Adaptable systems and methods for processing enterprise data

• US 10,157,347 B1
• Filed: 11/04/2014
• Issued: 12/18/2018
• Est. Priority Date: 11/04/2013
• Status: Active Grant

• Current Assignee: Predii, Inc.
• Original Assignee: Predii, Inc.
• Inventors: Kasturi, Tilak B., Ho, Hieu, Dalal, Aniket
• Primary Examiner(s): Holmes, Michael B
• Application Number:
  US14/533,085
• Time in Patent Office: 1,505 Days
• CPC Class Codes: G06F 16/3322 using system suggestions G06F16/3325 takes precedence
• CPC Class Codes: G06F 16/3326 using relevance feedback from the user, e.g. relevance feedback on documents, documents sets, document terms or passages
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A system, or platform, for processing enterprise data is configured to adapt to different domains and analyze data from various data sources and provide enriched results. The platform includes a data extraction and consumption module to translate domain specific data into defined abstractions, breaking it down for consumption by a feature extraction engine. A core engine, which includes a number of machine learning modules, such as a feature extraction engine, analyzes the data stream and produces data fed back to the clients via various interfaces. A learning engine incrementally and dynamically updates the training data for the machine learning by consuming and processing validation or feedback data. The platform includes a data viewer and a services layer that exposes the enriched data results. Integrated domain modeling allows the system to adapt and scale to different domains to support a wide range of enterprises.

Claims

1. A method, comprising:

building a domain model for processing data specific to an enterprise;

applying historical enterprise data and contextual data to the model to build entity relations annotated with contextual data;

wherein said method comprises sending said enterprise data and/or said contextual data to one or more of an importer module, an extraction module, a learning module, a classification module, a feedback module, a scoring module, and/or an export module;

extracting features of interest from the annotated entity relations defined in the domain model;

classifying and clustering the features to develop enterprise-specific metadata;

storing the metadata for use by the domain model; and

receiving and applying expert feedback to improve the model.

2. The method of claim 1, wherein building the domain model comprises:

adapting enterprise data by allowing expert to identify the features of interest;

auto-suggesting appropriate learning algorithm; and

interacting with the expert to learn domain knowledge.

3. The method of claim 2, further comprising:

receiving a query from an end user for information related to the enterprise;

generating a response from the metadata; and

returning the response to the end user.

4. The method of claim 1, wherein the enterprise data comprises unstructured data, said method further comprising:

associating and adapting a domain-specific meaning to the unstructured data using the extracted features; and

generating metadata associated with the enterprise data based on the domain model.

5. The method of claim 4, further comprising applying a plurality of domain-specific taxonomy terms from the domain model and a selected algorithm from an algorithm catalog to the enterprise data to extract meaning from the enterprise data.

6. The method of claim 1, wherein the domain model comprises a plurality of parameters associated with a domain corresponding to the enterprise data, and building the domain model comprises:

selecting a domain from a domain catalog;

selecting a domain problem for the selected domain, wherein the catalog includes suggested features for the domain problem;

looping through the suggested features;

using a feature extractor on source data samples to optimize the plurality of parameters;

and saving the optimized parameters in the domain model.

7. The method of claim 6, wherein building the domain model further comprises:

looping through an algorithm catalog until defined thresholds are met;

auto-suggesting an algorithm based on data volume parameters;

evaluating results from the suggested algorithm using source data samples;

allowing a domain expert to validate the results; and

saving the results to the domain model.

8. A method comprising:

identifying, by a natural language processor, that a computer-readable enterprise data represents terms of a natural human language that match one or more taxonomy terms within a defined taxonomy searchable by the natural language processor;

associating, by the natural language processor, a meaning with the enterprise data based on the terms of the natural human language represented by the enterprise data that match the one or more taxonomy terms;

generating, by the natural language processor, metadata that represents the meaning associated with the enterprise data;

providing the metadata to a data processing machine for generating content based at least in part on the metadata; and

wherein said method comprises sending said enterprise data and/or said metadata to one or more of an importer module, an extraction module, a learning module, a classification module, a feedback module, a scoring module, and/or an export module.

9. The method of claim 8, wherein the natural human language processor comprises a domain model development module for developing a domain model comprising the defined taxonomy applicable to the enterprise data.

10. A computing device comprising a processor and a computer-readable medium storing program instructions, that when executed by the processor, cause the computing device to perform a set of functions comprising:

identifying that a computer-readable vehicle enterprise data represents terms of a natural human language that match one or more taxonomy terms within a defined taxonomy searchable by the processor;

associating, by the processor, a meaning with the enterprise data based on the terms of the natural human language represented by the enterprise data that match the one or more taxonomy terms;

generating, by the processor, metadata that represents the meaning associated with the enterprise data;

providing the metadata to a data processing machine for generating content based at least in part on the metadata; and

wherein said functions comprise sending said enterprise data and/or said metadata to one or more of an importer module, an extraction module, a learning module, a classification module, a feedback module, a scoring module, and/or an export module.

11. The computing device of claim 10, wherein the natural human language processor comprises a domain model development module for developing a domain model comprising the defined taxonomy applicable to the enterprise data.