Text citation authenticity verification methods, systems, devices, media and equipment
By extracting the candidate citation identifier set from the generated text and performing a difference operation with the index set of materials already in the database, the problem of mismatch between generated content and real materials and difficulty in tracing the source in generative AI text systems is solved, and fast and automatic citation authenticity verification is achieved.
Patent Information
- Application Number
- CN202611429265.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-15
- Publication Date
- 2026-11-17
AI Technical Summary
In existing technologies, generative AI text systems lack independent pre-verification when generating content, resulting in a mismatch between the generated content and the real material, making it difficult to trace the source. Illegal citations may be directly stored on disk, and existing verification methods cannot effectively prevent fictitious citations.
By extracting a set of candidate reference identifiers from the generated text and performing a difference operation with the unique set of material indexes of the material data already stored in the controlled storage area, the verification result is output to ensure that all reference identifiers in the generated text actually exist in the material data already stored.
It enables automatic and rapid verification of the authenticity of generated text citations, accurately identifies fictitious or erroneous citations, improves the automation and traceability of verification, and reduces the consumption of storage and computing resources.
Smart Images

Figure CN123115405A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of text verification technology, and more specifically, to a method, system, apparatus, medium, and device for verifying the authenticity of text citations. Background Technology
[0002] With the application of generative AI text systems, they typically consist of a source library, an inference buffer, and a text writing library. The model generates text based on prompts in the inference buffer, and the authenticity of citations mainly relies on semantic inference from a large model. In existing architectures, the source library is often open for reading and writing, while citation constraints are only stored in the inference buffer, and the writing interface lacks independent pre-validation. When the buffer is pruned, constraints are easily lost, and the model may detach from the underlying source library and fabricate citation identifiers, resulting in generated content that does not match the real source material. Furthermore, the assertions lack an index binding with the source material, making traceability difficult, and illegal citations may even be directly written to disk.
[0003] While related technologies exist, such as citation verification based on soft constraints of prompt words and citation verification based on large model self-checking, which can standardize citation formats or remind models to use citations cautiously to some extent, these methods still do not address the aforementioned architectural flaws: authenticity verification is still mainly concentrated in the semantic reasoning layer of the large model, lacking independent pre-verification, and assertions and source material lack index binding. Therefore, relying solely on soft constraints of prompt words cannot eliminate fictitious citations, and the problems of mismatched generated content with real source material and difficulty in tracing its origin still exist. Summary of the Invention
[0004] This disclosure provides at least one method, system, apparatus, medium, and device for verifying the authenticity of text citations. By performing a difference operation between the set of candidate citation identifiers in the generated text to be verified and the set of unique material indexes corresponding to the material data already stored in the controlled storage area, and outputting the verification result based on whether the difference set is an empty set, automatic, fast, and traceable verification of the authenticity of generated text citations is achieved.
[0005] This disclosure provides a method for verifying the authenticity of text citations, including: Obtain the generated text and material index set to be verified; wherein, each unique material index in the material index set corresponds to a piece of material data that has been stored in the controlled storage area, the controlled storage area being an underlying read-only storage medium, which stores the material data and the unique material index corresponding to each piece of material data; The generated text to be verified is subjected to citation identifier extraction processing to obtain a set of candidate citation identifiers contained in the generated text; Calculate the difference between the candidate reference identifier set and the material index set; and output the reference authenticity verification result of the generated text based on whether the difference set is an empty set; wherein each reference identifier in the difference set belongs to the candidate reference identifier set and does not belong to the material index set.
[0006] This disclosure provides a text citation authenticity verification system, which includes an underlying read-only storage medium, a middle-layer buffer controller, and a text verification processor. The underlying read-only storage medium is used to store the material data that has been entered into the database and the unique material index corresponding to each piece of material data; The middle-layer buffer controller is used to read the unique material index of all the material data that has been stored in the underlying read-only storage medium, form a material index set by all the read unique material indexes, and synchronize the material index set to the text verification processor. The text verification processor is used to obtain the generated text to be verified, perform reference identifier extraction processing on the generated text to obtain a candidate reference identifier set, calculate the difference between the candidate reference identifier set and the material index set, and output the reference authenticity verification result of the generated text according to whether the difference set is an empty set.
