Method, electronic device, and computer-readable storage medium for determining answer to question of product
A method for determining an answer to a question of a product includes: receiving a question for a target product; obtaining a knowledge graph constructed in advance for the target product, in which, the knowledge graph includes entities and a relationship among the entities, and the entities at least include one question for the target product and a corresponding answer; and determining an answer corresponding to the question based on the relationship among the entities included in the knowledge graph.
1. A method for determining an answer to a question of a product, comprising:
receiving a question for a target product;
obtaining a knowledge graph constructed in advance for the target product, wherein the knowledge graph comprises entities and a relationship among the entities, and the entities at least comprise one question for the target product and a corresponding answer;
determining an answer corresponding to the question based on the relationship among the entities comprised in the knowledge graph;
wherein the knowledge graph is constructed in advance by:
obtaining a sample question and a sample answer for the target product;
performing semantic understanding on the sample question and the sample answer by converting a statement with an irregular syntax into a statement with a regular syntax including a statement in a subject-predicate-object (SPO) form;
determining a correlation between the sample question and the sample answer subjected to the semantic understanding; and
generating the knowledge graph by taking a degree of the correlation as a length of an edge and taking the sample question and the sample answer as the entities;
wherein the method further comprises outputting the answer through at least one of an audio signal and a visual signal;
wherein the method is implemented in an after-sales support computer system comprising:
one or more processing circuits configured to implement operations of the method;
a display screen configured to display the visual signal for outputting the answer;
a speaker configured to output the audio signal for outputting the answer;
wherein
the question is a sample question and comprises at least one of:
a question for a structural damage of the target product;
a question for reporting an error when the target product is in use; and
a question that needs guidance on how to use a certain function of the target product;
the question is obtained through after-sales personnel receiving at least one of a voice call from a user of the target product, and a forum;
the answer is a sample answer;
the one or more processing circuits are further configured to
summarize the sample question and the sample answer subjected to the semantic understanding as SPO language fragments represented as word vectors {S, P, O},
calculate an angle and/or distance between a question word vector corresponding to the sample question and an answer word vector corresponding to the sample answer;
determine a correlation between every sample question and every sample answer based on a calculation result, with smaller angle between the question word vector and the answer word vector and smaller distance between the question word vector and the answer word vector representing greater correlation;
wherein
the correlation is represented by a three-component group {Q, R, s}, where Q represents the sample question, R represents the sample answer, and s represents a correlation between the sample question and the sample answer; and
based on a value of s, the correlation between the sample question and the sample answer is classified as a strong correlation when s is greater than a first preset value, a weak correlation when s is smaller than the first preset value and greater than a second preset value, and irrelevant when s is smaller than the second preset value.
2. The method of claim 1 , wherein the determining the correlation between the sample question and the sample answer subjected to the semantic understanding comprises:
determining the correlation between the sample question and the sample answer subjected to the semantic understanding based on the angle and/or distance.
3. The method of claim 1 , wherein the determining the correlation between the sample question and the sample answer subjected to the semantic understanding comprises:
obtaining quality information of the sample answer and/or information of a user providing the sample answer;
determining a first correlation between the sample question and the sample answer subjected to the semantic understanding based on the quality information and/or information of the user;
determining a second correlation between the sample question and the sample answer subjected to the semantic understanding based on an angle and/or distance between a question word vector corresponding to the sample question and an answer word vector corresponding to the sample answer; and
determining the correlation between the sample question and the sample answer subjected to the semantic understanding based on the first correlation and the second correlation.
4. The method of claim 1 , wherein the sample question comprises a question received from a public media for the target product, and/or a question received after sales for the target product; and
the sample answer comprises an answer corresponding to the question received from the public media for the target product, and/or an answer corresponding to the question received after sales for the target product.
