System and method for incremental estimation of interlocutor intents and goals in turn-based electronic conversational flow
A system and method implemented on a computing device for incrementally discovering new intents and goals by collecting data from a first corpus of a plurality of newer digitally-recorded conversations, performing dimensionality reduction to prepare the extracted conversations for clustering, clustering the prepared conversations, identifying new intent and/or goal labels using a trained Artificial Intelligence (AI) engine, applying a model-based filter to remove new labels which overlap already-known labels in the first corpus, and outputting the newly-discovered labels in association with the extracted conversations into a computer-readable file.
1. A method implemented by a computing device for incrementally discovering an intent or a goal or both an intent and a goal of a party in an interlocutor digital conversation, the method comprising:
collecting data, by a computer system, from a first corpus of a plurality of digitally-recorded conversations by extracting conversations with a creation date newer than a threshold date;
performing dimensionality reduction, by a computer system, to prepare the extracted conversations for clustering;
clustering, by the computer system, the prepared extracted conversations;
subsequent to the clustering, identifying, by a computer system, one or more new intent labels or a new goal label or a combination of new intent labels and new goal labels using a trained Artificial Intelligence (AI) engine;
applying a model-based filter, by a computer system, to the one or more new intent labels or the new goal label or the combination of new intent labels and new goal labels to remove new labels which overlap already-known labels in the first corpus; and
outputting, by a computer system, the new labels in association with the extracted conversations into a computer-readable file.
2. The method as set forth in claim 1 wherein the collecting data comprises retrieving a current taxonomy.
3. The method as set forth in claim 1 further comprising:
counting, by a computer system, a number of extracted conversations; and
responsive to the number of extracted conversations falling below a threshold count, determining, by a computer system, the method without performing remaining steps.
4. The method as set forth in claim 1 further comprising storing, by a computer system, the extracted conversations into a second corpus.
5. The method as set forth in claim 4 further comprising:
extracting, by a computer system, from the first corpus one or more conversations which have associated model scores below a low score threshold; and
merging, by a computer system, one or more extracted low model score conversations into the second corpus.
6. The method as set forth in claim 1 wherein the dimensionality reduction comprises a Uniform Manifold Approximation and Projection process.
7. The method as set forth in claim 1 wherein the applying of the model-based filter comprises selecting extracted conversation clusters according to a model score threshold.
8. The method as set forth in claim 1 wherein the applying of the model-based filter comprises selecting extracted conversation clusters according to an entropy threshold.
9. A computer program product for incrementally discovering an intent or a goal or both an intent and a goal of a party in an interlocutor digital conversation, the computer program product comprising:
one or more tangible, non-transitory computer-readable memory devices which are not propagating signals per se; and
one or more program instructions encoded by the one or more tangible, non-transitory computer-readable memory devices configured to cause one or more computer processors to perform steps comprising:
collecting data from a first corpus of a plurality of digitally-recorded conversations by extracting conversations with a creation date newer than a threshold date;
performing dimensionality reduction to prepare the extracted conversations for clustering;
clustering the prepared extracted conversations;
subsequent to the clustering, identifying one or more new intent labels or a new goal label or a combination of new intent labels and new goal labels using a trained Artificial Intelligence (AI) engine;
applying a model-based filter to the one or more new intent labels or the new goal label or the combination of new intent labels and new goal labels to remove new labels which overlap already-known labels in the first corpus; and
outputting the new labels in association with the extracted conversations into a computer-readable file.
10. The computer program product as set forth in claim 9 wherein the collecting data comprises retrieving a current taxonomy.
11. The computer program product as set forth in claim 9 wherein the one or more program instructions further comprise one or more program instructions configured to cause one or more computer processors to perform the steps comprising:
counting a number of extracted conversations; and
responsive to the number of extracted conversations falling below a threshold count, determining a method without performing remaining steps.
12. The computer program product as set forth in claim 9 wherein the one or more program instructions further comprise one or more program instructions configured to cause one or more computer processors to perform the step comprising storing the extracted conversations into a second corpus.
13. The computer program product as set forth in claim 12 wherein the one or more program instructions further comprise one or more program instructions configured to cause one or more computer processors to perform the steps comprising:
extracting from the first corpus one or more conversations which have associated model scores below a low score threshold; and
merging one or more extracted low model score conversations into the second corpus.
14. The computer program product as set forth in claim 9 wherein the dimensionality reduction comprises a Uniform Manifold Approximation and Projection process.
15. The computer program product as set forth in claim 9 wherein the applying of the model-based filter comprises selecting extracted conversation clusters according to a model score threshold.
16. The computer program product as set forth in claim 9 wherein the applying of the model-based filter comprises selecting extracted conversation clusters according to an entropy threshold.
17. A system for incrementally discovering an intent or a goal or both an intent and a goal of a party in an interlocutor digital conversation, the system comprising:
one or more computer processors;
one or more tangible, non-transitory computer-readable memory devices which are not propagating signals per se; and
one or more program instructions encoded by the one or more tangible, non-transitory computer-readable memory devices configured to cause the one or more computer processors to perform steps comprising:
collecting data from a first corpus of a plurality of digitally-recorded conversations by extracting conversations with a creation date newer than a threshold date;
performing dimensionality reduction to prepare the extracted conversations for clustering;
clustering the prepared extracted conversations;
subsequent to the clustering, identifying one or more new intent labels or a new goal label or a combination of new intent labels and new goal labels using a trained Artificial Intelligence (AI) engine;
applying a model-based filter to the one or more new intent labels or the new goal label or the combination of new intent labels and new goal labels to remove new labels which overlap already-known labels in the first corpus; and
outputting the new labels in association with the extracted conversations into a computer-readable file.
18. The system as set forth in claim 17 wherein the collecting data comprises retrieving a current taxonomy.
19. The system as set forth in claim 17 wherein the one or more program instructions further comprise one or more program instructions configured to cause one or more computer processors to perform the steps comprising:
counting a number of extracted conversations; and
responsive to the number of extracted conversations falling below a threshold count, determining a method without performing remaining steps.
20. The system as set forth in claim 17 wherein the one or more program instructions further comprise one or more program instructions configured to cause one or more computer processors to perform the step comprising storing the extracted conversations into a second corpus.
21. The system as set forth in claim 17 wherein the dimensionality reduction comprises a Uniform Manifold Approximation and Projection process.
22. The system as set forth in claim 17 wherein the applying of the model-based filter comprises selecting extracted conversation clusters according to a model score threshold.
23. The system as set forth in claim 17 wherein the applying of the model-based filter comprises selecting extracted conversation clusters according to an entropy threshold.
24. The system as set forth in claim 20 wherein the one or more program instructions further comprise one or more program instructions configured to cause one or more computer processors to perform the steps comprising: extracting from the first corpus one or more conversations which have associated model scores below a low score threshold; and merging the one or more extracted low model score conversations into the second corpus.