Human-computer interaction methods for refrigerators

By combining the current question, dialogue content, and domain information to filter answers, the problem of incorrect responses caused by the lack of context utilization in existing technologies is solved, resulting in more accurate and relevant answer output and improving user experience.

CN117555998BActive Publication Date: 2026-11-17QINDAO HAIER REFRIGERATOR CO LTD +1
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210927342.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-03
Publication Date
2026-11-17
Estimated Expiration
2042-08-03

AI Technical Summary

Technical Problem

Existing retrieval methods fail to effectively utilize the contextual information of dialogue in human-computer interaction, leading to an increased possibility of incorrect responses and reducing the robustness of the dialogue system.

Method used

When filtering answers, the system not only utilizes the current question but also combines the previous dialogue content and the domain of the question. Through slot information, sentence structure matching, and weight assignment, multiple answer sets are formed, and the answer with the highest weight is finally output.

Benefits of technology

It improved the relevance and accuracy of answers, reduced the likelihood of incorrect responses, and enhanced the user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117555998B_ABST
    Figure CN117555998B_ABST
Patent Text Reader

Abstract

This invention provides a human-computer interaction method for a refrigerator, comprising: obtaining a first set of answers based on the current question; assigning a weight xN to the answers in the first set of answers, where N is the standard value of the weight; obtaining an article by combining the questions and output answers from y rounds of dialogue; obtaining a second set of answers based on the nouns in the article; assigning a weight mN to the answers based on the frequency m of each answer in the second set; obtaining a third set of answers within the domain of the current question; assigning a weight xN to the answers in the third set; obtaining the union of the first, second, and third set of answers; summing the weights of identical answers; and outputting the answer with the highest weight as the target answer; wherein x, y, and m are all integers greater than zero, and y ≤ x. This human-computer interaction method is highly relevant to the above and can obtain more accurate answers.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to a human-computer interaction method for a refrigerator. Background Technology

[0002] Current retrieval methods mostly rely on user input to retrieve answers without considering the context of the dialogue. This increases the likelihood of incorrect responses and significantly reduces the robustness of the dialogue system. Improving the quality of responses and enhancing user experience is a topic worthy of serious research. Summary of the Invention

[0003] The purpose of this invention is to provide a human-computer interaction method for a refrigerator. When filtering answers to questions, the method not only utilizes the current question but also the dialogue content preceding the current question to search for answers. This makes the answers more relevant to the preceding text, reduces the possibility of generating incorrect responses, and improves the quality of the response statements.

[0004] To achieve one of the above-mentioned objectives, one embodiment of the present invention provides a human-computer interaction method for a refrigerator, comprising:

[0005] Obtain the current question, filter the answers based on the current question to obtain the first set of answers, and assign a weight xN to the answers in the first set of answers, where N is the standard value of the weight;

[0006] The current problem is in the y-th round of dialogue. We obtain the questions and answers from the y-th round of dialogue and form an article. We filter the words in the article to obtain a second set of answers and assign weights mN to the corresponding answers based on the number of times the answers appear in the second set.

[0007] Obtain answers within the domain of the current question to form a third set of answers, and assign a weight xN to the answers in the third set of answers;

[0008] Obtain the union of the first set of answers, the second set of answers, and the third set of answers. Then, sum the weights of the answers with the same weights and output the answer with the highest weight as the target answer.

[0009] Where x, y, and m are all integers greater than zero, and y ≤ x.

[0010] As a further improvement to one embodiment of the present invention, "obtaining the current question and filtering to obtain a first set of answers based on the current question" specifically includes:

[0011] Extract the slot information of each word in the current question, and determine the sentence structure of the current question based on the slot information of each word and the order of the slot information;

[0012] The first question set is formed by selecting questions from the question bank that have the same sentence structure as the current question.

[0013] Match the questions in the first question set with the words in the slot corresponding to the current question, and filter out the top 50 questions with the best matching values ​​to form the second question set;

[0014] The answers to the questions in the second set of questions are selected from the answer database to form the first set of answers.

[0015] As a further improvement to one embodiment of the present invention, "matching the questions in the first set of questions with the words in the slots corresponding to the current question, and filtering out the questions with the top 50 matching values" specifically includes:

[0016] Extract the words from each slot of the questions in the first set of questions and the words from each slot of the current question. Match the words in the same slot one by one to get the matching value. Add the matching values ​​of the words in each slot of the questions in the first set of questions and select the top 50 questions with the best matching values.