[0007] This disclosure provides a text citation authenticity verification device, including: The data acquisition module is used to acquire the generated text and material index set to be verified; wherein, each unique material index in the material index set corresponds to a piece of material data that has been stored in the controlled storage area [fsj1.1][fsj1.2], and the controlled storage area is an underlying read-only storage medium that stores the material data and the unique material index corresponding to each piece of material data; The identifier extraction module is used to extract citation identifiers from the generated text to be verified, thereby obtaining a set of candidate citation identifiers contained in the generated text. The text verification module is used to calculate the difference between the candidate reference identifier set and the material index set; and output the reference authenticity verification result of the generated text based on whether the difference set is an empty set; wherein each reference identifier in the difference set belongs to the candidate reference identifier set and does not belong to the material index set.
[0008] This disclosure provides a computer device, including a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the computer device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, a text citation authenticity verification method as described in any of the above possible embodiments is performed.
[0009] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the text citation authenticity verification method as described in any of the possible embodiments above.
[0010] The text citation authenticity verification method, system, device, medium, and equipment provided in this disclosure perform a difference operation between the set of candidate citation identifiers in the generated text to be verified and the set of unique material indexes corresponding to the material data already stored in the controlled storage area. The citation authenticity verification result is output based on whether the difference set is empty. This allows for the rapid determination of whether all candidate citation identifiers in the generated text truly exist in the material data already stored, accurately identifying fictitious citations, erroneous citations, or untraceable citations, and achieving automatic determination of the authenticity of citations in the generated text to be verified.
[0011] Here, the controlled storage area is the underlying read-only storage medium, which only provides read access to the unique material index. The source range of the citation identifier in the generated text is limited to the material data that has been stored in the database, thus restricting fictitious citation identifiers from the data source. At the same time, since the verification uses the unique material index as the comparison object, it is not necessary to read the full text of the material data, which can reduce the consumption of storage and computing resources and improve the automation, repeatability and traceability of the verification.
[0012] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0013] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings referenced in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.
[0014] Figure 1A flowchart of a text citation authenticity verification method provided by an embodiment of this disclosure is shown; Figure 2 A flowchart of a material import method provided by an embodiment of this disclosure is shown; Figure 3 This diagram illustrates the structure of a text citation authenticity verification system provided in an embodiment of the present disclosure. Figure 4 This diagram illustrates the structure of another text citation authenticity verification system provided in an embodiment of this disclosure; Figure 5 This diagram illustrates the structure of a text citation authenticity verification device provided in an embodiment of the present disclosure. Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of this disclosure is shown. Detailed Implementation
[0015] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.
[0016] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0017] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.
[0018] To facilitate understanding of this embodiment, the executing entity of the text citation authenticity verification method provided in this disclosure will first be described in detail. The executing entity of the text citation authenticity verification method provided in this disclosure is a computer device. This computer device can be a terminal device or a server. The terminal device can also be a mobile device, user terminal, terminal, handheld device, computing device, vehicle-mounted device, wearable device, etc. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, and artificial intelligence platforms. Optionally, this method can also be applied to an implementation environment composed of computer devices and servers.
[0019] This disclosure provides a method for verifying the authenticity of text citations. The method is used to verify the authenticity of citation identifiers in the generated text to be verified, and to determine whether all candidate citation identifiers in the generated text to be verified exist in the unique material index of the material data already stored in the database.
[0020] In some possible embodiments, this method can be applied to a three-tier isolated storage architecture. This three-tier isolated storage architecture includes a bottom-layer read-only storage medium, a middle-layer buffer controller, and an upper-layer inference memory. The bottom-layer read-only storage medium stores the material data already stored and the unique material index corresponding to each piece of material data. The bottom-layer read-only storage medium does not grant direct read / write permissions to the upper-layer inference memory. The middle-layer buffer controller reads the unique material indexes of all the material data already stored from the bottom-layer read-only storage medium, assembles all the read unique material indexes into a material index set, and synchronizes the material index set to the upper-layer inference memory. The upper-layer inference memory is used to deploy a generative artificial intelligence model. The generative artificial intelligence model generates the generated text to be verified in the upper-layer inference memory. The upper-layer inference memory does not have direct read / write permissions to the bottom-layer read-only storage medium and can only obtain the material index set through the middle-layer buffer controller. Based on this, the text verification processor executes the text citation authenticity verification method provided in this embodiment to perform citation authenticity verification on the generated text to be verified.
[0021] The following detailed description, with reference to the accompanying drawings, illustrates a method for verifying the authenticity of text citations provided in this disclosure. (See also...) Figure 1 The diagram shows a flowchart of a text citation authenticity verification method provided in this embodiment of the present disclosure. The method includes the following steps S101 to S103: S101, obtain the generated text and material index set to be verified.
