Multi-magnitudinal vectors with resolution based on source vector features

• US 10,019,261 B2
• Filed: 01/14/2014
• Issued: 07/10/2018
• Est. Priority Date: 04/13/2007
• Status: Active Grant

• Current Assignee: Optum360, LLC (UnitedHealth Group Incorporated)
• Original Assignee: A-Life Medical, LLC
• Inventors: Heinze, Daniel T., Morsch, Mark L.
• Primary Examiner(s): Smith, Paulinho E
• Application Number:
  US14/154,527
• Publication Number:
  US 20140129803A1
• Time in Patent Office: 1,638 Days
• CPC Class Codes: G06F 16/316 Indexing structures
• CPC Class Codes: G06F 16/3347 using vector based model
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 9/30036 Instructions to perform operations on packed data, e.g. vector, tile or matrix operations

Abstract

Methods, systems and computer program products for resolving multiple magnitudes assigned to a target vector are disclosed. A target vector that includes one or more target vector dimensions is received. One of the target vector dimensions is processed to determine a total number of magnitudes assigned to the processed target vector dimension. Also, a source vector that includes one or more source vector dimensions is received. The received source vector is processed to determine a total number of features associated with the source vector. When it is detected that the total number of magnitudes assigned to the processed target vector dimension exceeds one, one of the assigned magnitudes is selected based on one of the determined features associated with the source vector.

Claims

1. A non-transitory computer readable medium encoded with instructions executable by a processor of a computing system for performing vector comparison in natural language processing, the instructions comprising instructions for:

receiving a target vector and a source vector, the target vector including a first target vector dimension comprising a first word or phrase and a second target vector dimension comprising a second word or phrase;

determining a first number of possible magnitudes that are assignable to the first target vector dimension, wherein the magnitudes comprise weights;

determining a number of possible features associated with the source vector;

assigning a magnitude from the first number of possible magnitudes to the first target vector dimension based on a first feature of the determined number of possible features associated with the source vector;

determining a second number of magnitudes that are assignable to the second target vector dimension, wherein the magnitudes comprise weights;

assigning a magnitude from the second number of magnitudes to the second target vector dimension based on a second feature of the determined number of features associated with the source vector.

2. The non-transitory computer readable medium of claim 1, wherein the instructions further comprise instructions for:

comparing the target vector with the source vector to obtain a similarity measure.

3. The non-transitory computer readable medium of claim 1, wherein the instructions further comprise instructions for:

if the total number of magnitudes assigned to the target vector dimension is one, selecting the one assigned magnitude.

4. The non-transitory computer readable medium of claim 1, wherein receiving a target vector including a target vector dimension and a source vector comprises:

receiving the target vector from a first storage and receiving the source vector from a second storage.

5. The non-transitory computer readable medium of claim 1, wherein assigning a magnitude from the number of magnitudes assigned to the target vector dimension based on a feature of the determined number of features associated with the source vector comprises:

testing the feature associated with the source vector.

6. The non-transitory computer readable medium of claim 5, wherein testing the one of the determined number of features associated with the source vector is based on a morphological characteristic of the feature, a syntactic characteristic of the feature, a proximity of a first source vector dimension to a second source vector dimension, a frequency of appearance of one or more source vector dimensions, time or date information associated with the source vector, an originating location of the source vector, a location of an event described in a source document, a location of analysis of a source document, a purpose for analyzing the source document, or a combination thereof.

7. The non-transitory computer readable medium of claim 1, wherein assigning a magnitude from the number of magnitudes assigned to the target vector dimension based on a feature of the determined number of features associated with the source vector comprises:

assigning a magnitude from the number of magnitudes based on a magnitude associated with the feature.

8. A method for performing vector comparison in natural language processing, the method comprising:

receiving a target vector including a target vector dimension comprising a target word or phrase and a source vector including a source vector dimension comprising a source word or phrase;

determining whether the source vector dimension has more than one possible magnitude wherein the magnitude comprises a weight;

if the source vector dimension has more than one possible magnitude;

determining whether the source vector dimension exists in the target vector;

if the source vector dimension exists in the target vector, associating the target vector dimension with a first magnitude; and

if the source vector dimension does not exist in the target vector, associating the target vector dimension with a second magnitude; and

if the source vector dimension does not have more than one possible magnitude;

associating the target vector dimension with the first magnitude, the second magnitude, or a third magnitude based on a magnitude of the source vector dimension.