[0017] As a further improvement to one embodiment of the present invention, "obtaining a second set of answers based on the nouns in the article" specifically includes:

[0018] Extract the components, slot information, and relationships of nouns in the sentences from the article, match them with answers in the answer database, and select the top 50 matching answers to form a second answer set.

[0019] As a further improvement to one embodiment of the present invention, "obtaining answers within the domain of the current question to form a third set of answers" specifically includes:

[0020] Extract the slot information of each word in the current question, and determine the sentence structure of the current question based on the slot information of each word and the order of the slot information;

[0021] Match the sentence structure of the current question with the preset sentence structures in each domain, obtain the domain to which the current question belongs according to the priority order of the preset domains, and combine all the answers in the domain to which the current question belongs to form the third answer set.

[0022] As a further improvement to one embodiment of the present invention, questions from each round of dialogue are added to a question bank, so that the content of the question bank increases as the dialogue progresses.

[0023] As a further improvement of one embodiment of the present invention, the relationship between x and y satisfies: x≤2y.

[0024] As a further improvement of one embodiment of the present invention, the time interval between each two rounds is ≤5 minutes.

[0025] As a further improvement of one embodiment of the present invention, the time interval between each two rounds is ≤2 minutes.

[0026] As a further improvement of one embodiment of the present invention, the answers in the first answer set, the second answer set, and the third answer set are all taken from a preset answer database.

[0027] One or more technical solutions provided by this invention have at least the following technical effects or advantages:

[0028] The human-computer interaction method provided by this invention filters out a first set of answers, a second set of answers, and a third set of answers based on the current question, the dialogue content preceding the current question, and the domain to which the current dialogue belongs. It then assigns weights to the answers in the first, second, and third sets of answers based on the influence of the dialogue timing on the answers. Answers in the first and third sets that are directly related to the current question are given higher weights, while answers in the second set, filtered from the preceding context, are indirectly related to the current question and are given lower weights. This results in more reasonable answers that are better relevant to the preceding context. Attached Figure Description

[0029] Figure 1 This is a flowchart of the human-computer interaction method in an embodiment of the present invention.

[0030] Figure 2 This is a flowchart for obtaining the first set of answers in human-computer interaction methods.

[0031] Figure 3 This is a flowchart for obtaining the second set of answers in human-computer interaction methods.

[0032] Figure 4 This is a flowchart for obtaining the third set of answers in human-computer interaction methods. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Furthermore, it should be understood that although the terms "first," "second," etc., may be used herein to describe various objects, these described objects should not be limited by these terms. These terms are only used to distinguish these described objects from one another. For example, a first set of answers may be referred to as a second set of answers, and similarly, a second set of answers may be referred to as a first set of answers, without departing from the scope of protection of this application.

[0035] This invention provides a human-computer interaction method for a refrigerator, such as... Figure 1 As shown, it includes the following steps:

[0036] Obtain the current question, filter the answers based on the current question to obtain the first set of answers, and assign a weight xN to the answers in the first set of answers, where N is the standard value of the weight;

[0037] The current problem is in the y-th round of dialogue. We obtain the questions and answers from the y-th round of dialogue and form an article. We filter the words in the article to obtain a second set of answers and assign weights mN to the corresponding answers based on the number of times the answers appear in the second set.

[0038] Obtain answers within the domain of the current question to form a third set of answers, and assign a weight xN to the answers in the third set of answers.

[0039] Obtain the union of the first set of answers, the second set of answers, and the third set of answers. Then, sum the weights of the answers with the same weights and output the answer with the highest weight as the target answer.

[0040] Where x, y, and m are all integers greater than zero, and y ≤ x. Figure 1 In the example, the current problem is the 10th round, and we take x = y = 10.

[0041] Specifically, such as Figure 2 As shown, "obtaining the current question and filtering it to get the first set of answers" specifically includes the following steps:

[0042] Extract the slot information of each word in the current question, and determine the sentence structure of the current question based on the slot information of each word and the order of the slot information;

[0043] The first question set is formed by selecting questions from the question bank that have the same sentence structure as the current question.

