Method for verifying dominant knowledge, electronic equipment and storage medium
By converting implicit knowledge in the language model into explicit knowledge and using the model knowledge base for data comparison, the problems of low efficiency and poor reliability of explicit knowledge verification in the existing technology are solved, and efficient and accurate display is achieved Sexual knowledge verification.
Patent Information
- Application Number
- CN202311262667.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-27
- Publication Date
- 2025-02-11
AI Technical Summary
The prior art faces complexity and verification problems when constructing and verifying explicit knowledge. Traditional manual verification methods are costly, inefficient and prone to omissions and errors.
By converting implicit knowledge in the language model into explicit knowledge and using the model knowledge base for data comparison, we verify whether the conclusion fields of explicit knowledge match the data content, thereby determining the correctness and consistency of explicit knowledge.
It improves the efficiency and accuracy of explicit knowledge verification, reduces labor and time costs, and improves the reliability of verification results.
Smart Images
Figure CN119415666A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure generally relates to the field of natural language processing technology. More specifically, the present disclosure relates to a method, electronic device and storage medium for verifying explicit knowledge. Background Art
[0002] In recent years, language models have been widely used in various fields and have shown amazing capabilities and stability. Especially in various natural language processing tasks, language models can acquire and use this knowledge for reasoning and decision-making by learning a large amount of data.
[0003] In order to further enhance the reasoning ability of language models, the existing technology adopts a Chain-of-Thought (CoT) technology, the core idea of which is to guide the model to solve problems in a step-by-step reasoning manner. In the specific implementation process, the model first starts the reasoning process by generating an initial answer, and then gradually improves and corrects this answer until the final solution is obtained. The Chain-of-Thought technology provides a structured way to guide the model's reasoning process, making it more accurate and reliable when solving complex problems.
[0004] However, the knowledge used in the above reasoning process exists in an invisible form inside the model, which is not easy for people to understand and verify. Therefore, in order to make the knowledge of the language model reason in a way that people can understand, it is necessary to transform this knowledge into the form of explicit knowledge, which is conducive to improving the interpretability and reliability of the model.
[0005] At present, the construction of explicit knowledge faces two major problems: complexity and verification. Verification means that the constructed explicit knowledge needs to be verified for its correctness and consistency, which is the key to ensuring the reliability of the model. However, due to the huge amount of knowledge contained in the language model, traditional manual verification methods are often costly and inefficient, and are prone to omissions and errors.
[0006] In view of this, there is an urgent need to provide an explicit knowledge verification solution to replace the traditional manual verification method and complete the verification of explicit knowledge efficiently and accurately. Summary of the invention
[0007] In order to at least solve one or more technical problems mentioned above, the present disclosure proposes an explicit knowledge verification solution in multiple aspects.
[0008] In a first aspect, the present disclosure provides a method for verifying explicit knowledge, including: converting implicit knowledge in a language model into explicit knowledge; searching for data content reflecting the implicit knowledge from a model knowledge base based on premise fields of the explicit knowledge and their field values; determining whether the field value of a conclusion field of the explicit knowledge matches the data content; and in response to the conclusion field of the explicit knowledge matching the data content, determining that the explicit knowledge has passed verification.
[0009] In some embodiments, converting implicit knowledge in a language model into explicit knowledge includes: converting implicit knowledge in a language model into explicit knowledge in a textual expression form; and converting explicit knowledge in a textual expression form into explicit knowledge in a formal expression, wherein the explicit knowledge in a formal expression includes a premise field and a conclusion field.
[0010] In some embodiments, before searching for data content reflecting implicit knowledge from the model knowledge base based on the premise fields and field values of explicit knowledge, the method also includes: constructing an association relationship between the premise fields and the table headers in the model knowledge base; and constructing an association relationship between the conclusion fields and the table headers in the model knowledge base.
[0011] In some embodiments, searching for data content reflecting implicit knowledge from a model knowledge base based on premise fields of explicit knowledge and their field values includes: constructing a database query statement based on premise fields of explicit knowledge and their field values; and querying the model knowledge base based on the database query statement to obtain data content.
[0012] In some embodiments, determining whether the field value of the conclusion field of explicit knowledge matches the data content includes: determining whether the field value of the premise field of explicit knowledge matches the data content; and in response to the field value of the premise field of explicit knowledge matching the data content, determining whether the field value of the conclusion field of explicit knowledge matches the data content.
[0013] In some embodiments, after searching the model knowledge base for data content reflecting implicit knowledge, the method further includes: calculating the similarity between the field value of the conclusion field of explicit knowledge and the data content to obtain the probability of the conclusion in the explicit knowledge being true.
