US 11,687,730 B1Granted

Automated conversation goal discovery using neural networks and deep multi-view clustering

AssigneeAsapp, Inc.

InventorsYi Yang, Hugh Nicholas Perkins

Life of this patent

Estimating the term…

Loading patent…

Abstract

The present disclosure describes a system, method, and computer program for automatically discovering goals from conversations using neural networks and deep multi-view clustering. A dataset of conversations is partitioned into two views. Vector representations of each view are then generated and clustered in an alternating fashion between views for a number of iterations (i.e., the system alternates between views in generating and clustering vector representations of a view). A first neural network encoder generates the vector representations for the first view, and a second neural network encoder generates the vector representations for the second view. With each semi-iteration, cluster assignments from one view are used to update the encoder for the other view, thus encouraging the two neural network encoders to yield similar cluster assignments. After all the iterations are complete, a user interface enables a user to label each first-view cluster with a goal, where a subset of example utterances is displayed for each cluster.

Claims20

Dependent claims

A method, performed by a computer system, for automatically discovering goals from conversations between two or more entities, the method comprising:

(a) obtaining a dataset of conversations;
(b) obtaining a first view and a second view of the dataset, wherein a first view of a conversation comprises a first set of utterances from a first entity, and a second view of the same conversation comprises a second set of utterances in the conversation from both the first entity and one or more other entities;
(c) for each conversation in the dataset, using a first neural network encoder to encode the first view of the conversation into a first-view vector;
(d) clustering the first-view vectors into K clusters to obtain first-view cluster assignments, wherein K is an integer greater than 1 and wherein the clustering is unsupervised;
(e) optimizing a second neural network encoder to predict the first-view cluster assignments using only the second-view data and the first-view cluster assignments;
(f) for each conversation in the dataset, using the optimized second neural network to encode the second-view of the conversation into a second-view vector;
(g) clustering the second-view vectors into K clusters to obtain second-view cluster assignments, wherein the clustering is unsupervised;
(h) optimizing the first neural network encoder to predict the second-view cluster assignments using only the first-view data and the second-view cluster assignments;
(i) generating new first-view vectors using the optimized first neural network encoder;
(j) clustering the new first-view vectors into K clusters to obtain new first-view cluster assignments, wherein the clustering is unsupervised;
(k) repeating steps (e)-(j) for a number of iterations until the first neural network encoder and the second neural network encoder yield similar cluster assignments such that the first-view cluster assignment and the second-view cluster assignment are in agreement above a threshold;
and
(l) providing a user interface that enables a user to label each first-view cluster with a goal, wherein providing the user interface includes displaying a subset of the first set of utterances for each of the first-view clusters.

10 more elements

A non-transitory computer-readable medium comprising a computer program, that, when executed by a computer system, enables the computer system to perform the following method for automatically discovering goals from conversations between two or more entities, the method comprising:

(a) obtaining a dataset of conversations;
(b) obtaining a first view and a second view of the dataset, wherein a first view of a conversation comprises a first set of utterances from a first entity, and a second view of the same conversation comprises a second set of utterances in the conversation from both the first entity and one or more other entities;
(c) for each conversation in the dataset, using a first neural network encoder to encode the first view of the conversation into a first-view vector;
(d) clustering the first-view vectors into K clusters to obtain first-view cluster assignments, wherein K is an integer greater than 1 and wherein the clustering is unsupervised;
(e) optimizing a second neural network encoder to predict the first-view cluster assignments using only the second-view data and the first-view cluster assignments;
(f) for each conversation in the dataset, using the optimized second neural network to encode the second-view of the conversation into a second-view vector;
(g) clustering the second-view vectors into K clusters to obtain second-view cluster assignments, wherein the clustering is unsupervised;
(h) optimizing the first neural network encoder to predict the second-view cluster assignments using only the first-view data and the second-view cluster assignments;
(i) generating new first-view vectors using the optimized first neural network encoder;
(j) clustering the new first-view vectors into K clusters to obtain new first-view cluster assignments, wherein the clustering is unsupervised;
(k) repeating steps (e)-(j) for a number of iterations until the first neural network encoder and the second neural network encoder yield similar cluster assignments such that the first-view cluster assignment and the second-view cluster assignment are in agreement above a threshold;
and
(l) providing a user interface that enables a user to label each first-view cluster with a goal, wherein providing the user interface includes displaying a subset of the first set of utterances for each of the first-view clusters.

10 more elements

A computer system for automatically discovering goals from conversations between two or more entities, the system comprising:

one or more processors;
one or more memory units coupled to the one or more processors, wherein the one or more memory units store instructions that, when executed by the one or more processors, cause the system to perform the operations of:
(a) obtaining a dataset of conversations;
(b) obtaining a first view and a second view of the data set, wherein a first view of a conversation comprises a first set of utterances from a first entity, and a second view of the same conversation comprises a second set of utterances in the conversation from both the first entity and one or more other entities;
(c) for each conversation in the dataset, using a first neural network encoder to encode the first view of the conversation into a first-view vector;
(d) clustering the first-view vectors into K clusters to obtain first-view cluster assignments, wherein K is an integer greater than 1 and wherein the clustering is unsupervised;
(e) optimizing a second neural network encoder to predict the first-view cluster assignments using only the second-view data and the first-view cluster assignments;
(f) for each conversation in the dataset, using the optimized second neural network to encode the second-view of the conversation into a second-view vector;
(g) clustering the second-view vectors into K clusters to obtain second-view cluster assignments, wherein the clustering is unsupervised;
(h) optimizing the first neural network encoder to predict the second-view cluster assignments using only the first-view data and the second-view cluster assignments;
(i) generating new first-view vectors using the optimized first neural network encoder;
(j) clustering the new first-view vectors into K clusters to obtain new first-view cluster assignments, wherein the clustering is unsupervised;
(k) repeating steps (e)-(j) for a number of iterations until the first neural network encoder and the second neural network encoder yield similar cluster assignments such that the first-view cluster assignment and the second-view cluster assignment are in agreement above a threshold;
and
(l) providing a user interface that enables a user to label each first-view cluster with a goal, wherein providing the user interface includes displaying a subset of the first set of utterances for each of the first-view clusters.

12 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/319,797Filed 13 May 2021 · continuation
Priority: 21 May 2019
Granted: 27 Jun 2023
Term ends: Not estimatedThe source records don’t support an estimate.

Classification12 codes

G06F 40/35
Semantic analysis: discourse or dialogue representation
G06F 40/284
Recognition of textual entities: lexical analysis, e.g. tokenisation or collocates
G06N 3/0442
Recurrent networks, e.g. Hopfield networks: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
G06N 3/0455
Combinations of networks: auto-encoder networks; Encoder-decoder networks