System and method for clustering interest patterns based on a plurality of priority values

• US 10,460,242 B1
• Filed: 06/06/2018
• Issued: 10/29/2019
• Est. Priority Date: 06/06/2018
• Status: Active Grant

• Current Assignee: Fulcrum Management Solutions Ltd.
• Original Assignee: Fulcrum Management Solutions Ltd.
• Inventors: Procter, Thomas John
• Primary Examiner(s): Nilsson, Eric
• Application Number:
  US16/001,375
• Time in Patent Office: 510 Days
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

A system and method for clustering interest for a plurality of participant devices based on open-ended, free-form communication between a plurality of user devices using priority value responses from the plurality of participant devices based on distributed thought objects associated to the open-ended, free-form communication. The system and method using a ratings matrix, comprising a plurality of priority values, that is permutated by assigning participant devices into interest clusters by first suing a strict association method and then increasing cohorts by using a tolerant association method.

Claims

1. A system for clustering interest for a plurality of participant objects and a plurality of thought objects based on priority values comprising:

a network-connected interest clustering computer comprising a processor, a memory, and programming instructions, the programming instructions, when executed by the processor, cause the processor to cluster interest for a plurality of participant objects and thought objects comprising;

a device interface;

a plurality of user devices;

a project controller;

a matrix calculator;

a question object;

a plurality of participant objects;

a plurality of thought objects;

a score calculator;

a pattern analyzer;

wherein the device interface receives a plurality of connections from the plurality of user devices on a network;

wherein the project controller;

associates each user device to a participant object of the plurality of participant objects;

receives a question object comprising, at least, an arrangement of information from a first user device of the plurality of user devices;

sends the question object to at least a portion of the plurality of user devices;

receives the plurality of thought objects from at least a portion of the user devices;

sends the plurality of the thought objects to at least a portion of the user devices;

receives a plurality of priority value responses from a plurality of participant devices, each priority value response associated to a thought object of the plurality of thought objects and to a corresponding participant object of the plurality of participant objects;

wherein the matrix calculator;

computes a ratings matrix, the ratings matrix based on the plurality of priority values, the ratings matrix comprised of at least a portion of the plurality of participant devices and at least a portion of the plurality of thought objects;

permutes the ratings matrix into an interest-based submatrix, the permutation identifying a plurality of cohorts comprising at least a portion of the plurality of participant objects and at least a portion of the plurality of thought objects based on similar patterns of priority value responses;

wherein the participant devices are comprised from at least a portion of the user devices;

wherein each priority value fall within a predefined range;

wherein to compute the submatrix using a strict association level, the matrix calculator is operable to;

remove participant objects of the plurality of participant objects, that have no associated priority values;

remove thought objects of the plurality of thought objects, that have no associated priority values;

convert the plurality of priority values into a range centered around zero value;

filter at least a portion of the plurality of thought objects based on an associated polarization score, the polarization score calculated by summing the absolute values of the plurality of priority values assigned to an associated thought object and subtracting the absolute value of a sum of the plurality of priority values;

filter at least a portion of the plurality of participant objects based on an associated passion score, the passion score calculated by summing the absolute values of the plurality of priority values assigned by an associated participant object;

assign each participant object of the plurality of participant objects to its own cohort of a plurality of cohorts;

wherein the score calculator is operable to;

calculate a polarization score for each thought object of the plurality of thought objects;

calculate a passion score for each participant object of the plurality of participant objects;

wherein the pattern analyzer is operable to compare a plurality of assigned priority value responses associated to a first cohort, to a second plurality of assigned priority value responses associated to a second cohort, to determine an agreement, wherein if there is agreement, merging the first cohort with the second cohort;

wherein the agreement between the first priority value response and the second priority value response is determined by comparing, by the pattern analyzer, the signs associated to the first plurality of assigned priority value responses and the second plurality of assigned priority value responses;