[0022] It is understandable that the generated text to be verified refers to the text content generated by the generative artificial intelligence model in the inference memory and has not yet been verified for authenticity. It is the natural language text output by the large model based on the prompts or task instructions input by the user. The text may contain citations of the source data, which are used to indicate the source of the source data on which the text content is based. The generated text to be verified can contain one or more arguments, each argument corresponding to at least one citation. The citation can be embedded in the text in the form of parentheses or square brackets. For example, the generated text may contain citations in the form of (Zhang San, 2024) or [1], which respectively represent the year of authorship or number of a source data. Here, an argument refers to a declarative text unit in the generated text to be verified that expresses complete semantics. Each argument corresponds to at least one citation, which can be obtained by segmenting the generated text at the sentence level or paragraph level. For example, the generated text to be verified is that the accuracy of artificial intelligence in medical image diagnosis has reached more than 95% [1], but this conclusion is mainly based on a single-center dataset, and the generalization ability still needs to be verified [2]. After segmenting according to sentence granularity, two statements are obtained. The first statement is that the accuracy of artificial intelligence in medical image diagnosis has reached more than 95%, and its corresponding citation is [1]. The second statement is that this conclusion is mainly based on a single-center dataset, and the generalization ability still needs to be verified, and its corresponding citation is [2].
[0023] In some possible embodiments, the above-mentioned generative artificial intelligence model is a natural language generation model, which has the ability to generate natural language text based on prompts or task instructions input by the user. It can be a pre-trained language model based on the Transformer architecture, an open-source large language model, or a model obtained by fine-tuning an open-source large language model. No specific limitation is made here.
[0024] Specifically, the material index set represents the collection of unique material indexes for all material data already stored in the controlled storage area. It serves as the baseline data set for difference comparison with candidate reference identifiers in the generated text to be verified. Each unique material index in the material index set corresponds to one piece of material data already stored in the underlying read-only storage medium. This correspondence is established and solidified when the material data is stored. The controlled storage area refers to the storage region used to store the stored material data and unique material indexes. The controlled storage area can be various types of read-only storage media. The underlying read-only storage medium is one embodiment of the controlled storage area and can be a distributed key-value system or a local relational database, etc., as long as it satisfies the atomic read / write capability of the unique material indexes; no specific limitations are made here. It stores material data and the unique material index corresponding to each piece of material data. The material index set consists of the unique material indexes corresponding to each piece of material data already stored in the controlled storage area.
[0025] Understandably, the underlying read-only storage medium is a storage medium used to persistently store the material data that has been added to the database and its unique material index. It has access restrictions that do not grant direct read and write permissions to the upper-layer inference memory. It can store the material data that has been added to the database and the unique material index corresponding to each piece of material data. For the upper-layer inference memory, the underlying read-only storage medium can only provide the unique material index through the middle-layer buffer controller. The upper-layer inference memory does not have direct read and write permissions to the underlying read-only storage medium.
[0026] Here, source data refers to structured or unstructured text data obtained from external retrieval interfaces and processed for database storage. It serves as the original data that generative AI models can reference when generating text. Each piece of source data has a unique source index, which is used to uniquely identify the source data in the underlying read-only storage medium. The unique source index is an identifier assigned by the database processor when the source data is stored. It can be in the form of a primary key field, hash value, or serial number, and is used as a basis for comparing the authenticity of references during subsequent verification processes.
[0027] For example, during the material storage stage, considering that the retrieval interface may return empty results or error messages, reference will be made to... Figure 2 As shown, the above material index set can be constructed by following the steps S201~S204: S201, in response to the material retrieval request, calls the retrieval interface to obtain material data.
[0028] Here, a material retrieval request is a request signal initiated by the material ingestion processor when it needs to obtain material data from an external data source. It may include information such as search keywords, search scope, and search time range to instruct the retrieval interface to perform the corresponding material data query operation. Upon receiving this request, the external retrieval interface can be called to obtain material data matching the retrieval request. The retrieval interface refers to the application programming interface provided by the external data source, which can return the corresponding material data based on the retrieval request, serving as the initial data source for material ingestion. The corresponding material data can be obtained from external data sources such as academic literature databases and institutional knowledge bases through interface calls, data pushes, etc., but this embodiment does not limit this.
[0029] For example, the processor for storing materials can be an independent process running on the server side, the external data source can be an academic literature database, the retrieval interface can be an application programming interface provided by the academic literature database, and the material retrieval request can be a request message containing search keywords and search scope, which can instruct the retrieval interface to query matching material data in the specified data source, thereby calling the retrieval interface and obtaining the returned material data. For example, the search keyword is generative AI healthcare, the search scope is academic papers in the past five years, and the retrieval interface returns data such as the titles, authors, and abstracts of several papers.
