US 10,885,278 B2Granted

Auto tele-interview solution

AssigneeRoyal Bank of Canada

InventorsChai K. Lam, Xuong Hue Tran, Kulbinder Mann, Lori May Beesack, Edward C. Wong

Life of this patent

Estimating the term…

Loading patent…

Abstract

Computer-implemented systems and methods are provided for improved generation and control of conversations. A computing device is utilized to control or simulate conversation using estimated contextual cues extracted from profile information or prior responses. The computing device is configured to automatically tailor a flow of a conversation to an effort to improve relevancy and engagement without the need of a human operator to manually tailor the conversation, which, for example, could be impractically expensive. A structured workflow is maintained in the form of a series of conversation decisions, and a machine learning engine is utilized to maintain a continuously trained data structure that generates predictions that bias conversation decisions (e.g., by weighting tree options) for tailoring the conversation flow.

Claims20

Dependent claims

A computer system for dynamically determining, from a set of question data objects, a next best question data object for presentment to a user, the computer system comprising:

a first data storage configured to store one or more data structures representative of a decision tree learning network maintaining one or more interconnected computing units representing at least:
(i) a first set of features representative of one or more observed characteristics of a computing platform or a network connection being used by the user;
(ii) a second set of features representative of an estimated current intent, mental state or disposition of the user;
and (iii) a third set of features representative of information retrieved from a stored profile of the user;
a second data storage configured to maintain the stored profile of the user, the profile including at least one or more data fields that are updated as new elements of information are received from the user;
a processor configured to:
train the network using a set of historical training data from previous recordings of interviews to establish weighted interconnections between the one or more interconnected computing units, wherein the set of historical training data is used to adjust a probability ranking at each node is adjusted based on a frequency of a path taken in the set of historical training data;
monitor the one or more observed characteristics of the computing platform or the network connection being used by the user in provisioning of an incoming answer string and update the first set of features based on the one or more observed characteristics;
parse the incoming answer string from the user into one or more string tokens, and using a natural language processing engine, process the one or more string tokens to estimate the current intent, mental state or disposition of the user and to update the second set of features;
process the one or more string tokens to extract the new elements of information from the incoming answer string;
process each question data object from the set of question data objects through the network to establish a match score for each question data object;
and
select the next best question data object from the set of question data objects based at least on the match score corresponding to each question data object.

10 more elements

A computer implemented method for dynamically determining, from a set of question data objects, a next best question data object for presentment to a user, the method performed by a computer processor, the method comprising:

maintaining one or more data structures representative of a decision tree learning network storing one or more interconnected computing units representing at least:
(i) a first set of features representative of one or more observed characteristics of a computing platform or a network connection being used by the user;
(ii) a second set of features representative of an estimated current intent, mental state or disposition of the user;
and (iii) a third set of features representative of information retrieved from a stored profile of the user;
maintaining the stored profile of the user, the profile including at least one or more data fields that are updated as new elements of information are received from the user;
training the network using a set of historical training data from previous recordings of interviews to establish weighted interconnections between the one or more interconnected computing units, the set of historical training data used to adjust a probability ranking at each node is adjusted based on a frequency of a path taken in the set of historical training data;
monitoring the one or more observed characteristics of the computing platform or the network connection being used by the user in provisioning of an incoming answer string and update the first set of features based on the one or more observed characteristics;
parsing the incoming answer string from the user into one or more string tokens, and using a natural language processing engine, processing the one or more string tokens to estimate the current intent, mental state or disposition of the user and to update the second set of features;
processing the one or more string tokens to extract the new elements of information from the incoming answer string;
processing each question data object from the set of question data objects through the network to establish a match score for each question data object;
and
selecting the next best question data object from the set of question data objects based at least on the match score corresponding to each question data object.

9 more elements

A non-transitory computer readable medium storing machine interpretable instructions, which when executed by a computer processor, cause the computer processor to perform steps of a comprising:

maintaining one or more data structures representative of a decision tree learning network storing one or more interconnected computing units representing at least:
(i) a first set of features representative of one or more observed characteristics of a computing platform or a network connection being used by the user;
(ii) a second set of features representative of an estimated current intent, mental state or disposition of the user;
and (iii) a third set of features representative of information retrieved from a stored profile of the user;
maintaining the stored profile of the user, the profile including at least one or more data fields that are updated as new elements of information are received from the user;
training the network using a set of historical training data from previous recordings of interviews to establish weighted interconnections between the one or more interconnected computing units, the set of historical training data used to adjust a probability ranking at each node is adjusted based on a frequency of a path taken in the set of historical training data;
monitoring the one or more observed characteristics of the computing platform or the network connection being used by the user in provisioning of an incoming answer string and update the first set of features based on the one or more observed characteristics;
parsing the incoming answer string from the user into one or more string tokens, and using a natural language processing engine, processing the one or more string tokens to estimate the current intent, mental state or disposition of the user and to update the second set of features;
processing the one or more string tokens to extract the new elements of information from the incoming answer string;
processing each question data object from the set of question data objects through the network to establish a match score for each question data object;
and
selecting the next best question data object from the set of question data objects based at least on the match score corresponding to each question data object.

9 more elements

Dependencies are inferred from the claim text.

Facts

Application: 16/163,212Filed 17 Oct 2018 · us provisional application
Priority: 17 Oct 2017
Published: US 2019/0114321 A118 Apr 2019
Granted: 5 Jan 2021
Term ends: Not estimatedThe source records don’t support an estimate.

Classification17 codes

G06F 40/30
Handling natural language data: semantic analysis
G06F 16/2246
Indexing structures: trees, e.g. B+trees
G06F 16/24522
Query translation: translation of natural language queries to structured queries
G06F 40/205
Natural language analysis: parsing