[0014] In some embodiments, converting explicit knowledge in textual expression form into explicit knowledge in formal expression includes: using prompt information of natural language to guide a language model so that it generates explicit knowledge in formal expression based on the explicit knowledge in textual expression form.
[0015] In some embodiments, before converting implicit knowledge in the language model into explicit knowledge, the method also includes: constructing an ontology and semantic rules, wherein the ontology includes operators and concepts; forming constituent elements of a model knowledge base based on the operators and concepts; and formally describing the elements based on the semantic rules to construct a model knowledge base of the language model.
[0016] In a second aspect, the present disclosure provides an electronic device comprising: a processor; and a memory storing executable program instructions, which, when executed by the processor, enables the device to implement any method according to the first aspect.
[0017] In a third aspect, the present disclosure provides a computer-readable storage medium having computer-readable instructions stored thereon, which, when executed by one or more processors, implement the method of any one of the first aspects.
[0018] Through the method for verifying explicit knowledge provided above, the disclosed embodiment searches for matching data content from the model knowledge base through the premise field of explicit knowledge and its field value. Compared with manually consulting professional field knowledge, this search method of the disclosed embodiment is more efficient and convenient. The disclosed embodiment further verifies the consistency of explicit knowledge and implicit knowledge and the correctness of the explicit knowledge based on the matching result of the field value of the conclusion field of explicit knowledge and the data content. Since the data content reflects implicit knowledge and provides a quantifiable basis for verification, the correctness and consistency verification of explicit knowledge can be completed only through data comparison operations. Compared with traditional manual verification schemes, the reliability of the verification results of this method is improved, and the manpower and time consumed are reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] By reading the detailed description below with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0020] Figure 1 An exemplary flow chart showing an explicit knowledge verification method of some embodiments of the present disclosure;
[0021] Figure 2 An exemplary flow chart showing the explicit knowledge verification method of other embodiments of the present disclosure;
[0022] Figure 3 An exemplary flow chart showing the explicit knowledge verification method of some other embodiments of the present disclosure;
[0023] Figure 4An exemplary flow chart showing an explicit knowledge verification method of some embodiments of the present disclosure;
[0024] Figure 5 An exemplary flow chart of a method for matching explicit knowledge and implicit knowledge according to some embodiments of the present disclosure is shown;
[0025] Figure 6 An exemplary flow chart showing an explicit knowledge verification method of some embodiments of the present disclosure;
[0026] Figure 7 An exemplary flow chart showing an explicit knowledge verification method of some embodiments of the present disclosure;
[0027] Figure 8 An exemplary structural block diagram of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION
[0028] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.
[0029] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.
[0030] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations.
[0031] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.
[0032] The specific implementation of the present disclosure is described in detail below with reference to the accompanying drawings.
[0033] Exemplary application scenarios
[0034] In various natural language processing tasks, language models can acquire and use knowledge to make inferences and decisions by learning from a large amount of data. However, the knowledge used by language models for reasoning exists in an invisible form within the model, which is not easy for people to understand and verify. Therefore, in order to make the knowledge of the language model reason in a way that people can understand, it is necessary to convert this knowledge into explicit knowledge, which is conducive to improving the interpretability and reliability of the model.
[0035] Existing technologies usually use manual methods to construct explicit knowledge, and language models usually contain a huge amount of knowledge. Manually extracting and organizing this knowledge requires professional domain knowledge and a large amount of human resources, and is prone to errors and omissions. Therefore, verifying the correctness and consistency of the constructed explicit knowledge is a crucial link.
[0036] However, given the huge amount of knowledge contained in language models, traditional manual verification methods are often costly and inefficient. Taking the scenario of converting the knowledge of language models into a set of rules as an example, the verification process requires manual verification of the applicability and correctness of each rule one by one, which is undoubtedly a huge workload. In addition, the manual verification process is affected by human subjective factors, and omissions and errors are prone to occur during verification, making it difficult to guarantee the reliability and accuracy of the verification results.
[0037] Exemplary Application Scenarios
[0038] In view of this, the disclosed embodiment provides an explicit knowledge verification scheme, which locates the data of the model knowledge base through the premise fields in the explicit knowledge, so as to efficiently find the data content that can reflect the corresponding implicit knowledge, and then compares the data through the conclusion fields of the explicit knowledge, thereby generating reliable verification results.
[0039] Before introducing the explicit knowledge verification method of the disclosed embodiment, the language model is first introduced. The current language models can show excellent performance in various natural language processing tasks. Specifically, these language models form a model knowledge base by learning a large amount of data, and can use this knowledge for reasoning and decision-making.
