US-12619668-B1 - Computer-implemented methods utilizing machine learning to generate a question and answer pair for a conversational agent

US12619668B1US 12619668 B1US12619668 B1US 12619668B1US-12619668-B1

Abstract

Techniques for utilizing machine learning to generate a question and answer pair for a conversational agent are described. According to some examples, a computer-implemented method includes receiving a search indication from a user; determining a plurality of corresponding aspects for a plurality of suppliers based on a set of user reviews; generating, by one or more machine learning models, one or more contextually relevant aspects from the plurality of corresponding aspects based on the search indication for individual ones of the plurality of suppliers; generating, by the one or more machine learning models, a corresponding supplier related question for the individual ones of the plurality of suppliers based on the one or more contextually relevant aspects; selecting a supplier related question from the corresponding supplier related questions; generating, by the one or more machine learning models, a corresponding answer to the supplier related question based on one or more of the set of user reviews; and causing the supplier related question and the corresponding answer to be presented the user.

Inventors

Shaunak MISHRA
Amjad Y. A. Abu Jbara
Shanmugavelayutham Muthukrishnan
Atul Kamat
Akshit Mehta
Brian Saltzman
Jeremiah Morgan
Jimin Patel
Changwei Hu
Abinand Kishore
Phani Teja Anumanchupallik

Assignees

AMAZON TECHNOLOGIES, INC.

Dates

Publication Date: 20260505
Application Date: 20240923

Claims (20)

1 . A computer-implemented method comprising: receiving, by a conversational agent service, a search query from a user; selecting, by the conversational agent service, a supplier from a plurality of suppliers based on a relevance of the supplier to the search query; determining, by the conversational agent service, a plurality of aspects for the supplier based on a set of user reviews; generating, in real time by one or more machine learning models of the conversational agent service, one or more contextually relevant aspects from the plurality of aspects based on the search query; generating, in real time by the one or more machine learning models of the conversational agent service, a supplier related question and corresponding answer based on the one or more contextually relevant aspects; and presenting, by the conversational agent service, the supplier related question in real time and the corresponding answer to the user.
2 . The computer-implemented method of claim 1 , wherein the selecting the supplier comprises: determining a corresponding predicted click-through value based on the search query for the plurality of suppliers; and determining the supplier based on the corresponding predicted click-through values.
3 . The computer-implemented method of claim 1 , wherein the presenting comprises: presenting the supplier related question to the user; and presenting the corresponding answer in response to an indication received from the user.
4 . A computer-implemented method comprising: receiving an indication from a user; determining a plurality of corresponding aspects for a plurality of suppliers based on a set of user based content; generating, by one or more machine learning models, one or more contextually relevant aspects from the plurality of corresponding aspects based on the indication for individual ones of the plurality of suppliers; generating, by the one or more machine learning models, a corresponding supplier related question for the individual ones of the plurality of suppliers based on the one or more contextually relevant aspects; selecting a supplier related question from the corresponding supplier related questions; generating, by the one or more machine learning models, a corresponding answer to the supplier related question based on one or more of the set of user based content; and causing the supplier related question and the corresponding answer to be presented the user.
5 . The computer-implemented method of claim 4 , wherein the supplier related question is for a product of the supplier.
6 . The computer-implemented method of claim 4 , wherein the generating, by the one or more machine learning models, the corresponding answer to the supplier related question is also based on an input of supplier guidelines into the one or more machine learning models.
7 . The computer-implemented method of claim 4 , wherein the selecting the supplier related question comprises: determining a corresponding predicted click-through value for the indication for the corresponding supplier related questions; and determining the supplier related question based on the corresponding predicted click-through values.
8 . The computer-implemented method of claim 4 , wherein the indication is a search query from the user.
9 . The computer-implemented method of claim 4 , wherein the indication is an access of an online shopping page for a product.
10 . The computer-implemented method of claim 4 , further comprising updating a cached version of the set of user based content in response to receiving the indication.
11 . The computer-implemented method of claim 4 , wherein the causing the supplier related question and the corresponding answer to be presented the user comprises: causing the supplier related question to be presented to the user; and causing the corresponding answer to be presented to the user in response to another indication received from the user.
12 . The computer-implemented method of claim 4 , further comprising causing one or more other questions generated by a conversational agent service to be presented to the user with the supplier related question.
13 . The computer-implemented method of claim 12 , further comprising ranking, by the one or more machine learning models, the supplier related question and the one or more other questions, and causing the supplier related question and the one or more other questions to be presented based on the ranking.
14 . The computer-implemented method of claim 4 , further comprising determining the one or more of the set of user based content, by the one or more machine learning models, based on the one or more contextually relevant aspects.
15 . A non-transitory computer-readable medium storing code that, when executed by a device, causes the device to perform a method comprising: in response to receiving an indication from a user, determining a plurality of corresponding aspects for a plurality of suppliers based on a set of user based content; generating, by one or more machine learning models, one or more contextually relevant aspects from the plurality of corresponding aspects based on the indication for individual ones of the plurality of suppliers; generating, by the one or more machine learning models, a corresponding supplier related question for the individual ones of the plurality of suppliers based on the one or more contextually relevant aspects; selecting a supplier related question from the corresponding supplier related questions; generating, by the one or more machine learning models, a corresponding answer to the supplier related question based on one or more of the set of user based content; and causing the supplier related question and the corresponding answer to be presented the user.
16 . The non-transitory computer-readable medium of claim 15 , wherein the selecting the supplier related question comprises: determining a corresponding predicted click-through value for the indication for the corresponding supplier related questions; and determining the supplier related question based on the corresponding predicted click-through values.
17 . The non-transitory computer-readable medium of claim 15 , wherein the method further comprises updating a cached version of the set of user based content in response to receiving the indication.
18 . The non-transitory computer-readable medium of claim 15 , wherein the causing the supplier related question and the corresponding answer to be presented the user comprises: causing the supplier related question to be presented to the user; and causing the corresponding answer to be presented to the user in response to another indication received from the user.
19 . The non-transitory computer-readable medium of claim 15 , wherein the method further comprises causing one or more other questions generated by a conversational agent service to be presented to the user with the supplier related question.
20 . The non-transitory computer-readable medium of claim 19 , wherein the method further comprises ranking, by the one or more machine learning models, the supplier related question and the one or more other questions, and causing the supplier related question and the one or more other questions to be presented based on the ranking.