wherein the predefined range is a zero-centered scale;

wherein if two cohorts, the matrix calculator is operable to assign a binarized label to each participant object and to each thought object based on associated priority values;

further wherein if other than two cohorts remain, the score calculator is further operable to calculate a passion score for each remaining cohort;

wherein if the passion score is less than a predefined threshold, matrix calculator is operable to iteratively remove at least a portion of participant objects, the at least portion of participant objects associated to one or more remaining cohorts with the lowest passion score,

wherein if the passion score is greater than the predefined threshold;

the score calculator is further operable to iteratively calculate a polarization score for the at least portion of thought objects;

the matrix calculator is further operable to iteratively remove at least a portion of the plurality of thought objects, the at least portion of the plurality of thought objects associated to one or more thought objects with the lowest polarization score.

2. The system of claim 1, wherein the pattern analyzer is further operable to:

label the at least one thought object, if it is determined that at least one additional thought object, of the plurality of thought objects, should be labeled;

label the at least one participant object, if it is determined, by the pattern analyzer, that at least one additional participant object, of the plurality of participant objects, should be labeled;

remove the at least one thought object, by an assignment of a zero label, if it is determined, by the pattern analyzer, that at least one thought object with a non-zero label, of the plurality of thought objects, should be removed;

remove the at least one participant object, by an assignment of a zero label, if it is determined, by the pattern analyzer, that at least one participant object with a non-zero label, of the plurality of participant objects, should be removed;

wherein if at least one thought object or at least one participant object is labeled or removed iteratively labeling or removing, or both, the at least one thought object or the at least one participant object, or both;

wherein removing the at least one thought object is performed by labeling the at least one thought object as zero;

wherein removing the at least one participant object is performed by labeling the at least one participant object as zero;

wherein the determination to remove the at least one thought object is based on a comparison of patterns of priority values assigned to the at least one thought object by one or more participant objects of the plurality of participant objects, indicating an incompatibility between the label of the at least one thought object, the labels of the one or more participant objects, and the signs of the priority values assigned to the at least one thought object by the one or more participant objects;

wherein the determination to remove the at least one participant object is based on a comparison of patterns of priority values assigned by the at least one participant object to one or more thought objects of the plurality of thought objects, indicating an incompatibility between the label of the at least one participant object, the labels of the one or more thought objects, and the signs of the priority values assigned by the at least one participant object to the one or more thought objects.

3. The system of claim 2, wherein labeling the at least one thought object comprises:

a computation of a positive submatrix from at least a first portion of priority values from the ratings matrix, the at least first portion of priority values associated to a participant object with a positive label and to a thought object of the plurality of thought objects labeled as zero;

an analysis of one or more priority value responses associated to a first thought object from the positive submatrix;

a comparison of the number of positive priority value responses to the number of negative priority value responses;

wherein if the first thought object is associated to more positive priority value responses than negative priority value responses, a positive label to the first thought object is assigned to the first thought object, otherwise a negative label is assigned to the first thought object;

a computation of a negative submatrix from at least a second portion of priority values from the ratings matrix, the at least second portion of priority values associated to a participant object with a negative label and to a thought object of the plurality of thought objects labeled as zero;

an analysis of one or more priority value responses associated to a first thought object from the negative submatrix;

a comparison of the number of positive priority value responses to the number of negative priority value responses;

wherein if the first thought object is associated to more negative priority value responses than negative priority value responses, a negative label is assigned to the first thought object, otherwise a positive label is assigned to the first thought object;

wherein the matrix calculator is operable to iteratively compute, analyze and compare each thought object associated to the positive submatrix;

wherein the matrix calculator is operable to iteratively compute, analyze and compare each thought object associated to the negative submatrix.

4. The system of claim 3, wherein the comparison of the number of positive priority value responses to the number of negative priority value responses is based on a ratio.