[0030] S202, when the search interface returns an empty result or an error message, adjust the search parameters and call the search interface again to perform a second search.
[0031] Specifically, when the search interface returns empty results or an error message, it means that no valid material data was obtained under the current search conditions. At this time, the search parameters can be adjusted and the search interface can be called again for a second search to continue to obtain material data. The second search is the same as the first search in terms of the search interface call method, only the search parameters are different. The search parameter adjustment includes keyword expansion, synonym replacement, or widening of the search scope to increase the probability of the second search hitting valid material data.
[0032] Here, adjusting search parameters refers to expanding or modifying parameters such as search keywords and search scope. Adjusted search parameters can broaden the search's reach and supplement the initial search conditions. This can include keyword expansion, synonym replacement, and broadening the search scope. For example, if the initial search using keyword A returns no results, keyword A can be replaced with its synonyms or broader terms for a new search.
[0033] S203, when the secondary search returns an empty result or an error message, an entry failure signal is returned, prohibiting the allocation of the unique material index to the material data and prohibiting the writing of the material data to the controlled storage area.
[0034] Here, when a secondary search returns an empty result or an error message, it indicates that valid material data cannot be obtained under the current search conditions. At this time, no false index records are generated, an entry failure signal is returned, and it is prohibited to assign a unique material index to the material data. At the same time, it is prohibited to write the material data to the controlled storage area.
[0035] Understandably, not generating an index record when no valid material is found is a prerequisite for ensuring the reliability of the material index set during subsequent verification of citation authenticity, eliminating the need to allocate an index for non-existent material data. This ensures from the data source that there are no false index records in the controlled storage area.
[0036] S204, when the retrieval interface returns valid material data or the secondary retrieval returns valid material data, the valid material data is written into the controlled storage area, and a unique material index is assigned to the valid material data.
[0037] Specifically, when the retrieval interface returns valid material data or the secondary retrieval returns valid material data, it is confirmed that valid material data has been obtained. The valid material data can be written to the controlled storage area, and a unique material index can be assigned to the valid material data to establish a correspondence between the material data and the unique material index.
[0038] In this way, through the material entry process of S201~S204 described above, it can be ensured that each piece of material data stored in the controlled storage area has a corresponding unique material index, thereby enabling the construction of a material index set composed of unique material indexes of all material data that have been entered into the storage, which can then serve as a comparison benchmark for subsequent verification of the authenticity of references.
[0039] In some possible embodiments, the source data may contain structured feature fields. After writing the valid source data into the controlled storage area, to facilitate subsequent source tracing and matching of assertions in the generated text, the structured feature fields in the valid source data can be extracted and associated with the valid source data in the controlled storage area. Here, structured feature fields refer to field information obtained after structuring the content of the source data. These can include types such as research question fields, research method fields, and core conclusion fields, which can be used to quickly locate the core content of the source data during subsequent source tracing. When verifying citations in the generated text, the structured feature fields of the corresponding source data can be retrieved through the mapping relationship between assertions and source indexes, thereby quickly determining the research question, research method, and core conclusion of the source material on which the assertion is based, without the need for manual verification by searching the entire original source material line by line.
[0040] S102, the generated text to be verified is subjected to citation identifier extraction processing to obtain a set of candidate citation identifiers contained in the generated text.
[0041] Furthermore, after obtaining the generated text and material index set to be verified, citation identifier extraction processing can be performed on the generated text to be verified. This involves scanning and extracting citation identifiers from the generated text, traversing the entire text content to obtain a set of candidate citation identifiers contained within the generated text. In this set, each element is a citation identifier extracted from the generated text to be verified, and can include citation identifiers in parentheses and square brackets, respectively, corresponding to different citation formats.
[0042] Specifically, the generated text can be parsed using preset citation mark extraction rules. The generated text to be verified is scanned character by character or segment by segment according to the citation mark extraction rules, identifying and extracting text segments that conform to the citation mark format characteristics. This yields a set of candidate citation marks contained in the generated text, where each citation mark comes from the generated text itself. The citation mark extraction rules refer to pre-defined parsing rules used to identify citation marks from natural language text. These rules define the format characteristics and boundary conditions of citation marks and can be used to identify text segments in the generated text that have citation mark format characteristics, such as author-year combinations enclosed in parentheses or numerical codes enclosed in square brackets.
