Knowledge discovery agent system

• US 10,049,162 B2
• Filed: 11/06/2015
• Issued: 08/14/2018
• Est. Priority Date: 05/22/2002
• Status: Expired due to Term

• Current Assignee: Digital Reasoning Systems, Inc (Smarsh Incorporated)
• Original Assignee: Digital Reasoning Systems, Inc (Smarsh Incorporated)
• Inventors: Estes, Timothy W.
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): LAMARDO, VIKER ALEJANDRO
• Application Number:
  US14/935,183
• Publication Number:
  US 20160140236A1
• Time in Patent Office: 1,012 Days
• Field of Search: 706 12
• CPC Class Codes: G06F 16/2425 Iterative querying; Query formulation based on the results of a preceding query
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/907 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: H04N 21/251 Learning process for intelligent management, e.g. learning user preferences for recommending movies details of learning user preferences for the retrieval of video data in a video database G06F16/739; computer systems using learning methods G06N3/08
• CPC Class Codes: H04N 21/466 Learning process for intelligent management, e.g. learning user preferences for recommending movies services using the results of monitoring in broadcast systems H04H60/61
• CPC Class Codes: H04N 21/4662 characterized by learning algorithms
• CPC Class Codes: H04N 21/4668 for recommending content, e.g. movies
• CPC Class Codes: H04N 21/4826 using recommendation lists, e.g. of programs or channels sorted out according to their score

Abstract

A system and method for processing information in unstructured or structured form, comprising a computer running in a distributed network with one or more data agents. Associations of natural language artifacts may be learned from natural language artifacts in unstructured data sources, and semantic and syntactic relationships may be learned in structured data sources, using grouping based on a criteria of shared features that are dynamically determined without the use of a priori classifications, by employing conditional probability constraints.

Claims

1. A system, comprising:

at least one processor in a distributed computer network;

at least one data storage device in the distributed computer network, in communication with the at least one processor, and configured to store computer-executable instructions and program data; and

at least one data agent in the distributed computer network and in communication with the at least one processor and configured to perform specific functions in response to instructions from the at least one processor;

wherein the at least one data agent is configured to perform functions that include creating a directed or undirected graph representation of features from unstructured data of at least one unstructured data source, wherein the unstructured data source comprises text data from a text corpus;

wherein the graph representation of the features is configured for use with at least one machine learning function, the at least one machine learning function including learning semantic associations between words using calculations of similarity of the words based on usage of the words in context over time, wherein the contexts correspond to semantic units and constituents of the semantic units correspond to elements;

wherein the similarity of words determined from extracted semantic units that re-occur across the text corpus are used for forming a compressed knowledge representation of a discovered pattern of relationships between semantic units and respective elements; and

an organic software agent, of the at least one data agent, having a belief state formed from the compressed knowledge representation and that modifies its own source code for at least one of decisions and execution of plans by the organic software agent, wherein modifying the source code comprises at least one of autonomous dynamic creation and manipulation of executable code by the organic software agent that, when executed, causes the at least one data agent to perform personalized agent services for a user in response to a user interaction with the system.

2. The system of claim 1, wherein the relative position of the words in the context for each of the calculations of similarity is used as one of the features.

3. The system of claim 2, wherein the similarity calculations are represented at least in part as vectors and distances between the vectors.

4. The system of claim 3, wherein hierarchies of similarity are calculated using grouping functions or relative distances such that one symbol is shown to be semantically more general or senior in a hierarchy than another.

5. The system of claim 4, wherein the semantic seniority is represented as a graph or tree derived without an a priori structure.

6. The system of claim 1, configured to support tasks in a computer-executable intelligent assistant or intelligent agent, wherein the tasks include prioritizing alerts or risks associated with unstructured data, such that user interaction with the system improves the ranking based on analysis of the user interaction, actively or passively.

7. The system of claim 1, wherein the extraction of semantic units comprises extracting at least one of:

one or more of subject, predicate, object, and modifier structures; and

tuples.

8. A computer-implemented method, comprising:

performing, by at least one data agent in a distributed computer network, functions that include creating a directed or undirected graph representation of features from unstructured data of at least one unstructured data source, wherein the unstructured data source comprises text data from a text corpus;

wherein the graph representation of the features is configured for use with at least one machine learning function, the at least one machine learning function including learning semantic associations between words using calculations of similarity of the words based on usage of the words in context over time, wherein the contexts correspond to semantic units and constituents of the semantic units correspond to elements;

wherein the similarity of words determined from extracted semantic units that re-occur across the text corpus are used for forming a compressed knowledge representation of a discovered pattern of relationships between semantic units and respective elements; and

wherein the compressed knowledge representation forms a belief state of an organic software agent of the at least one data agent;

modifying, by the organic software agent, its own source code for at least one of decisions and execution of plans by the organic software agent, wherein modifying the source code comprises at least one of autonomous dynamic creation and manipulation of executable code by the organic software agent that, when executed, causes the at least one data agent to perform personalized agent services for a user in response to a user interaction with the system.

9. The computer-implemented method of claim 8, wherein the extraction of semantic units comprises extracting at least one of:

one or more of subject, predicate, object, and modifier structures; and

tuples.

10. A non-transitory computer-readable medium storing instructions which, when executed by at least one processor in a distributed computer network, cause at least one computer to perform functions that include:

performing, by at least one data agent in the distributed computer network, functions that include creating a directed or undirected graph representation of features from unstructured data of at least one unstructured data source, wherein the unstructured data source comprises text data from a text corpus;

wherein the graph representation of the features is configured for use with at least one machine learning function, the at least one machine learning function including learning semantic associations between words using calculations of similarity of the words based on usage of the words in context over time, wherein the contexts correspond to semantic units and constituents of the semantic units correspond to elements;

wherein the similarity of words determined from extracted semantic units that re-occur across the text corpus are used for forming a compressed knowledge representation of a discovered pattern of relationships between semantic units and respective elements; and

wherein the compressed knowledge representation forms a belief state of an organic software agent, of the at least one data agent, that modifies its own source code for at least one of decisions and execution of plans by the organic software agent, wherein modifying the source code comprises at least one of dynamic creation and manipulation of executable code by the organic software agent that, when executed, causes the at least one data agent to perform personalized agent services for a user in response to a user interaction with the system.

11. The non-transitory computer-readable medium of claim 10, wherein the extraction of semantic units comprises extracting at least one of:

one or more of subject, predicate, object, and modifier structures; and

tuples.