5. The system of claim 4, wherein the matrix calculator is further operable to compare a temporary label associated to a common thought object;

wherein the common thought object is present in both the positive submatrix to the negative submatrix;

wherein if the temporary label associated to the common thought object in the positive submatrix is the same as the temporary label associated to the common thought object in the negative submatrix, assign the matched temporary label to the label of the common thought object.

6. The system of claim 2, wherein labeling the at least one participant object comprises:

a computation of a positive submatrix from at least a third portion of priority values from the ratings matrix, the at least third portion associated to a participant object labeled as zero and to a thought object of the plurality of thought objects with a positive label;

an analysis of one or more priority value responses associated to a first participant object from the positive submatrix;

a comparison of the number of positive priority value responses to the number of negative priority value responses associated to the first participant object;

wherein if the first participant object is associated more positive priority value responses than negative priority value responses, a positive label is assigned to the first participant object, otherwise a negative label is assigned to the first participant object;

a computation of a negative submatrix from at least a fourth portion of priority values from the ratings matrix, the at least fourth portion associated to a participant object labeled as zero and to a thought object of the plurality of thought objects with a negative label;

an analysis of one or more priority value responses associated to a first participant object from the negative submatrix;

a comparison of the number of positive priority value responses to the number of negative priority value responses associated to the first participant object;

wherein if the first participant object is associated to more negative priority value responses than positive priority value responses, a negative label is assigned to the first participant object, otherwise a positive label assigning to the first participant object;

wherein the matrix calculator is operable to iteratively compute, analyze and compare each participant object associated to the positive submatrix;

wherein the matrix calculator is operable to iteratively compute, analyze and compare each participant object associated to the negative submatrix.

7. The system of claim 6, wherein the comparison of the number of positive priority value responses to the number of negative priority value responses is based on a ratio.

8. The system of claim 7, wherein the matrix calculator is further operable to compare a temporary label associated to a common participant object;

wherein the common participant object is present in both the positive submatrix to the negative submatrix;

wherein if the temporary label associated to the common participant object in the positive submatrix is the same as the temporary label associated to the common participant object in the negative submatrix, assign the matched temporary label to the label of the common participant object.

9. A method for clustering thought objects and participant objects based on interest for a plurality of objects executed by a network-connected processor according to a plurality of programmable instructions, the method comprising:

receiving, at a device interface, a plurality of connections from a plurality of user devices on a network;

associating, by a project controller, each user device to a participant object of a plurality of participant objects;

receiving, at the project controller, one or more question objects comprising, at least, an arrangement of information from a first user device of the plurality of user devices, the project controller communicatively connected to the device interface;

sending, by the project controller, the one or more question objects to at least a portion of the plurality of user devices;

receiving, at the project controller, a plurality of thought objects from at least a portion of the user devices;

sending, by the project controller, the plurality of the thought objects to at least a portion of the user devices;

receiving, at the project controller, a plurality of priority value responses from a plurality of participant devices, each priority value response associated to a thought object of the plurality of thought objects and to a corresponding participant object of the plurality of participant objects;

computing, at a matrix calculator, a ratings matrix, the ratings matrix based on the plurality of priority values, the ratings matrix comprised of at least a portion of the plurality of participant devices and at least a portion of the plurality of thought objects;

permuting, by the matrix calculator, the ratings matrix into an interest-based submatrix, the permutation identifying a plurality of cohorts comprising at least a portion of the plurality of participant objects and at least a portion of the plurality of thought objects based on similar patterns of priority value responses;

wherein the participant devices are comprised from at least a portion of the user devices;

wherein each priority value falls within a predefined range;

wherein the submatrix is computed using a strict association level comprising the steps of;

(a) removing, by the matrix calculator, participant objects of the plurality of participant objects, that have no associated priority values;

(b) removing, by the matrix calculator, thought objects of the plurality of thought objects, that have no associated priority values;

(c) converting, by the matrix calculator, the plurality of priority values into a range centered around zero value;