5. The method of claim 1 , wherein, in response to determining a plurality of answers, the method further comprises:
determining a plurality of correlations between the plurality of answers and the question based on the knowledge graph;
sorting the plurality of answers based on a descending order of the plurality of correlations; and
outputting the plurality of answers sorted.
6. An after-sales support computer system, comprising:
a processor; and
memory storing instructions for execution by the processor;
wherein the processor is configured to:
receive a question for a target product;
obtain a knowledge graph constructed in advance for the target product, wherein the knowledge graph comprises entities and a relationship among the entities, and the entities at least comprise one question for the target product and a corresponding answer;
determine an answer corresponding to the question based on the relationship among the entities comprised in the knowledge graph;
wherein the processor is further configured to construct the knowledge graph in advance by:
obtaining a sample question and a sample answer for the target product;
performing semantic understanding on the sample question and the sample answer by converting a statement with an irregular syntax into a statement with a regular syntax including a statement in a subject-predicate-object (SPO) form;
determining a correlation between the sample question and the sample answer subjected to the semantic understanding; and
generating the knowledge graph by taking a degree of the correlation as a length of an edge and taking the sample question and the sample answer as the entities;
wherein the after-sales support computer system is configured to output answer through at least one of an audio signal and a visual signal, and comprises:
one or more processing circuits configured to implement operations of the method;
a display screen configured to display the visual signal for outputting the answer;
a speaker configured to output the audio signal for outputting the answer;
wherein
the question is a sample question and comprises at least one of:
a question for a structural damage of the target product;
a question for reporting an error when the target product is in use; and
a question that needs guidance on how to use a certain function of the target product;
the question is obtained through after-sales personnel receiving at least one of a voice call from a user of the target product, and a forum;
the answer is a sample answer;
the one or more processing circuits are further configured to
summarize the sample question and the sample answer subjected to the semantic understanding as SPO language fragments represented as word vectors {S, P, O},
calculate an angle and/or distance between a question word vector corresponding to the sample question and an answer word vector corresponding to the sample answer;
determine a correlation between every sample question and every sample answer based on a calculation result, with smaller angle between the question word vector and the answer word vector and smaller distance between the question word vector and the answer word vector representing greater correlation;
wherein
the correlation is represented by a three-component group {Q, R, s}, where Q represents the sample question, R represents the sample answer, and s represents a correlation between the sample question and the sample answer; and
based on a value of s, the correlation between the sample question and the sample answer is classified as a strong correlation when s is greater than a first preset value, a weak correlation when s is smaller than the first preset value and greater than a second preset value, and irrelevant when s is smaller than the second preset value.
7. The electronic device of claim 6 , wherein the determining the correlation between the sample question and the sample answer subjected to the semantic understanding comprises:
determining the correlation between the sample question and the sample answer subjected to the semantic understanding based on the angle and/or distance.
8. The electronic device of claim 6 , wherein the determining the correlation between the sample question and the sample answer subjected to the semantic understanding comprises:
obtaining quality information of the sample answer and/or information of a user providing the sample answer;
determining a first correlation between the sample question and the sample answer subjected to the semantic understanding based on the quality information and/or information of the user;
determining a second correlation between the sample question and the sample answer subjected to the semantic understanding based on an angle and/or distance between a question word vector corresponding to the sample question and an answer word vector corresponding to the sample answer; and
determining the correlation between the sample question and the sample answer subjected to the semantic understanding based on the first correlation and the second correlation.
9. The electronic device of claim 6 , wherein the sample question comprises a question received from a public media for the target product, and/or a question received after sales for the target product; and
the sample answer comprises an answer corresponding to the question received from the public media for the target product, and/or an answer corresponding to the question received after sales for the target product.
10. The electronic device of claim 6 , wherein in response to determining a plurality of answers, the processor is further configured to:
determine a plurality of correlations between the plurality of answers and the question based on the knowledge graph;
sort the plurality of answers based on a descending order of the plurality of correlations; and
output the plurality of answers sorted.