[0040] In the model knowledge base, knowledge is stored in the form of implicit knowledge. To facilitate understanding, the following example is provided to illustrate the difference between explicit knowledge and implicit knowledge. Explicit knowledge refers to knowledge expressed in a human-readable form, such as rules, facts, and concepts expressed in text. Assuming that the text expression of explicit knowledge is that the demand for catering has decreased and the operating income of the catering company has decreased, then the implicit knowledge corresponding to this knowledge in the model database can be discrete data such as the catering consumption expenditure data of residents in a certain period of time and the operating income data of the catering company.
[0041] It should be noted that the above description of the expression forms of explicit knowledge and implicit knowledge is merely an example provided to facilitate understanding by those skilled in the art and does not constitute a limitation of the present disclosure.
[0042] Based on the content described above, in order to solve the problems existing in the existing explicit knowledge verification methods, Figure 1 An exemplary flow chart of an explicit knowledge verification method 100 according to some embodiments of the present disclosure is shown.
[0043] like Figure 1 As shown, in step S101, the implicit knowledge in the language model is converted into explicit knowledge. In the embodiment of the present disclosure, the implicit knowledge is stored in the model knowledge base of the language model in the form of discrete data. In practical applications, some embodiments of the present disclosure can manually construct explicit knowledge by manually analyzing the data content in the model knowledge base. In other embodiments, the language model can also be guided to output explicit knowledge in the form of text expression with the help of natural language prompt information.
[0044] It should be noted that this disclosure does not have strict requirements on the method of generating explicit knowledge. In practical applications, any method that can convert implicit knowledge into explicit knowledge is applicable to this disclosure. Therefore, this disclosure does not impose too many restrictions on the method of generating explicit knowledge.
[0045] In step S102, data content reflecting implicit knowledge is searched from the model knowledge base according to the premise field and field value of explicit knowledge. In some embodiments, in order to facilitate people to understand and use explicit knowledge for reasoning, explicit knowledge can be divided into premise and conclusion.
[0046] To facilitate understanding, let's take the explicit knowledge that the demand for catering has decreased and the operating income of catering companies has decreased as an example. The decrease in demand for catering can be understood as the premise, so the field corresponding to the string is the premise field, and the decrease in the operating income of catering companies can be understood as the conclusion. Similarly, the field corresponding to the string is the conclusion field.
[0047] Given that the implicit knowledge corresponding to the explicit knowledge in the model database can be discretely distributed data such as residents' catering consumption expenditure data and catering company's operating income data within a certain period of time, and based on the correlation between catering demand and residents' catering consumption expenditure, the time period in which residents' catering consumption expenditure data shows a downward trend can be found in the model database, and then the catering company's operating income data within this time period can be found. At this time, the data content obtained reflects implicit knowledge and the implicit knowledge is related to the above explicit knowledge.
[0048] In this embodiment, step S102 can be understood as a data location process, which uses the premise fields of explicit knowledge to quickly locate data in the model database, so as to efficiently obtain data content that can reflect implicit knowledge and use it for explicit knowledge verification. Compared with the traditional manual method, this data location process is much more efficient.
[0049] In step S103, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. In this embodiment, after using the premise field to find the data content with the same premise from the model knowledge base, by comparing the data content of the conclusion with the field value of the conclusion field of the explicit knowledge, it can be determined whether the constructed explicit knowledge and the implicit knowledge used by the language model can draw the same conclusion under the same premise, thereby verifying whether the explicit knowledge and the implicit knowledge are consistent and whether the constructed explicit knowledge is correct.
[0050] In step S104, in response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge has passed the verification. In this embodiment, when the conclusion field of the explicit knowledge matches the data content, it means that the conclusion of the constructed explicit knowledge is consistent with the implicit knowledge used by the language model under the same premise, which means that the idea of reasoning using the explicit knowledge can accurately explain the reasoning logic of the language model, so it can be determined that the explicit knowledge has passed the verification.
[0051] Furthermore, since the data content provides a quantifiable basis for verification, in some embodiments, the similarity between the field value of the conclusion field and the found data content can also be calculated to obtain the probability of the conclusion of the explicit knowledge being true. Therefore, some other embodiments of the present disclosure also provide an explicit knowledge verification method, which Figure 1 Based on the method shown, the accuracy of the verification process is further improved by similarity calculation.