(d) calculating, by a score calculator, a polarization score for each thought object by summing the absolute values of the plurality of priority values assigned to an associated thought object and subtracting the absolute value of a sum of the plurality of priority values;

(e) filtering, by the matrix calculator, at least a portion of the plurality of thought objects based on an associated polarization score;

(f) calculating, by a score calculator, a passion score for each participant object of the plurality of participant objects by summing the absolute values of the plurality of priority values assigned by an associated participant object;

(g) filtering, by the matrix calculator, at least a portion of the plurality of participant objects based on an associated passion score;

(h) assigning, by the matrix calculator, each participant object of the plurality of participant objects to its own cohort of a plurality of cohorts;

(i) selecting, by a pattern analyzer, a first cohort of the plurality of cohorts;

(j) iteratively comparing, by the pattern analyzer, a plurality of assigned priority value responses associated to the first cohort of the plurality of cohorts, to each plurality of assigned priority value responses associated to each remaining cohort of the plurality of cohorts, to determine an agreement;

(k) if there is agreement, merging the first cohort with the second cohort;

(l) selecting, by the pattern analyzer, a next cohort;

(m) repeating steps (j), (k), and (l) for all cohorts;

wherein the agreement between the first priority value response and the second priority value response is determined by comparing, by the pattern analyzer, signs associated to the first plurality of assigned priority value responses and signs associated to the second plurality of assigned priority value responses;

wherein the predefined range is a zero-centered scale;

if two cohorts remain;

(n) assigning, by matrix calculator, a binarized label to each participant object and to each thought object based on associated priority values;

otherwise;

(o) calculating, by the score calculator, a passion score for at least a portion of the remaining cohorts;

if the lowest of the passion scores associated to the remaining cohorts is less than a predefined threshold;

(p) removing, by the pattern analyzer, at least a portion of participant objects, the at least portion of participant objects associated to one or more remaining cohorts with the lowest passion score;

(q) returning to step (d);

otherwise;

(r) calculating, by the score calculator, a polarization score for the at least portion of the remaining thought objects;

(s) removing, by the matrix calculator, at least a portion of the plurality of thought objects, the at least portion of the plurality of thought objects associated to one or more thought objects with the lowest polarization score;

(t) returning to step (d).

10. The method of claim 9, further comprising the steps of:

(u) if it is determined, by the pattern analyzer, that at least one additional thought object, of the plurality of thought objects, should be labeled, labeling the at least one thought object;

(v) if it is determined, by the pattern analyzer, that at least one additional participant object, of the plurality of participant objects, should be labeled, labeling the at least one participant object;

(w) if it is determined, by the pattern analyzer, that at least one thought object with a non-zero label, of the plurality of thought objects should be removed, removing the at least one thought object by assigning a zero label;

(x) if it is determined, by the pattern analyzer, that at least one participant object with a non-zero label, of the plurality of participant objects, should be removed, removing the at least one participant object by assigning a zero label;

(y) if at least one thought object or at least one participant object is labeled or removed in steps (u), (v), (w), or (x), repeating steps (u), (v), (w), and (x);

wherein steps (u), (v), (w), and (x) are performed in any order;

wherein the determination to remove the at least one additional thought object is based on a comparison of patterns of priority values assigned to the at least one thought object by one or more participant objects of the plurality of participant objects, indicating an incompatibility between the label of the at least one thought object, the labels of the one or more participant objects, and the signs of the priority values assigned to the at least one thought object by the one or more participant objects;

wherein the determination to remove the at least one additional participant object is based on a comparison of patterns of priority values assigned by the at least one participant object to one or more thought objects of the plurality of thought objects, indicating an incompatibility between the label of the at least one participant object, the labels of the one or more thought objects of the plurality of thought objects, and the signs of the priority values assigned by the at least one participant object to one or more thought objects of the plurality of thought objects.