9. The method of claim 8, wherein receiving a target vector including a target vector dimension and a source vector including a source vector dimension comprises:

receiving the target vector from a first storage and receiving the source vector from a second storage.

10. The method of claim 8, wherein the first magnitude is a high magnitude, the second magnitude is a low magnitude, and the third magnitude is a default magnitude.

11. The method of claim 8, wherein associating the target vector dimension with the first magnitude, second magnitude, or a third magnitude based on a magnitude of the source vector dimension comprises:

determining whether the source vector dimension has the first magnitude;

if the source vector dimension has the first magnitude, associating the target vector dimension with the first magnitude;

if the source vector dimension does not have the first magnitude, determining whether the source vector dimension has the second magnitude;

if the source vector dimension has the second magnitude, associating the target vector dimension with the second magnitude; and

if the source vector dimension does not have the second magnitude, associating the target vector with the third magnitude.

12. The method of claim 8, further comprising:

determining a number of unique dimensions of the target vector; and

determining a number of unique dimensions of the source vector.

13. The method of claim 8, further comprising:

comparing the target vector and source vector to provide a comparison score, the comparison score based on the magnitude of the target vector dimension and the source vector dimension.

14. A system for performing vector comparison in natural language processing, the system comprising at least one processing unit coupled to a memory, wherein the memory is encoded with computer executable instructions that, when executed, cause the at least one processing unit to:

receive a single-magnitude source vector including a plurality of source vector dimensions each of the plurality of source vector dimensions comprising a source word or phrase, and a multi-magnitudinal target vector including a plurality of target vector dimensions, each of the plurality of target vector dimensions comprising a target word or phrase, wherein each source vector dimension of the plurality of source vector dimensions is associated with a single possible magnitude and one or more of the target vector dimension of the plurality of target vector dimensions is associated with multiple possible magnitudes, wherein each of the magnitudes comprises a weight;

determine which of the plurality of target vector dimensions are associated with multiple possible magnitudes;

resolve each of the plurality of target vector dimensions associated with multiple possible magnitudes to an actual magnitude based on a feature of a respective source vector dimension of the single-magnitude source vector; and

compare the target vector with the source vector to obtain a similarity measure.

15. The system of claim 14, wherein the instructions to resolve each of the plurality of target vector dimensions associated with multiple possible magnitudes to an actual magnitude based on a feature of a respective source vector dimension of the single-magnitude source vector cause the processor to:

test the feature of the respective source vector dimension of the single-magnitude source vector based on a morphological characteristic of the feature, a syntactic characteristic of the feature, a proximity of a first source vector dimension to a second source vector dimension, a frequency of appearance of one or more source vector dimensions, time or date information associated with the source vector, an originating location of the source vector, a location of an event described in a source document, a location of analysis of a source document, purpose for analyzing the source document, or a combination thereof.

16. The system of claim 14, wherein the instructions to compare the target vector with the source vector to obtain a similarity measure cause the processor to:

determine a vector difference between the target vector and the source vector.

17. The system of claim 14, wherein the instructions to resolve each of the plurality of target vector dimensions associated with multiple possible magnitudes to an actual magnitude based on a feature of a respective source vector dimension of the single-magnitude source vector cause the processor to:

resolve each of the plurality of target vector dimensions to a high magnitude, a low magnitude, or a default magnitude.

18. The system of claim 14, wherein the instructions to receive a single-magnitude source vector and a multi-magnitudinal target vector including a plurality of target vector dimensions cause the processor to:

receive the multi-magnitudinal target vector from a semantic data storage and receiving the single-magnitude source vector from a source data storage.

19. The system of claim 14, wherein each possible magnitude is associated with a respective indexed data structure.

20. The system of claim 14, wherein the single-magnitude source vector is associated with one or more source documents.

21. A non-transitory computer readable medium encoded with instructions for performing vector comparison in natural language processing, the instructions executable by a processor of a computing system, the instructions comprising instructions for:

receiving a target vector including a plurality of target vector dimensions, each of the plurality of target vector dimensions comprising a target word or phrase, and a source vector;

determining a number of possible magnitudes that are assignable to each of the plurality of target vector dimensions, wherein the magnitudes comprise weights;

determining a number of features associated with the source vector; and

assigning a magnitude from the number of possible magnitudes to each of the plurality of target vector dimensions based on a respective feature of the determined number of features associated with the source vector.