile specifying a plurality of attribute values for the plurality of principle attributes, for each of a plurality of providers, by processing a respective corpus of electronic documents associated with each respective provider using a second trained machine learning model. A plurality of match coefficients based on comparing the user profile and the plurality of provider profiles are determined. Finally, one or more online interactions between the user and the target provider are influenced based on the determined match coefficients.

Building and matching electronic standards profiles using machine learning

Method and apparatus for generating profiles using machine learning and influencing online interactions are provided. The methods include receiving, from a first user of a plurality of users, a first set of electronic documents, where each electronic document in the first set of electronic documents corresponds to a respective user in the plurality of users. The methods also include identifying a plurality of user profiles, where each of the plurality of user profiles was generated by processing a corpus of electronic documents associated with each respective user using a first trained machine learning model. The methods include determining a plurality of match coefficients, based on comparing a plurality of user profiles associated with each respective user in the plurality of users, filtering the first set of electronic documents based on the plurality of match coefficients, and providing the filtered first set of electronic documents to the first user.

Building and matching electronic user profiles using machine learning

Disclosed aspects relate to managing a set of offers using a dialogue. An adaptive profile may be received with respect to a client. The adaptive profile may indicate a set of client profile data, a set of client event data, and a set of client context data. A dialogue may be established with the client based on the adaptive profile. A set of offers may be resolved by an offer management engine based on the dialogue. The set of offers may be presented to the client.

Managing a set of offers using a dialogue

In a general purpose computer, a method of extracting snippets includes receiving textual content and a plurality of available topics, dividing the textual content into a plurality of snippets, converting each of the snippets to a vector, determining a distance between coadjacent snippets of the plurality of snippets in the textual content, determining an update to the plurality of snippets by merging each of the pairs of coadjacent snippets having a respective distance less than a second threshold, wherein an updated plurality of snippets includes merged snippets, generating a plurality of clusters from the updated plurality of snippets, each cluster associated with one topic selected from the plurality of available topics, and generating, for each of the snippets of the updated plurality of snippets, an affinity score for each of the clusters, each affinity score measuring an assignment strength of a given snippet to a given cluster, and a dominant topic among the at least one identified topic.

Incremental discovery of salient topics during customer interaction

In one embodiment, a system includes at least one processor and logic configured to receive a request to process data stored in one or more index records in a keyed index of a database, determine an active key range for the one or more index records, and group the one or more index records into a number of groups for parallel processing thereof. In another embodiment, a method for utilizing a keyed index includes receiving a request to process data stored in index records in a keyed index of a database, determining an active key range for the index records, wherein the active key range is determined based on any of: a type of the index records, a horizontal pointer to a next index record, and a low key value for each index record, and grouping the index records into a number of groups for parallel processing thereof.

A resource reservoir is provided. The resource reservoir is a pool of resources that are allocated to and shared between a plurality of reserve virtual machines. A reserve virtual machine of the plurality of reserve virtual machines corresponds to a primary virtual machine from a plurality of primary virtual machines. Each reserve virtual machine of the plurality of reserve virtual machines fulfills a resource requirement from the resource reservoir. The sum of the resources requirements of the plurality of reserve virtual machines is greater than a sum of total resources within the resource reservoir. In response to determining a decrease in performance of the primary virtual machine, at least a portion of workload of the primary virtual machine is applied to the corresponding reserve virtual machine. The resources used by the reserve virtual are indicated as unavailable to the remaining reserve virtual machines from the resource reservoir.

Optimization of virtual machines

Techniques for managing a plurality of threads on a multi-threading processing core. Embodiments provide an instruction count threshold condition that determines how many countable instructions of a thread the multi-threading processing core will execute before context switching to another one of the plurality of threads. A first plurality of instructions for a first one of the plurality of threads is processed on the multi-threading processing core. Embodiments determine, for each of the first plurality of instructions, whether the instruction is a countable instruction, wherein at least one of the first plurality of instructions is not a countable instruction. A count of the countable instructions is maintained. Upon determining that the instruction count threshold condition is satisfied, based on the maintained count, embodiments context switch the multi-threading processing core to process a second plurality of instructions for a second one of the plurality of threads.

System management and instruction counting

Virtual Machine (VM) images in a virtualized environment are updated through the use of patches. A virtualization data processing system includes a hypervisor that manages a VM image. The hypervisor is configured to retrieve a patch for an instance of the VM image from a secure site. The hypervisor blocks all other network access to the VM image. The hypervisor is configured to apply the patch to the instance of the VM image and unblock all network access to the VM image

Growing community of inventors

Fishkill, NY, United States of America

John J Thomas