Description

BACKGROUND Suppliers (such as businesses, media distribution services, etc.) can employ one or more data centers to deliver content (such as web sites, web content, or other digital data) to users or clients. In certain examples, a conversational agent, e.g., a “chatbot” or “virtual assistant”, engages in a natural language conversation with a computer user. Certain conversational agents are designed to provide information, answer questions, perform tasks, and/or assist with various activities in response to queries or requests received from such users. Conversational agents may be found in a wide range of applications and systems, and are capable of understanding and interpreting human languages, as well as generating responses in a human-like manner, using text or speech that is rational, easy to understand, and contextually relevant. Moreover, unlike humans, conversational agents may operate at any time of day or on any day of the year, and provide responses to users of any computer-based systems. Certain suppliers (e.g., of products and/or services) may desire to utilize a conversational agent to connect with its user. However, certain customers (e.g., users) may not articulate the questions or expend the effort to find a supplier (e.g., the supplier's brands and/or products) that is right for them. The manual process of gathering and synthesizing information from social media, reviews, videos, and in-store visits is time-consuming and often leaves customers uncertain about the relevance and applicability of the information to their specific needs and preferences. Certain customers seek deeper insights to discover products, learn about key attributes, and find a supplier that aligns with their preferences. BRIEF DESCRIPTION OF DRAWINGS Various examples in accordance with the present disclosure will be described with reference to the following drawings. FIG. 1 is a diagram illustrating a conversational agent application of a user (e.g., client) device and a conversational agent service/system generating a question and answer pair using machine learning according to some examples. FIG. 2 is a diagram illustrating an environment including a provider network (including a conversational agent service/system) communicatively coupled to a user (e.g., client) device (including a conversational agent application) according to some examples. FIG. 3 is a flow diagram illustrating operations of a method of utilizing machine learning at the granularity of candidate questions to generate a question and a corresponding answer according to some examples. FIG. 4 is a flow diagram illustrating operations of a method of utilizing machine learning at the granularity of candidate suppliers to generate a question and a corresponding answer according to some examples. FIG. 5 is a diagram illustrating an architecture of a machine learning model for review retrieval according to some examples. FIG. 6 is a diagram illustrating a graphical user interface (GUI) presenting a question generated by a machine learning model and presenting the response to that question according to some examples. FIG. 7 is a diagram illustrating an environment for creating, training, and using one or more machine learning models according to some examples. FIG. 8 is a flow diagram illustrating operations of a method of utilizing a search indication to create a question and answer pair by one or more machine learning models according to some examples. FIG. 9 illustrates an example provider network environment according to some examples. FIG. 10 is a block diagram of an example provider network that provides a storage service and a hardware virtualization service to customers according to some examples. FIG. 11 is a block diagram illustrating an example computer system that may be used in some examples. FIG. 12 illustrates a logical arrangement of a set of general components of an exemplary computing device that can be utilized in accordance with various examples. FIG. 13 illustrates an example of an environment for implementing aspects in accordance with various examples. DETAILED DESCRIPTION The present disclosure relates to methods, apparatus, systems, and non-transitory computer-readable storage media for utilizing a machine learning model to generate a question and answer pair for a conversational agent. Certain examples herein are directed to a computer-implemented service that utilizes machine learning at the granularity of candidate questions to generate a question and a corresponding answer. Other examples herein utilize machine learning at the granularity of candidate suppliers (e.g., a supplier's brand(s) and/or product(s)\service(s)) to generate a question and a corresponding answer. In certain examples, a brand is the name of a product and/or service. To overcome the technical problems discussed above, certain examples herein utilize machine learning for a conversational agent to generate a relevant question and answer pair (e.g., a branded