International Business Machines Corporation

Sheng Hua Bao

Zhong Su

Jian Ye Chen

William Scott Spangler

Xiaoyang Gao

Xin Yang

Richard L Martin

Brian Scott Dreher

Yanyan Han

Li Ni Zhang

Ramani Ranjan Routray

Xiang Sean Zhou

Rashmi Gangadharaiah

Cartic Ramakrishnan

Feng Wang

Shiwan Zhao

Wen Liu

Shi Lei Zhang

Yong Qin

Qin Shi

Meenakshi Nagarajan

Shilpi Ahuja

Pathirage D Perera

David Martinez Iraola

Jie Cui

Eitan D Farchi

Honglei Guo

Hui Su

Orna Raz

Zhili Guo

Xian Wu

Hong Lei Guo

Wei Hong Qian

Pathirage D S U Perera

Rui Wang

Rui Ma

Shuo Zhang

Wei Jiang

Ke Ke Cai

Dan G Tecuci

Su Yan

Dong Xu Duan

Hrishikesh Sathe

Pathirage Dinindu Sujan Udayanga Perera

Nakul Chakrapani

Vivek Krishnamurthy

Chengen Lu

Maja Vukovic

Donna Karen Byron

Yu Deng

Roman Vaculin

Min Li

Jim Alain Laredo

Hui Zhu

Keke Cai

Michael D Pfeifer

Davide Pasetto

Qi Hu

Changhua Sun

Marcel Zalmanovici

Jian Wei Zhang

Quan Yuan

Bruce Light Hillsberg

Chang Hua Sun

Nan Liu

Priscilla S Moraes

Daniel Thornton Pierce

Henry Haitian Chen

Hong Lin Guo

Abhinandan Kelgere Ramesh

Claire Abu-Assal

Xianying Liu

Nitish Aggarwal

Tongkai Shao

A mechanism is provided in a data processing system comprising at least one processor and a memory comprising instructions which, when executed by the at least one processor, causes the at least one processor to implement a real-time prediction engine for real-time predication of chemical properties through combining calculated, structured, and unstructured data at large scale. Offline components executing within the real-time prediction engine store a computational representation for each of a plurality of chemical structures in a unified storage. Each computational representation maps a respective chemical structure to a vector of calculated chemical structure features and properties, unstructured chemical features and properties, and structured chemical features and properties. The offline components train a computational real-time predictive model based on the computational representations. A user interface executing within the real-time prediction engine receives a request specifying one or more chemical compounds. An analytics jobs manager executing within the real-time prediction engine predicts one or more properties of the one or more chemical compounds using the computational real-time predictive model. The analytics jobs manager outputs the one or more properties of the one or more chemical compounds to the user interface.

Real-time prediction of chemical properties through combining calculated, structured and unstructured data at large scale

A computer system accesses a storage device. Contents of an object of a request are sorted. The contents of an object of a request are sorted. A hash key is generated to access information in the storage device based on the sorted contents of the object, wherein objects with non-critical differences are mapped to the same hash key. The information in the storage device is accessed based on the generated hash key to produce a response to the request. Embodiments of the present invention further include a method and program product for accessing a storage device in substantially the same manner described above.

Accessing cached data using hash keys

A mechanism is provided in a data processing system having a processor and a memory. The memory comprises instructions which are executed by the processor to cause the processor to implement a training system for finding an optimal surface for hierarchical classification task on an ontology. The training system receives a training data set and a hierarchical classification ontology data structure. The training system generates a neural network architecture based on the training data set and the hierarchical classification ontology data structure. The neural network architecture comprises an indicative layer, a parent tier (PT) output and a lower leaf tier (LLT) output. The training system trains the neural network architecture to classify the training data set to leaf nodes at the LLT output and parent nodes at the PT output. The indicative layer in the neural network architecture determines a surface that passes through each path from a root to a leaf node in the hierarchical ontology data structure. The training system trains a classifier model for a cognitive system using the surface and the training data set.

Optimizing hierarchical classification with adaptive node collapses

A computer system processes a group of inputs. A group of entities that is input for processing is intercepted. The intercepted group is expanded into individual entities. Each of the individual entities is processed to produce results for each individual entity. The results for each individual entity are intercepted and merged to produce results for the group of entities. Embodiments of the present invention further include a method and program product for processing a group of inputs in substantially the same manner described above.

Processing entity groups to generate analytics

A computer system trains a machine learning model. A vector representation is generated for each document in a collection of documents. The documents are clustered based on the vector representations of the documents to produce a plurality of clusters. A training set is produced by selecting one or more documents from each cluster, wherein the selected documents represent a sample of the collection of documents to train the machine learning model. The machine learning model is trained by applying the training set to the machine learning model. Embodiments of the present invention further include a method and program product for training a machine learning model in substantially the same manner described above.

Generating training sets to train machine learning models

A mechanism is provided in a data processing system for training a computer implemented model. The mechanism determines an operation for which the computer implemented model is to be trained. The mechanism performs a statistical analysis of an enterprise dataset for an enterprise to generate one or more statistical distributions of cases and features correlated with the operation for which the computer implemented model is to be trained. The mechanism selects a subset of cases in the enterprise dataset for annotation based on the one or more statistical distributions of cases and features. The mechanism annotates the selected subset of cases to generate an annotated training dataset. The mechanism trains the computer implemented model, using the annotated training dataset, to perform the operation.