[0052] Combine the following Figure 2 This explicit knowledge verification method is explained. Figure 2 An exemplary flow chart of an explicit knowledge verification method 200 according to some other embodiments of the present disclosure is shown. Figure 2As shown, in step S201, the implicit knowledge in the language model is converted into explicit knowledge. It should be noted that the content of step S201 in this embodiment is consistent with step S101 in the previous embodiment, which has been fully described in the previous embodiment and will not be repeated here.
[0053] In step S202, according to the premise fields and field values of explicit knowledge, data contents reflecting implicit knowledge are searched from the model knowledge base. It should be noted that the contents of step S202 in this embodiment are consistent with those of step S102 in the previous embodiment, which has been described in detail in the previous embodiment and will not be further described here.
[0054] In step S203, the similarity between the field value of the conclusion field of the explicit knowledge and the data content is calculated to obtain the probability of the conclusion in the explicit knowledge being true. In practical applications, the source of knowledge includes data statistics, which reflect and analyze the development law of things as a whole. Therefore, there may be some abnormal data or data deviations that make it difficult for the field value of the actually calculated conclusion field to reach 100% similarity with the data content.
[0055] However, such abnormal data or data deviation will not have a serious impact on the reliability and accuracy of knowledge, but will interfere with the verification process. In order to prevent abnormal data or data deviation from affecting the verification results, this disclosure introduces similarity to characterize the degree of matching between the field value of the conclusion field and the data content, thereby improving the fault tolerance of the verification process.
[0056] In step S204, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. In this embodiment, step S204 can determine whether the field value of the conclusion field of the explicit knowledge matches the data content based on the probability of establishment obtained in step S203. Exemplarily, this embodiment can pre-set a probability threshold. When the probability of establishment is greater than or equal to the probability threshold, it is determined that the probability of establishment is high, indicating that the field value of the conclusion field of the explicit knowledge successfully matches the data content. When the probability of establishment is less than the probability threshold, it is determined that the probability of establishment is low, indicating that the field value of the conclusion field of the explicit knowledge fails to match the data content.
[0057] In step S205, in response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge has passed the verification. It should be noted that the content of step S205 in this embodiment is consistent with that of step S104 in the previous embodiment, which has been fully described in the previous embodiment and will not be repeated here.
[0058] Relative to Figure 1 The explicit knowledge verification method shown, Figure 2The method shown uses similarity to divide the degree of matching in the matching step and sets a certain tolerance interval for it, so as to eliminate the interference of abnormal data or data errors in the verification process, and further improve the reliability and accuracy of the verification result.
[0059] According to the contents described in the previous embodiments, it can be understood that the present disclosure needs to find the corresponding data content in the model knowledge base of the language model based on the premise field and conclusion field of the explicit knowledge to judge the degree of matching, so as to complete the verification of the accuracy and consistency of the explicit knowledge.
[0060] In order to efficiently and accurately complete data location and data comparison, after obtaining explicit knowledge, it is necessary to clarify the premise fields and conclusion fields of the explicit knowledge. To this end, some embodiments of the present disclosure provide another explicit knowledge verification method, which converts the expression form of explicit knowledge to facilitate the extraction of premise fields and conclusion fields from explicit knowledge.
[0061] Figure 3 An exemplary flow chart of an explicit knowledge verification method 300 according to some other embodiments of the present disclosure is shown. Figure 3 As shown, in step S301, the implicit knowledge in the language model is converted into explicit knowledge in the form of textual expression. In this embodiment, the explicit knowledge in the form of textual expression can be understood as explicit knowledge in the form of text and described in natural language, which is easy for people to understand and express. For example, the decline in catering demand and the decline in the operating income of catering companies are explicit knowledge in the form of textual expression.
[0062] In step S302, the explicit knowledge in the form of text expression is converted into explicit knowledge in the form of formal expression. In this embodiment, the formal expression can be understood as an expression form borrowing symbols and formulas from mathematics or logic. Taking the decline in catering demand and the decline in the operating income of the catering company in the previous text as an example, its formal expression can be as follows:
[0063] premise:
[0064] GetEventChange(event=demand,industry=food)==down;
[0065] IndustryHasCompany(industry=food,Company=Acompany)==True;
[0066] in conclusion:
[0067] GetCompanyIndex(Company=Acompany, companyIndex=businessIncome)==down.
[0068] It can be seen from the above examples that the formally expressed explicit knowledge includes premise fields and conclusion fields. Furthermore, the specific terms therein can also be associated with the table headers in the model knowledge base to facilitate subsequent data location.