[0043] For example, the citation identifier extraction rule can use regular expressions to match the generated text to be validated in order to extract all citation identifiers that conform to the regular expression pattern. For example, the regular expression can include a first matching pattern and a second matching pattern. The first matching pattern is used to match citation identifiers in parentheses, such as (Zhang San, 2024), and the second matching pattern is used to match citation identifiers in square brackets, such as [1]. The generated text to be validated can be matched separately, and both types of citation identifiers can be extracted to form a candidate citation identifier set.
[0044] In some other embodiments, other text parsing methods such as abstract syntax tree parsing and named entity recognition can also be used to extract the above-mentioned reference identifiers. This embodiment does not limit this.
[0045] In this disclosure, by extracting citation identifiers from the generated text to be verified, scattered citation identifiers in the text can be uniformly extracted into a set of candidate citation identifiers, providing structured comparison data for subsequent difference operations with the material index set, thereby improving the automation level of citation authenticity verification.
[0046] In some possible embodiments, the generated text to be verified may also be empty text. Therefore, before performing citation identifier extraction processing on the generated text to be verified, it is possible to determine whether the generated text to be verified is empty, and thus determine whether to skip the citation identifier extraction processing. When the generated text to be verified is empty text, there is no need to perform citation identifier extraction processing and difference operation. The step of outputting the citation authenticity verification result of the generated text based on whether the difference set is empty (i.e., step S103) can be directly executed, skipping the citation identifier extraction processing, and determining the verification result as failing. In this way, a verification judgment can be made for empty text without performing the complete extraction and comparison process, improving the processing efficiency of citation authenticity verification and reducing unnecessary computational overhead.
[0047] S103, calculate the difference between the candidate reference identifier set and the material index set; and output the reference authenticity verification result of the generated text based on whether the difference set is an empty set.
[0048] Specifically, by calculating the difference between the candidate reference identifier set and the material index set, it can be determined whether the generated text to be verified contains reference identifiers that do not belong to the material index set, i.e., whether there are reference identifiers for material data that has not been entered into the database. Here, the difference set is the set obtained by performing a set difference operation between the candidate reference identifier set and the material index set. Each reference identifier in this set belongs to the candidate reference identifier set but not to the material index set. This indicates whether the generated text to be verified references material data that does not exist in the controlled storage area, and the authenticity of the references in the generated text to be verified can be determined accordingly.
[0049] Furthermore, the authenticity verification result of the generated text to be verified can be determined based on whether the difference set is empty. When the difference set is empty, it means that each reference identifier in the candidate reference identifier set exists in the material index set. In this case, the generated text to be verified does not reference material data that has not been entered into the database. Indication information indicating that the reference verification of the generated text has passed can be output, indicating that all references in the generated text are genuine references. For example, a verification pass flag or a verification pass message can be output. When the difference set is not empty, it means that at least one reference identifier in the candidate reference identifier set does not belong to the material index set. In this case, the generated text to be verified references material data that has not been entered into the database. Indication information indicating that the reference verification of the generated text has failed can be output, indicating that there are fictitious references in the generated text. For example, a verification fail flag or a verification fail message can be output. The reference identifiers in the difference set are marked as difference reference identifiers to identify which specific reference identifiers in the generated text to be verified did not find corresponding material data in the controlled storage area, which facilitates subsequent location and processing.
[0050] In some possible embodiments, the generated text to be verified can be written to persistent storage medium after the verification passes. Therefore, when the reference authenticity verification result is passed, the generated text to be verified can be written to persistent storage medium to complete the normal disk storage of the text. When the reference authenticity verification result is failed, a write rejection instruction can be issued to the text write queue to prevent the generated text to be verified from being written to persistent storage medium. In this way, the reference authenticity verification result is used as a prerequisite for writing text to persistent storage medium, which can intercept generated text containing fictitious reference identifiers and reduce the disk space occupation of persistent storage medium by invalid text.
[0051] In some possible embodiments, to quickly locate the source data upon which each assertion in the generated text is based during subsequent verification, a two-way index binding process can be further executed when the difference set is empty and the authenticity verification result is passed. The two-way index binding process establishes a correspondence between assertions and source data, facilitating subsequent automated traceability. Specifically, it can include: establishing a mapping relationship between each assertion in the generated text and its corresponding unique source index; that is, associating the assertion identifier of each assertion with its corresponding unique source index in the source index set, forming a two-way mapping relationship between assertions and source indexes; and storing this two-way mapping relationship in a mapping register. During subsequent verification, the structured fields of the source data can be retrieved back through the index, eliminating the need for manual verification of each original source data item, thus achieving automated traceability.