[0044] The first set of questions is matched with the words in the slot corresponding to the current question. The top 50 questions with the best matching values ​​are selected as the second set of questions. The matching process specifically includes:

[0045] Extract the words from each slot of the questions in the first set of questions and the words from each slot of the current question. Match the words in the same slot one by one to get the matching value. Add the matching values ​​of the words in each slot of the questions in the first set of questions and select the top 50 questions with the best matching values.

[0046] The answers to the questions in the second set of questions are selected from the answer database to form the first set of answers.

[0047] Each word in a sentence corresponds to a slot. Slot information is obtained based on the part of speech of the word in the slot and the function of the word in the sentence. Then, slot names that match the slot information are selected from a preset set of slot names. These slot names are arranged in order to obtain the sentence structure.

[0048] For example, can milk and durian be eaten together? The words "milk," "and," "durian," "can," "together," "eat," "ma," and "?" correspond to their respective slots. The slots for "milk" and "durian" are named 'FD_slots', the slot for "and" is named 'verb_and', the slot for "can" is named 'verb_can', the slot for "together" is named 'eating_way', the slot for "eat" is named 'verb_eat', the slot for "ma" is named 'yn_form', and the slot for "?" is named '?'.

[0049] The sentence structure for "Can milk and durian be eaten together?" is: 'FD_slots', 'verb_and', FD_slots', 'verb_can', 'eating_way', 'verb_eat', 'yn_form', '?'.

[0050] How can pork be cooked? The words "pork," "can," "how," "make," and "?" each correspond to their respective slots. The slot for "pork" is named 'FD_slots', the slot for "can" is named 'verb_can', the slot for "how" is named 'wh_how', the slot for "make" is named 'yn_cook', and the slot for "?" is named '?'.

[0051] The sentence structure for "How can pork be cooked?" is: 'FD_slots', 'verb_can', 'wh_how', 'yn_cook', '?'.

[0052] Questions with the same sentence structure have the highest similarity. Therefore, once the sentence structure of the current question is determined, questions with the same sentence structure can be found in the question bank based on the sentence structure.

[0053] However, questions with the same sentence structure can vary greatly due to the influence of nouns, verbs, locative words, and other auxiliary words. Therefore, it is necessary to perform similarity matching between the words in each slot of the questions in the first question set and the words in the corresponding slot of the current question. After matching each word in each slot, the matching values ​​of each word in the questions in the first question set are added together, and the top 50 questions with the highest matching values ​​are taken as the second question set.

[0054] Each question in the question bank has a unique corresponding answer in the answer bank. The answers to the top 50 questions with the highest matching values ​​are combined to form the first answer set.

[0055] Preferably, after each round of dialogue, the questions from the dialogue are added to the question bank, so that the content of the question bank increases as the dialogue progresses.

[0056] The first set of answers is filtered based on the current question, and has a high degree of accuracy. Therefore, the answers in the first set of answers are given a higher weight. The weight coefficient x changes according to the number of rounds in the current dialogue, and always remains no less than the number of rounds in the current dialogue.

[0057] Furthermore, such as Figure 3 As shown, the aforementioned "obtaining the second set of answers by filtering the nouns in the article" specifically includes:

[0058] Extract the components, slot information, and relationships of nouns in the sentences from the article, match them with answers in the answer database, and select the top 50 matching answers to form a second answer set.

[0059] The context of a sentence varies depending on the noun's grammatical function, slot information, and relationship with other words. Matching these elements with answers in the answer database yields a large number of results. Since the second answer set corresponds to the current question and previous rounds of dialogue, it contains a significant amount of content. Therefore, more answers need to be selected to form the second answer set. Duplicate answers may appear in the second answer set. The weights are calculated by multiplying the standard weight N (the number of times an answer appears, m) by the number of repetitions, resulting in weight mN.

[0060] like Figure 4 As shown, "obtaining answers within the domain of the current question to form a third set of answers" specifically includes:

[0061] Extract the slot information of each word in the current question, and determine the sentence structure of the current question based on the slot information of each word and the order of the slot information;

[0062] Match the sentence structure of the current question with the preset sentence structures in each domain, obtain the domain to which the current question belongs according to the priority order of the preset domains, and combine all the answers in the domain to which the current question belongs to form the third answer set.

[0063] Each domain has its own preset sentence structure. If the sentence structure of the current question exists only in one domain, then that domain is the domain to which the current question belongs. If the sentence structure of the current question appears in multiple domains, then the first domain with the sentence structure of the current question is selected as the domain to which the current question belongs, according to the preset domain priority order.