[0069] Compared with textual expressions, formal expressions are more obscure and difficult to understand, and do not conform to the description rules of natural language. Therefore, manually converting the expression form of explicit knowledge also requires manpower and is prone to errors and omissions. Therefore, in actual applications, in addition to manual methods, a language model can also be used to execute step S302. For example, the prompt information prompt in natural language can be used to guide the language model to generate explicit knowledge in formal expression based on the explicit knowledge in textual expression.
[0070] The explicit knowledge expressed in formal form is clearer and more explicit in structure. This form of expression reduces the difficulty of dividing the premise field and the conclusion field, thereby improving the verification efficiency of explicit knowledge.
[0071] In step S303, according to the premise fields and field values of the explicit knowledge, the data content reflecting the implicit knowledge is searched from the model knowledge base. It should be noted that the content of step S303 in this embodiment is consistent with that of step S102 and step S202 in the previous embodiment, which has been fully described in the previous embodiment and will not be further elaborated here.
[0072] In step S304, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. It should be noted that step S304 in this embodiment is consistent with the content of step S103 and step S204 in the previous embodiment, which has been fully described in the previous embodiment and will not be repeated here.
[0073] Furthermore, the method shown in this embodiment can also introduce similarity to characterize the degree of matching. After step S303, the similarity between the field value of the conclusion field and the data content is calculated, and the probability of the conclusion being true is determined based on this, and then it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. It can be understood that the above combined with Figure 1 to Figure 2 The same features as described in this article apply to Figure 3 Description.
[0074] In step S305, in response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge has passed the verification. It should be noted that the content of step S303 in this embodiment is consistent with that of step S104 and step S205 in the previous embodiment, which has been fully described in the previous embodiment and will not be further described here.
[0075] Further, those skilled in the art will appreciate that Figure 3 After dividing the premise field and conclusion field of explicit knowledge by the method shown or other methods, some embodiments of the present disclosure can also locate data in the model knowledge base through structured query language.
[0076] For example, Figure 4 An exemplary flow chart of an explicit knowledge verification method 400 of some embodiments of the present disclosure is shown, as shown in FIG. Figure 4 As shown, in step S401, the implicit knowledge in the language model is converted into explicit knowledge. It should be noted that the content of step S401 in this embodiment is consistent with step S101 in the previous embodiment, which has been fully described in the previous embodiment and will not be repeated here.
[0077] Furthermore, the above combined Figure 3 The expression form conversion method of explicit knowledge shown is also applicable to step S401 of this embodiment. For example, after the implicit knowledge in the language model is converted into explicit knowledge in text expression form, it is converted into explicit knowledge in formal expression using a language model or other tools.
[0078] In step S402, a database query statement is constructed according to the premise fields of the explicit knowledge and their field values. In this embodiment, the database query statement includes but is not limited to an SQL statement.
[0079] In step S403, a query is performed in the model knowledge base based on the database query statement to obtain data content. For example, in this embodiment, an SQL statement can be written based on the premise fields and field values of the explicit knowledge, and then the data is located by executing the SQL statement to obtain the located data content.
[0080] In step S404, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. It should be noted that the content of step S404 in this embodiment is consistent with that of step S103, step S204 and step S304 in the previous embodiment, which has been fully described in the previous embodiment and will not be further described here.
[0081] Furthermore, the method of using similarity to characterize the degree of matching in the previous embodiment is also applicable to the present embodiment. After step S403, the present embodiment can also calculate the similarity between the field value of the conclusion field of the explicit knowledge and the data content to obtain the probability of the conclusion in the explicit knowledge being true, and then determine whether the field value of the conclusion field of the explicit knowledge matches the data content.
[0082] In step S405, in response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge has passed the verification. It should be noted that the content of step S405 in this embodiment is consistent with that of step S104, step S205 and step S305 in the previous embodiment, which has been fully described in the previous embodiment and will not be further described here.
[0083] In combination Figures 1 to 4 In the illustrated embodiment, the verification process of explicit knowledge only determines the degree of matching of the conclusion field. However, in actual application, there is a problem of inaccurate data positioning. At this time, the data content found does not match the premise field. However, the consistency verification requires that the explicit knowledge and implicit knowledge must be compared under the same premise. Therefore, the mismatch between the data content and the premise field will affect the subsequent verification steps.
[0084] In response to the above problems, those skilled in the art can understand that the disclosed embodiment also provides a matching method for explicit knowledge and implicit knowledge, which not only needs to match the conclusion field, but also needs to match the premise field.
[0085] Figure 5 An exemplary flow chart of a method 500 for matching explicit knowledge and implicit knowledge according to some embodiments of the present disclosure is shown. Figure 5 As shown, in step S501, the premise field and conclusion field of the explicit knowledge are obtained. In this embodiment, the explicit knowledge may be formally expressed explicit knowledge, which may be obtained by processing the explicit knowledge in textual expression form using a language model.