11. A non-transitory computer-readable storage medium having stored therein instructions that, when executed by a processor of an electronic device, causes the electronic device to perform a method for determining an answer to a question of a product, the method comprising:
receiving a question for a target product;
obtaining a knowledge graph constructed in advance for the target product, wherein the knowledge graph comprises entities and a relationship among the entities, and the entities at least comprise one question for the target product and a corresponding answer;
determining an answer corresponding to the question based on the relationship among the entities comprised in the knowledge graph;
wherein the method further comprises constructing the knowledge graph in advance by:
obtaining a sample question and a sample answer for the target product;
performing semantic understanding on the sample question and the sample answer by converting a statement with an irregular syntax into a statement with a regular syntax including a statement in a subject-predicate-object (SPO) form;
determining a correlation between the sample question and the sample answer subjected to the semantic understanding; and
generating the knowledge graph by taking a degree of the correlation as a length of an edge and taking the sample question and the sample answer as the entities;
wherein the method further comprises outputting the answer through at least one of an audio signal and a visual signal;
wherein the method is implemented in an after-sales support computer system comprising:
one or more processing circuits configured to implement operations of the method;
a display screen configured to display the visual signal for outputting the answer;
a speaker configured to output the audio signal for outputting the answer;
wherein
the question is a sample question and comprises at least one of:
a question for a structural damage of the target product;
a question for reporting an error when the target product is in use; and
a question that needs guidance on how to use a certain function of the target product;
the question is obtained through after-sales personnel receiving at least one of a voice call from a user of the target product, and a forum;
the answer is a sample answer;
the one or more processing circuits are further configured to
summarize the sample question and the sample answer subjected to the semantic understanding as SPO language fragments represented as word vectors {S, P, O},
calculate an angle and/or distance between a question word vector corresponding to the sample question and an answer word vector corresponding to the sample answer;
determine a correlation between every sample question and every sample answer based on a calculation result, with smaller angle between the question word vector and the answer word vector and smaller distance between the question word vector and the answer word vector representing greater correlation;
wherein
the correlation is represented by a three-component group {Q, R, s}, where Q represents the sample question, R represents the sample answer, and s represents a correlation between the sample question and the sample answer; and
based on a value of s, the correlation between the sample question and the sample answer is classified as a strong correlation when s is greater than a first preset value, a weak correlation when s is smaller than the first preset value and greater than a second preset value, and irrelevant when s is smaller than the second preset value.
12. The non-transitory computer-readable storage medium of claim 11 , wherein the determining the correlation between the sample question and the sample answer subjected to the semantic understanding comprises:
determining the correlation between the sample question and the sample answer subjected to the semantic understanding based on the angle and/or distance.
13. The non-transitory computer-readable storage medium of claim 11 , wherein the determining the correlation between the sample question and the sample answer subjected to the semantic understanding comprises:
obtaining quality information of the sample answer and/or information of a user providing the sample answer;
determining a first correlation between the sample question and the sample answer subjected to the semantic understanding based on the quality information and/or information of the user;
determining a second correlation between the sample question and the sample answer subjected to the semantic understanding based on an angle and/or distance between a question word vector corresponding to the sample question and an answer word vector corresponding to the sample answer; and
determining the correlation between the sample question and the sample answer subjected to the semantic understanding based on the first correlation and the second correlation.
14. The non-transitory computer-readable storage medium of claim 11 , wherein the sample question comprises a question received from a public media for the target product, and/or a question received after sales for the target product; and
the sample answer comprises an answer corresponding to the question received from the public media for the target product, and/or an answer corresponding to the question received after sales for the target product.
15. The non-transitory computer-readable storage medium of claim 11 , wherein, in response to determining a plurality of answers, the method further comprises:
determining a plurality of correlations between the plurality of answers and the question based on the knowledge graph;
sorting the plurality of answers based on a descending order of the plurality of correlations; and
outputting the sorted plurality of answers.