[0064] Furthermore, the relationship between x and y also satisfies: x ≤ 2y. Although the weight coefficient x of the answers in the first and second answer sets satisfies x ≥ y, the weight coefficient x still has an upper limit, which is twice the round number y of the current question. This is to prevent the weights of the answers in the first and third answer sets from being too high, causing the weights of the answers in the second answer set to be completely incomparable to the weights of the answers in the first and third answer sets, thus losing the relevance to the preceding text.

[0065] Furthermore, the time interval between each two rounds should be ≤5 minutes. If the time interval between two rounds of dialogue is too long, the topic and content of the dialogue may have changed. At this time, the answer obtained by referencing the previous topic will have a large error.

[0066] Furthermore, the time interval between each two rounds is ≤2 minutes, further reducing the error in generating the answer.

[0067] Furthermore, the answers in the first, second, and third answer sets are all taken from a pre-defined answer database.

[0068] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0069] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1.A method of human-machine interaction of a refrigerator, characterized by, include: Obtain the current question and filter it to get the first set of answers, including: extracting the slot information of each word in the current question, and determining the sentence structure of the current question based on the slot information of each word and the order of the slot information; The first question set is formed by selecting questions from the question bank that have the same sentence structure as the current question. Match the questions in the first question set with the words in the slot corresponding to the current question, and filter out the top 50 questions with the best matching values ​​to form the second question set; The answers to the questions in the second set of questions are selected from the answer database to form the first set of answers; Assign weights xN to the answers in the first set of answers, where N is the standard value of the weights; The current problem is in the y-th round of dialogue. We obtain the questions and answers from the y-th round of dialogue and form an article. We filter the words in the article to obtain a second set of answers and assign weights mN to the corresponding answers based on the number of times the answers appear in the second set. Obtain answers within the domain of the current question to form a third set of answers, and assign a weight xN to the answers in the third set of answers; Obtain the union of the first set of answers, the second set of answers, and the third set of answers. Then, sum the weights of the answers with the same weights and output the answer with the highest weight as the target answer. Where x, y, and m are all integers greater than zero, and y ≤ x. 2.The refrigerator's human-machine interaction method of claim 1, characterized in that, "Matching the questions in the first set of questions with the words in the slot corresponding to the current question, and filtering out the top 50 questions with the best matching values" specifically includes: Extract the words from each slot of the questions in the first set of questions and the words from each slot of the current question. Match the words in the same slot one by one to get the matching value. Add the matching values ​​of the words in each slot of the questions in the first set of questions and select the top 50 questions with the best matching values. 3.The refrigerator's human-machine interaction method of claim 1, characterized in that, "Obtaining a second set of answers by filtering the nouns in the article" specifically includes: Extract the components, slot information, and relationships of nouns in the sentences from the article, match them with answers in the answer database, and select the top 50 matching answers to form a second answer set. 4.The refrigerator's human-machine interaction method of claim 1, characterized in that, "Obtaining answers within the domain of the current question to form a third set of answers" specifically includes: Extract the slot information of each word in the current question, and determine the sentence structure of the current question based on the slot information of each word and the order of the slot information; Match the sentence structure of the current question with the preset sentence structures in each domain, obtain the domain to which the current question belongs according to the priority order of the preset domains, and combine all the answers in the domain to which the current question belongs to form the third answer set. 5.The refrigerator's human-machine interaction method of claim 1, wherein, Add questions from each round of dialogue to a question bank, so that the content of the question bank increases as the dialogue progresses. 6.The refrigerator's human-machine interaction method of claim 1, wherein, The relationship between x and y satisfies: x ≤ 2y. 7.The refrigerator's human-machine interaction method of claim 1, wherein, The time interval between each two rounds should be ≤5 minutes. 8.The refrigerator's human-machine interaction method of claim 1, characterized in that, The time interval between each two rounds should be ≤2 minutes. 9.The refrigerator human-machine interaction method according to any one of claims 1-8, characterized in that, The answers in the first, second, and third answer sets are all taken from a pre-defined answer database.

Citation Information

Patent Citations

  • Deep question-and-answer service providing method and device based on artificial intelligence

    CN105159996A

  • Response statement generation method and device, computer equipment and storage medium

    CN114116998A