[0086] In step S502, it is determined whether the field value of the premise field of the explicit knowledge matches the data content. In this embodiment, if the judgment result is yes, it means that the premise field matches, and then step S503 is continued to complete the matching of the conclusion field. Further, if the judgment result is no, it can be considered that there is no data content related to the explicit knowledge in the model knowledge base. At this time, step S507 can be executed to determine that the explicit knowledge has not passed the verification or output a prompt message that the explicit knowledge cannot be verified. Furthermore, the explicit knowledge that cannot be verified can also be temporarily stored so that it can be verified again after the model knowledge base is updated.
[0087] In step S503, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. If the judgment result is yes, step S504 is executed. In some embodiments, the condition that step S503 determines that the field value of the conclusion field matches the data content is that the contents of the two are consistent. When the contents of the two are not completely consistent, the judgment result is no. At this time, step S505 can be executed to determine whether the similarity between the field value of the conclusion field of the explicit knowledge and the data content is greater than or equal to the probability threshold. If so, step S504 is executed to determine that the explicit knowledge has passed the verification; if not, step S506 is executed to determine that the explicit knowledge has not passed the verification.
[0088] It should be noted that the situation where the field value of the conclusion field is consistent with the data content can be understood as a situation where the similarity between the two is 100%. Therefore, in some embodiments, after the premise field is matched, the similarity between the field value of the conclusion field and the data content can be directly calculated and a determination can be made as to whether the similarity is greater than or equal to the probability threshold.
[0089] It should be further explained that the above Figure 5 The description is also applicable to the explicit knowledge verification method provided in any of the above embodiments.
[0090] Furthermore, before using the method shown in any of the above embodiments to complete the verification of explicit knowledge, it is necessary to establish an association relationship between the premise fields of the explicit knowledge and the model knowledge base to meet the needs of performing data location operations.
[0091] Based on this, those skilled in the art can understand that the present disclosure also provides an explicit knowledge verification method. Figure 6 An exemplary flow chart of an explicit knowledge verification method 600 of some embodiments of the present disclosure is shown, as shown in FIG. Figure 6 As shown, in step S601, an association relationship between the premise field and the table header in the model knowledge base is constructed. Taking the explicit knowledge formally expressed in the previous embodiment as an example, specific terms such as IndustryHasCompany and company can be matched with the table header in the model knowledge base in advance, thereby constructing an association relationship between the premise field and the table header in the model knowledge base.
[0092] In step S602, an association relationship between the conclusion field and the table header in the model knowledge base is constructed. Similar to step S601, a specific term such as GetCompanyIndex can be matched with the table header in the model knowledge base in advance, thereby constructing an association relationship between the conclusion field and the table header in the model knowledge base.
[0093] Then, step S603 to step S606 are executed in sequence to complete the verification of explicit knowledge. In step S603, the implicit knowledge in the language model is converted into explicit knowledge. In step S604, the data content reflecting the implicit knowledge is found from the model knowledge base according to the premise field and field value of the explicit knowledge. In step S605, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. In step S606, in response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge has passed the verification.
[0094] It should be noted that the specific implementation methods of the above steps S603 to S606 have been described in the previous section. Figures 1 to 5 The embodiment shown has been described in detail and will not be elaborated here.
[0095] Furthermore, in addition to the above combination Figure 6 In addition to the steps of establishing the association relationship described above, before using the method shown in any of the above embodiments to complete the verification of explicit knowledge, the following method can also be used in combination with Figure 7 The described method builds a language model and its model knowledge base.
[0096] Figure 7 An exemplary flow chart of an explicit knowledge verification method 700 of some embodiments of the present disclosure is shown, as shown in FIG. Figure 7 As shown, in step S701, ontology and semantic rules are constructed. Furthermore, ontology includes operators and concepts, wherein operators are used to realize calculations between concepts, each operator can be understood as a function, and concepts serve as input and output variables of the operator. Further, the input and output of the operator can be several concepts in the concept set, and new concepts can be obtained by calculations of operators and concepts. Concepts are sets composed of different instances. As described in the foregoing embodiments, the formally expressed knowledge includes premise fields and conclusion fields, wherein the premises and conclusions are composed of several knowledge equations, and each knowledge equation can be understood as an equation composed of concepts and operators.