11. The method of claim 10, wherein labeling the at least one thought object comprises the steps of:

computing, by the matrix calculator, a positive submatrix from at least a first portion of the plurality of priority values from the ratings matrix, the at least first portion of priority values associated to a participant object with a positive label and to a thought object of the plurality of thought objects labeled as zero;

analyzing, by the matrix calculator, one or more priority value responses associated to a first thought object from the positive submatrix;

comparing, by the matrix calculator, the number of positive priority value responses to the number of negative priority value responses associated to the first thought object;

if the first thought object is associated to more positive priority value responses than negative priority value responses, associating, by the matrix calculator, a positive label to the first thought object;

otherwise associating, by the matrix calculator, a negative label to the first thought object;

repeating the previous four steps for each thought object associated to the positive submatrix;

computing, by the matrix calculator, a negative submatrix from at least a second portion of priority values from the ratings matrix, the at least second portion of priority values associated to a participant object with a negative label and to a thought object of the plurality of thought objects labeled as zero;

analyzing, by the matrix calculator, one or more priority value responses associated to a first thought object from the negative submatrix;

comparing, by the matrix calculator, the number of positive priority value responses to the number of negative priority value responses associated to the first thought object;

if the first thought object is associated to more negative priority value responses than positive priority value responses, associating, by the matrix calculator, a negative label to the first thought object;

otherwise associating, by the matrix calculator, a positive label to the first thought object;

repeating the previous four steps for each thought object associated to the negative submatrix.

12. The method of claim 11, wherein the comparison of the number of positive priority value responses to the number of negative priority value responses is based on a ratio.

13. The method of claim 12, wherein the matrix calculator is further operable to compare a temporary label associated to a common thought object;

wherein the common thought object is present in both the positive submatrix to the negative submatrix;

wherein if the temporary label associated to the common thought object in the positive submatrix is in agreement with the temporary label associated to the common participant object in the negative submatrix, assigning the matched temporary label to the label of the common thought object.

14. The method of claim 10, wherein labeling the at least one participant object comprises the steps of:

computing, by the matrix calculator, a positive submatrix from at least a third portion of priority values from the ratings matrix, the at least third portion associated to a participant object labeled as zero and to a thought object of the plurality of thought objects with a positive label;

analyzing, by the matrix calculator, one or more priority value responses associated to a first participant object from the positive submatrix;

comparing, by the matrix calculator, the number of positive priority value responses to the number of negative priority value responses associated to the first participant object;

if the first participant object is associated to more positive priority value responses than negative priority value responses, associating, by the matrix calculator, a positive label to the first participant object;

otherwise associating, by the matrix calculator, a negative label to the first participant object;

repeating the previous four steps for each participant object associated to the positive submatrix;

computing, by the matrix calculator, a negative submatrix from at least a fourth portion of priority values from the ratings matrix, the at least fourth portion associated to a participant object labeled as zero and to a thought object of the plurality of thought objects with a negative label;

analyzing, by the matrix calculator, one or more priority value responses associated to a first participant object from the negative submatrix;

comparing, by the matrix calculator, the number of positive priority value responses to the number of negative priority value responses associated to the first participant object;

if the first participant object is associated to more negative priority value responses than positive priority value responses, associating, by the matrix calculator, a negative label to the first participant object;

otherwise associating, by the matrix calculator, a positive label to the first participant object;

repeating the previous four steps for each participant object associated to the negative submatrix.

15. The method of claim 14, wherein the comparison of the number of positive priority value responses to the number of negative priority value responses is based on a ratio.

16. The method of claim 15, wherein the matrix calculator is further operable to compare a temporary label associated to a common participant object;

wherein the common participant object is present in both the positive submatrix to the negative submatrix;

wherein if the temporary label associated to the common participant object in the positive submatrix is not in disagreement with the temporary label associated to the common participant object in the negative submatrix, assigning the matched temporary label to the label of the common participant object.