[0052] Here, the bidirectional mapping storage structure between assertions and materials is not limited to the form of a data table. The mapping register can adopt any storage organization method such as a relational table or a key-value mapping library, as long as the correspondence between assertion identifiers and unique material indexes can be established and maintained, so that the corresponding unique material index can be queried based on the assertion identifier, or the corresponding assertion identifier can be queried based on the unique material index. This embodiment does not limit this.
[0053] The text citation authenticity verification method, system, device, medium, and equipment provided in this disclosure perform a difference operation between the set of candidate citation identifiers in the generated text to be verified and the set of unique material indexes corresponding to the material data already stored in the controlled storage area. The citation authenticity verification result is output based on whether the difference set is empty. This allows for the rapid determination of whether all candidate citation identifiers in the generated text truly exist in the material data already stored, accurately identifying fictitious citations, erroneous citations, or untraceable citations, and achieving automatic determination of the authenticity of citations in the generated text to be verified.
[0054] Here, the controlled storage area is the underlying read-only storage medium, which only provides read access to the unique material index. The source range of the citation identifier in the generated text is limited to the material data that has been stored in the database, thus restricting fictitious citation identifiers from the data source. At the same time, since the verification uses the unique material index as the comparison object, it is not necessary to read the full text of the material data, which can reduce the consumption of storage and computing resources and improve the automation, repeatability and traceability of the verification.
[0055] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.
[0056] Based on the same inventive concept, this disclosure also provides a text citation authenticity verification system corresponding to the text citation authenticity verification method. Since the principle of the system in this disclosure for solving the problem is similar to the text citation authenticity verification method described above in this disclosure, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.
[0057] Reference Figure 3 The diagram shown is a schematic of a text citation authenticity verification system provided in an embodiment of this disclosure. The system includes a bottom-level read-only storage medium, a middle-level buffer controller, and a text verification processor. The underlying read-only storage medium is used to store the material data that has been entered into the database and the unique material index corresponding to each piece of material data; The middle-layer buffer controller is used to read the unique material index of all the material data that has been stored in the underlying read-only storage medium, form a material index set by all the read unique material indexes, and synchronize the material index set to the text verification processor. The text verification processor is used to obtain the generated text to be verified, perform reference identifier extraction processing on the generated text to obtain a candidate reference identifier set, calculate the difference between the candidate reference identifier set and the material index set, and output the reference authenticity verification result of the generated text according to whether the difference set is an empty set.
[0058] In some possible embodiments, refer to Figure 4 As shown, the system also includes a material import processor, specifically used for: In response to a material retrieval request, the retrieval interface is called to obtain material data; When the search interface returns an empty result or an error message, adjust the search parameters and call the search interface again to perform a second search. When the secondary search returns an empty result or an error message, an entry failure signal is returned, prohibiting the allocation of the unique material index to the material data and prohibiting the writing of the material data to the underlying read-only storage medium; When the retrieval interface returns valid material data or the secondary retrieval returns valid material data, the valid material data is written to the underlying read-only storage medium, and a unique material index is assigned to the valid material data.
[0059] In some possible embodiments, the material import processor is further configured to: Extract the structured feature fields from the effective material data; wherein, the structured feature fields include research question fields, research method fields, and core conclusion fields; The structured feature fields are associated with the valid material data and stored in the underlying read-only storage medium.
[0060] In some possible embodiments, the text validation processor is further configured to: Determine whether the generated text to be verified is empty text; When the generated text to be verified is empty text, the step of outputting the verification result of the reference authenticity of the generated text based on whether the difference set is an empty set is directly executed, and the verification result is determined to be unsuccessful.
[0061] In some possible embodiments, the text validation processor is specifically used for: The generated text is parsed using preset citation identifier extraction rules to obtain a set of candidate citation identifiers contained in the generated text; wherein, the citation identifier extraction rules are used to identify text segments in the generated text that have citation identifier format features.
[0062] In some possible embodiments, the text validation processor is specifically used for: When the difference set is an empty set, an indication message is output to indicate that the citation verification of the generated text has passed; When the difference set is a non-empty set, an indication message is output to indicate that the reference verification of the generated text fails, and the reference identifier in the difference set is marked as a difference reference identifier.