[0097] Semantic rules reflect the way that users of a language give meaning to a particular linguistic symbol. In other words, semantics is the interpretation of data symbols, while semantic rules are the definition of the organizational rules and structural relationships between these symbols. Data itself is meaningless, and only data that is given meaning can be used. Therefore, to some extent, semantic rules can be regarded as providing the meaning of the data itself and directly or indirectly reflecting the relationship between data.
[0098] In step S702, constituent elements of the model knowledge base are formed according to the operators and concepts. Based on the definitions of operators and concepts described above, it can be understood that constituent elements of the model knowledge base can be formed based on operators and concepts and their calculation results.
[0099] In step S703, the elements are formally described based on the semantic rules to construct a model knowledge base of the language model. Based on the definition of the semantic rules described above, after giving the data a certain meaning, a database that is interpretable in a certain field and has a logical representation can be formed by building connections between the data, that is, the model knowledge base described above.
[0100] After constructing the language model and its model knowledge base, then execute steps S704 to S707 in sequence to complete the verification of explicit knowledge. Among them, in step S704, the implicit knowledge in the language model is converted into explicit knowledge. In step S705, according to the premise field and field value of the explicit knowledge, the data content reflecting the implicit knowledge is found from the model knowledge base. In step S706, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content. In step S707, in response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge has passed the verification.
[0101] It should be noted that the specific implementation methods of the above steps S704 to S707 have been described in the previous section. Figures 1 to 5 The embodiment shown has been described in detail and will not be elaborated here.
[0102] Through the description above in combination with the accompanying drawings, those skilled in the art can understand that the disclosed embodiment provides a method for verifying explicit knowledge, which quickly locates and obtains relevant data content by combining the premise field and conclusion field of explicit knowledge and the model knowledge base, and then verifies the explicit knowledge through data comparison operations. The verification process uses quantifiable verification basis, which is more reliable and accurate than manual verification.
[0103] Furthermore, those skilled in the art can also understand that the disclosed embodiment also provides another explicit knowledge verification method, which uses similarity to characterize the degree of matching between the conclusion field and the data content, and provides a certain tolerance interval for the interference caused by abnormal data and data errors in the matching process, thereby further ensuring the reliability and accuracy of the verification results.
[0104] In order to implement the method steps described in the foregoing text of this disclosure in conjunction with the accompanying drawings at the software and hardware level, the present disclosure also provides the following Figure 8 The electronic device shown. Specifically, Figure 8 An exemplary structural block diagram of an electronic device 800 according to an embodiment of the present disclosure is shown.
[0105] like Figure 8 As shown, the electronic device 800 disclosed herein may include a processor 810 and a memory 820. Specifically, the memory 820 stores executable program instructions. When the program instructions are executed by the processor 810, the electronic device implements the above-mentioned Figures 1 to 7 The method steps are described.
[0106] It is understood that in order to clearly illustrate the solution of the present disclosure and avoid confusion with the prior art, Figure 8 The electronic device 800 only shows the components related to the embodiment of the present disclosure, and omits those components that may be necessary for implementing the embodiment of the present disclosure but belong to the scope of the prior art. Therefore, based on the content disclosed in the present disclosure, a person skilled in the art can clearly understand that the electronic device 800 of the present disclosure may also include components related to the embodiment of the present disclosure. Figure 8 The constituent elements shown in are different from the common constituent elements.
[0107] In an exemplary implementation scenario, the above-mentioned processor 810 can control the overall operation of the electronic device 800. For example, the processor 810 can control the operation of the electronic device 800 by executing the program stored in the memory 820. In terms of implementation, the processor 810 of the present disclosure can be implemented by a central processing unit (CPU), an application processor (Application Processor, AP), an artificial intelligence processor chip (Intelligent Processing Unit, IPU), etc. provided in the electronic device 800. Further, the processor 810 of the present disclosure can also be implemented in any appropriate manner. For example, the processor 810 can take the form of a computer-readable medium, logic gates, switches, application-specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers, and embedded microcontrollers, etc., such as a microprocessor or processor and a computer-readable program code (such as software or firmware) that can be executed by the (micro) processor.