Cognitive data preparation for deep learning model training

Mechanisms are provided to implement an annotation mechanism allows users to annotate documents with annotations for processing by a cognitive medical system. The annotation mechanism receives, via a user interface, a user selection of an electronic document for annotation, and determines one or more domains associated with the selected electronic document from an analysis of metadata associated with the selected electronic document. The annotation mechanism retrieves a predefined set of annotations associated with each determined domain, and presents the predefined set of annotations as user selectable elements. The annotation mechanism receives, via the user interface, a selection of one or more annotations in the predefined set of annotations to be associated with the selected portion of the selected electronic document, and generates annotation metadata associating the selected portion using the selected one or more annotations. The annotation mechanism then generates an annotated electronic document using the annotated metadata.

Annotating documents for processing by cognitive systems

From metadata corresponding to a narrative text, a first encoding is constructed, the first encoding comprising a standardized text string, the first encoding formed according to an encoding scheme. A specified portion of the standardized text string of the first encoding is marked as an anchor term. A correspondence between the first encoding and a second encoding is tested using the encoding scheme and a Natural Language Processing engine, responsive to finding the anchor term within the narrative text. The second encoding corresponds to a text window. The text window comprises a portion of the narrative text comprising an instance of the anchor term and a word within a predetermined distance from the instance. Responsive to the second encoding being identical to the first encoding, the narrative text is annotated, the annotating creating new data linking the narrative text with the second encoding.

Encoding-assisted annotation of narrative text

According to embodiments of the present invention, methods, systems and computer readable media are provided, in a cognitive data processing system, for implementing a predictive analytics system that utilizes entity and non-entity information. A collection of content is processed to extract defined entities pertaining to one or more domains. Semantic relationships are determined between objects within the collection of content, wherein the objects include undefined entities. The defined entities and objects are resolved based on entity definitions and the semantic relationships to determine defined entities and undefined entities for a resulting data set. The resulting data set is processed to identify one or more relationships between a defined entity and an undefined entity.

Combining semantic relationship information with entities and non-entities for predictive analytics in a cognitive system

A method, a system, and a computer program product are provided. A machine learning model is generated to process adverse event information and produce multiple corresponding medical codes associated with the adverse event information, wherein the multiple medical codes are semantically and hierarchically related in a medical taxonomy. The machine learning model includes multiple parallel output layers, each of which is associated with a corresponding medical code. The machine learning model is trained with training data elements, each of which includes adverse event information mapped to respective multiple medical codes, wherein results from each of the output layers adjusts the machine learning model. After completing the training, information pertaining to an adverse event is applied to the machine learning model to determine the corresponding multiple medical codes within the medical taxonomy.

Architecture for machine learning model to leverage hierarchical semantics between medical concepts in dictionaries

According to embodiments of the present invention, similarity metrics or measures of similarity may be combined using an adaptive weighting scheme. A subset of entities from a first set of entities that have a known relationship is randomly selected. The subset is combined with a second set of entities that have an unknown relationship to each other and/or to the first set of entities. At least two different measures of similarity (similarity metrics) between the first set and the combined second set (including the subset) is determined for each entity in the second set. For each entity in the second set, the at least two different measures of similarity are compared, and a weight is assigned adaptively to each measure of similarity based on the magnitude of the measure of similarity. The weighted measures of similarity are combined to determine an aggregate adaptively weighted similarity score for each entity.

Adaptive weighting of similarity metrics for predictive analytics of a cognitive system

Methods, systems, and computer program products are provided for processing a request regarding relationships among instances of entities. A graphical representation of instances of entities is generated and includes one or more source nodes, each representing an instance of an input entity of a request, and one or more related nodes, each representing an instance of a second entity related to one or more corresponding instances of the input entity and associated with a corresponding confidence score for the relationship. Each of the one or more related nodes associated with a confidence score satisfying a threshold is identified. One or more supplemental nodes are added to the graphical representation, each of which represents a corresponding instance of a third entity with a relationship to a corresponding instance of the second entity. The graphical representation is traversed to identify relationships between instances of entities and produce results for the request.

Searching for and determining relationships among entities

Methods, systems and computer readable media are provided for accessing data utilizing a multi-level table comprising generating a plurality of levels of the multi-level table, wherein a first level of the multi-level table includes a hyper-table with a plurality of hyper-cells each hyper-cell including information for a group of cells from an initial base table, wherein intermediate levels of the multi-level table each include a plurality of hyper-tables comprising hyper-cells with each hyper-table linked to and providing information for a corresponding hyper-cell of a hyper-table of a prior level, and wherein a plurality of tables of a terminal level includes information from cells of the initial base table with each table linked to and providing information for a corresponding hyper-cell. Data from the multi-level table is accessed by traversing links between the hyper-tables of the plurality of levels to access data within the tables of the terminal level.

Accessing data in a multi-level display for large data sets

Growing community of inventors

San Jose, CA, United States of America

Sheng Hua Bao