[0063] Those skilled in the art should understand that the underlying read-only storage medium, the middle buffer controller, the text verification processor, etc., described in this embodiment can be equivalently implemented as memory variables and software threads in a general-purpose computer system, or as hardware circuits such as application-specific integrated circuits and field-programmable gate arrays. Both the software and hardware equivalent solutions fall within the protection scope of this disclosure.
[0064] Based on the same inventive concept, this disclosure also provides a text citation authenticity verification device corresponding to the text citation authenticity verification method. Since the principle of the device in this disclosure for solving the problem is similar to the text citation authenticity verification method described above in this disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.
[0065] Reference Figure 5 The diagram shown is a schematic of a text citation authenticity verification device 500 provided in an embodiment of this disclosure. The device includes: The data acquisition module 501 is used to acquire the generated text and material index set to be verified; wherein, each unique material index in the material index set corresponds to a piece of material data that has been stored in the controlled storage area, and the controlled storage area is an underlying read-only storage medium that stores the material data and the unique material index corresponding to each piece of material data; The identifier extraction module 502 is used to perform citation identifier extraction processing on the generated text to be verified, and obtain a set of candidate citation identifiers contained in the generated text; The text verification module 503 is used to calculate the difference between the candidate reference identifier set and the material index set; and output the reference authenticity verification result of the generated text according to whether the difference set is an empty set; wherein, each reference identifier in the difference set belongs to the candidate reference identifier set and does not belong to the material index set.
[0066] In some possible embodiments, the data acquisition module 501 is further configured to: In response to a material retrieval request, the retrieval interface is called to obtain material data; When the search interface returns an empty result or an error message, adjust the search parameters and call the search interface again to perform a second search. When the secondary search returns an empty result or an error message, an entry failure signal is returned, prohibiting the allocation of the unique material index to the material data and prohibiting the writing of the material data to the controlled storage area; When the retrieval interface returns valid material data or the secondary retrieval returns valid material data, the valid material data is written to the controlled storage area, and a unique material index is assigned to the valid material data.
[0067] In some possible embodiments, the data acquisition module 501 is further configured to: Extract the structured feature fields from the effective material data; wherein, the structured feature fields include research question fields, research method fields, and core conclusion fields; The structured feature fields are associated with the valid material data and stored in the controlled storage area.
[0068] In some possible embodiments, the identifier extraction module 502 is further configured to: Determine whether the generated text to be verified is empty text; When the generated text to be verified is empty text, the step of outputting the verification result of the reference authenticity of the generated text based on whether the difference set is an empty set is directly executed, and the verification result is determined to be unsuccessful.
[0069] In some possible embodiments, the identifier extraction module 502 is specifically used for: The generated text is parsed using preset citation identifier extraction rules to obtain a set of candidate citation identifiers contained in the generated text; wherein, the citation identifier extraction rules are used to identify text segments in the generated text that have citation identifier format features.
[0070] In some possible embodiments, the text verification module 503 is specifically used for: When the difference set is an empty set, an indication message is output to indicate that the citation verification of the generated text has passed; When the difference set is a non-empty set, an indication message is output to indicate that the reference verification of the generated text fails, and the reference identifier in the difference set is marked as a difference reference identifier.
[0071] Based on the same technical concept, this disclosure also provides a computer device. (See also...) Figure 6 The diagram shows the structure of a computer device 600 provided in this embodiment of the present disclosure, including a processor 601, a memory 602, and a bus 603. The memory 602 stores execution instructions and includes a main memory 6021 and an external memory 6022. The main memory 6021, also called internal memory, is used to temporarily store computational data in the processor 601 and data exchanged with external memory 6022 such as a hard disk. The processor 601 exchanges data with the external memory 6022 through the main memory 6021.
[0072] In this embodiment, the memory 602 is specifically used to store application code that executes the scheme of this disclosure, and its execution is controlled by the processor 601. That is, when the computer device 600 is running, the processor 601 communicates with the memory 602 through the bus 603, so that the processor 601 executes the application code stored in the memory 602, and then executes the text reference authenticity verification method described in any of the foregoing embodiments.
[0073] The memory 602 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.
[0074] Processor 601 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor can be a microprocessor or any conventional processor.
[0075] It is understood that the structures illustrated in the embodiments of this disclosure do not constitute a specific limitation on the computer device 600. In other embodiments of this disclosure, the computer device 600 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.
[0076] This disclosure also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the text citation authenticity verification method described in the above method embodiments. The storage medium can be a volatile or non-volatile computer-readable storage medium.
[0077] This disclosure also provides a computer program product carrying program code. The program code includes instructions that can be used to execute the steps of the text citation authenticity verification method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.
[0078] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.