[0108] In terms of storage content, the memory 820 can be used to store hardware of various data and instructions processed in the electronic device 800. For example, the memory 820 can store processed data and data to be processed in the electronic device 800. The memory 820 can store data sets that have been processed or to be processed by the processor 810. In addition, the memory 820 can store applications, drivers, etc. to be driven by the electronic device 800. For example: the memory 820 can store various programs related to model building, data positioning, and data comparison to be executed by the processor 810. The memory 820 can be a DRAM, but the present disclosure is not limited to this. In terms of type, the memory 820 may include at least one of a volatile memory or a non-volatile memory. The non-volatile memory may include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a flash memory, a phase change RAM (PRAM), a magnetic RAM (MRAM), a resistive RAM (RRAM), a ferroelectric RAM (FRAM), and the like. The volatile memory may include a dynamic RAM (DRAM), a static RAM (SRAM), a synchronous DRAM (SDRAM), a PRAM, an MRAM, an RRAM, a ferroelectric RAM (FeRAM), etc. In an embodiment, the memory 820 may include at least one of a hard disk drive (HDD), a solid state drive (SSD), a high-density flash memory (CF), a secure digital (SD) card, a micro secure digital (Micro-SD) card, a mini secure digital (Mini-SD) card, an extreme digital (xD) card, caches, or a memory stick.
[0109] In summary, the specific functions implemented by the memory 820 and the processor 810 of the electronic device 800 provided in the embodiments of this specification can be explained in comparison with the aforementioned embodiments in this specification, and can achieve the technical effects of the aforementioned embodiments, and will not be repeated here.
[0110] Additionally or optionally, the present disclosure may also be implemented as a non-temporary machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) on which computer program instructions (or computer program, or computer instruction code) are stored. When the computer program instructions (or computer program, or computer instruction code) are executed by a processor of an electronic device (or electronic device, server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present disclosure.
[0111] Although multiple embodiments of the present disclosure have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art may think of many changes, modifications, and alternatives without departing from the thought and spirit of the present disclosure. It should be understood that in the process of practicing the present disclosure, various alternatives to the embodiments of the present disclosure described herein may be adopted. The attached claims are intended to define the scope of protection of the present disclosure, and therefore cover equivalents or alternatives within the scope of these claims.
Claims
1. A method for verifying explicit knowledge, characterized in that include: Convert implicit knowledge in language models into explicit knowledge; According to the premise fields and field values of the explicit knowledge, searching the model knowledge base for data content reflecting the implicit knowledge; Determining whether the field value of the conclusion field of the explicit knowledge matches the data content; and In response to the conclusion field of the explicit knowledge matching the data content, it is determined that the explicit knowledge passes verification.
2. The method according to claim 1, characterized in that The conversion of implicit knowledge in the language model into explicit knowledge includes: Convert implicit knowledge in language models into explicit knowledge in the form of textual expressions; and The explicit knowledge in the text expression form is converted into explicit knowledge in a formal expression, wherein the explicit knowledge in the formal expression includes a premise field and a conclusion field.
3. The method according to claim 1, characterized in that Before searching the model knowledge base for data content reflecting the implicit knowledge according to the premise fields and field values of the explicit knowledge, the method further includes: Constructing an association relationship between the premise field and the table header in the model knowledge base; and Construct an association relationship between the conclusion field and the table header in the model knowledge base.
4. The method according to claim 1, characterized in that: According to the premise fields and field values of the explicit knowledge, the data contents reflecting the implicit knowledge are searched from the model knowledge base, including: Constructing a database query statement according to the premise fields of the explicit knowledge and their field values; and A query is performed in the model knowledge base based on the database query statement to obtain the data content.
5. The method according to claim 1, characterized in that Wherein, judging whether the field value of the conclusion field of the explicit knowledge matches the data content includes: Determining whether the field value of the premise field of the explicit knowledge matches the data content; and In response to the field value of the premise field of the explicit knowledge matching the data content, it is determined whether the field value of the conclusion field of the explicit knowledge matches the data content.
6. The method according to claim 1, characterized in that After searching the model knowledge base for data content reflecting the implicit knowledge, the method further includes: The similarity between the field value of the conclusion field of the explicit knowledge and the data content is calculated to obtain the probability of the conclusion in the explicit knowledge being true.
7. The method according to claim 2, characterized in that The method of converting the explicit knowledge in the textual expression form into the explicit knowledge in the formal expression form includes: The language model is guided by prompt information in natural language so that it generates explicit knowledge in formal expression based on the explicit knowledge in the text expression form.
8. The method according to claim 7, characterized in that Before converting the implicit knowledge in the language model into explicit knowledge, the method further includes: Constructing ontology and semantic rules, where ontology includes operators and concepts; forming constituent elements of the model knowledge base according to the operators and the concepts; and The elements are formally described based on the semantic rules to construct a model knowledge base of the language model.
9. An electronic device, characterized in that: include: processor; as well as A memory storing executable program instructions, which, when executed by the processor, enables the device to implement the method according to any one of claims 1 to 8.
10. A computer-readable storage medium having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by one or more processors, the method according to any one of claims 1 to 8 is implemented.