[0079] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed systems and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.
[0080] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0081] In addition, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0082] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0083] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit them. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features within the scope of the technology disclosed in this disclosure. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure.
Claims
1. A method for verifying the authenticity of text citations, characterized in that, include: Obtain the generated text and material index set to be verified; wherein, each unique material index in the material index set corresponds to a piece of material data that has been stored in the controlled storage area, the controlled storage area being an underlying read-only storage medium, which stores the material data and the unique material index corresponding to each piece of material data; The generated text to be verified is subjected to citation identifier extraction processing to obtain a set of candidate citation identifiers contained in the generated text; Calculate the difference between the candidate reference identifier set and the material index set; and output the reference authenticity verification result of the generated text based on whether the difference set is an empty set; wherein each reference identifier in the difference set belongs to the candidate reference identifier set and does not belong to the material index set.
2. The text citation authenticity verification method according to claim 1, characterized in that, Before obtaining the set of generated text and material indexes to be verified, the following steps are included: In response to a material retrieval request, the retrieval interface is called to obtain material data; When the search interface returns an empty result or an error message, adjust the search parameters and call the search interface again to perform a second search. When the secondary search returns an empty result or an error message, an entry failure signal is returned, prohibiting the allocation of the unique material index to the material data and prohibiting the writing of the material data to the controlled storage area; When the retrieval interface returns valid material data or the secondary retrieval returns valid material data, the valid material data is written to the controlled storage area, and a unique material index is assigned to the valid material data.
3. The text citation authenticity verification method according to claim 2, characterized in that, After writing the valid material data into the controlled storage area, the method further includes: Extract the structured feature fields from the effective material data; wherein, the structured feature fields include research question fields, research method fields, and core conclusion fields; The structured feature fields are associated with the valid material data and stored in the controlled storage area.
4. The text citation authenticity verification method according to claim 1, characterized in that, Before performing the citation identifier extraction process on the generated text to be verified, the following steps are included: Determine whether the generated text to be verified is empty text; When the generated text to be verified is empty text, the step of outputting the verification result of the reference authenticity of the generated text based on whether the difference set is an empty set is directly executed, and the verification result is determined to be unsuccessful.
5. The text citation authenticity verification method according to claim 4, characterized in that, The step of extracting citation identifiers from the generated text includes: The generated text is parsed using preset citation identifier extraction rules to obtain a set of candidate citation identifiers contained in the generated text; wherein, the citation identifier extraction rules are used to identify text segments in the generated text that have citation identifier format features.
6. The text citation authenticity verification method according to claim 1, characterized in that, The step of outputting the authenticity verification result of the generated text based on whether the difference set is an empty set includes: When the difference set is an empty set, an indication message is output to indicate that the citation verification of the generated text has passed; When the difference set is a non-empty set, an indication message is output to indicate that the reference verification of the generated text fails, and the reference identifier in the difference set is marked as a difference reference identifier.
7. A text citation authenticity verification system, characterized in that, The text citation authenticity verification system includes an underlying read-only storage medium, a middle-layer buffer controller, and a text verification processor. The underlying read-only storage medium is used to store the material data that has been entered into the database and the unique material index corresponding to each piece of material data; The middle-layer buffer controller is used to read the unique material index of all the material data that has been stored in the underlying read-only storage medium, form a material index set by all the read unique material indexes, and synchronize the material index set to the text verification processor. The text verification processor is used to obtain the generated text to be verified, perform reference identifier extraction processing on the generated text to obtain a candidate reference identifier set, calculate the difference between the candidate reference identifier set and the material index set, and output the reference authenticity verification result of the generated text according to whether the difference set is an empty set.
8. A text citation authenticity verification device, characterized in that, include: The data acquisition module is used to acquire the generated text and material index set to be verified; wherein, each unique material index in the material index set corresponds to a piece of material data that has been stored in the controlled storage area, and the controlled storage area is an underlying read-only storage medium that stores the material data and the unique material index corresponding to each piece of material data; The identifier extraction module is used to extract citation identifiers from the generated text to be verified, thereby obtaining a set of candidate citation identifiers contained in the generated text. The text verification module is used to calculate the difference between the candidate reference identifier set and the material index set; and output the reference authenticity verification result of the generated text based on whether the difference set is an empty set; wherein each reference identifier in the difference set belongs to the candidate reference identifier set and does not belong to the material index set.
9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the text citation authenticity verification method according to any one of claims 1 to 6.
10. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the text citation authenticity verification method according to any one of claims 1 to 6.
