Information checking method of laboratory whole-process real-time input system

Abstract

The invention relates to the field of information checking, in particular to an information checking method of a laboratory whole-process real-time input system. The method comprises the following steps: firstly, acquiring an experiment operation record about a laboratory full-process real-time input system, segmenting contents in the experiment operation record, performing entity relationship extraction on each segmented statement, obtaining an entity vocabulary and a triple of each segmented statement, and based on the occurrence frequency of the entity vocabulary in the segmented statement in other segmented statements, performing entity relationship extraction on each segmented statement; the method comprises the following steps of: analyzing the reasonability of an entity extraction result according to an entity extraction result and the distribution similarity condition of triads containing entity vocabularies in other statement segments, performing same analysis under different segmentation modes so as to obtain optimal triads, and extracting the entity extraction result on the basis of all the optimal triads so as to obtain an entity extraction result. And carrying out information checking on experimental operation records of the whole-process real-time entry system of the laboratory. According to the invention, the information input into the system in real time in the whole process of the laboratory can be accurately checked, and the accuracy of information checking is improved.

Description

Technical Field

[0001] The present invention relates to the field of information verification, and in particular to an information verification method for a laboratory full-process real-time input system. Background Art

[0002] The importance of laboratory full-process input and verification of experimental operations is to ensure the safety of experimental operations, improve medical quality, reduce medical errors, and protect the lives of patients. Through a strict verification process, the possibility of errors in experimental operations can be minimized to ensure the smooth progress of experimental operations.

[0003] In the related technology, a dictionary matching method is usually used to extract triples representing entity relationships in the information of the laboratory's full-process real-time entry system, and the information is verified based on the extracted triples. However, due to untimely dictionary updates and the inability to accurately identify target words in a variety of different word order expressions, some longer academic terms are prone to partial interception during entity extraction, which makes it impossible to accurately verify the information of the laboratory's full-process real-time entry system through existing methods, thereby reducing the accuracy of information verification. Summary of the invention

[0004] In order to solve the technical problem that the existing method cannot accurately input the information of the laboratory full process real-time input system, and reduces the accuracy of information verification, the purpose of the present invention is to provide an information verification method for the laboratory full process real-time input system, and the technical scheme adopted is as follows: The present invention proposes an information verification method for a laboratory full-process real-time input system, the method comprising: Obtain experimental operation records for the real-time input system of the entire laboratory process; Segment the contents of the experimental operation record to obtain multiple segmented sentences of the experimental operation record; extract entity relations for each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence, wherein the triples contain two entity words, any one of the entity words is used as a target entity word, and papers containing the target entity words are selected from a database, and the commonness of the target entity words is obtained according to the number of papers containing the target entity words and the publication time of the papers; The segmented sentence where the target entity vocabulary is located is taken as the target segmented sentence, any triple of the target segmented sentence containing the target entity vocabulary is taken as the target triple, other segmented sentences except the target segmented sentence and containing the same triple as the target triple are taken as reference segmented sentences of the target triple, and the triple in each reference segmented sentence that is the same as the target triple is taken as the reference triple, and the rationality of the target triple is obtained according to the difference in the number of words between the two entity vocabulary of the target triple of the target segmented sentence and the number of words between the two entity vocabulary of the reference triple of each reference segmented sentence, and the number of the same entity vocabulary between the two entity vocabulary of the target triple of the target segmented sentence and the two entity vocabulary of the reference triple of each reference segmented sentence; the frequency of occurrence of the target entity vocabulary of the target segmented sentence is obtained according to the rationality of each triple containing the target entity vocabulary in the target segmented sentence and the number of the reference segmented sentences; Segment each segmented sentence in different ways, and extract entity relations for each segmented sentence in different segmentation ways, and obtain the extraction accuracy of target entity vocabulary in different segmentation ways according to the number of times the target entity vocabulary is extracted in different segmentation ways at the corresponding position, the frequency of occurrence of the target entity vocabulary when it is extracted in different segmentation ways at the corresponding position, the common degree of the target entity vocabulary, and the difference in the frequency of occurrence of the target entity vocabulary in different segmentation ways; based on the extraction accuracy of the entity vocabulary at the same position of each segmented sentence in different segmentation ways, obtain the best triplet of each segmented sentence; Based on all the best triples, the experimental operation records entered into the real-time system of the entire laboratory process are checked for information.

[0005] Furthermore, the multiple segmented statements of obtaining the experimental operation record include: The experimental operation record is divided by using periods as intervals to obtain multiple segmented sentences of the experimental operation record.

[0006] Further, the obtaining of multiple entity words and multiple triples of each segmented sentence includes: Based on the rule and dictionary method, entity relations are extracted for each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence.

[0007] Furthermore, obtaining the commonness of the target entity vocabulary includes: The papers containing the target entity vocabulary are used as reference papers of the target entity vocabulary, and the absolute value of the difference between the publication time of each reference paper of the target entity vocabulary and the current time is used as the publication time of each reference paper of the target entity vocabulary; in, Indicates the commonness of the target entity vocabulary; The number of reference papers representing the target entity vocabulary; The maximum number of reference papers representing all entity words; The first The publication time of the referenced paper.

[0008] Furthermore, the rationality of obtaining the target triplet includes: The number of characters between two entity words in the target triple of the target segmented sentence is used as the first interval distance of the target triple of the target segmented sentence; the number of characters between two entity words in the reference triple of each reference segmented sentence is used as the second interval distance of the reference triple of each reference segmented sentence; The set consisting of all entity words between two entity words of the target triple of the target segmented sentence is used as the first vocabulary set of the target triple of the target segmented sentence; the set consisting of all entity words between two entity words of the reference triple of each reference segmented sentence is used as the second vocabulary set of the reference triple of each reference segmented sentence; the intersection of the first vocabulary set and the second vocabulary set of the reference triple of each reference segmented sentence is used as the third vocabulary set of the reference triple of each reference segmented sentence; in, Indicates the rationality of the target triple; The first interval distance of the target triplet representing the target segmented sentence; Indicates The second interval distance of the reference triples of the reference segmented sentences; Indicates the number of entity words in the third word set of the reference triples of the reference segmented sentences; The number of reference segmentation sentences representing the target triples; Indicates the preset first adjustment parameter, The value range is .

[0009] Furthermore, the obtaining of the occurrence frequency of the target entity vocabulary of the target segmented sentence includes: in, Indicates the frequency of occurrence of the target entity vocabulary in the target segment sentence; Indicates that the target segment sentence contains the target entity vocabulary The rationality of the triples; Indicates that the target segment sentence contains the target entity vocabulary The number of reference segmentation sentences of triples; The maximum number of reference segment sentences that contain all triples of the target entity vocabulary in the target segment sentence; Indicates the number of triples containing the target entity vocabulary in the target segment sentence.

[0010] Furthermore, the segmenting of each segmented sentence in different ways includes: The number of words between two entity words in each triple in each segmented sentence is taken as the span range of each triple in each segmented sentence; Each segmented statement is segmented by increasing the span from the minimum span range to the maximum span range in each segmented statement with a step length of 1 character; When each span range segments a segmented statement, the position of the initial segment moves backward one character at a time, starting with the first character of each segmented statement, and stops traversing when the last character of the span range is the last character of the segmented statement; When the span ranges between segmented statements are consistent or closest, they are divided into one segmentation method.

[0011] Furthermore, the step of obtaining the extraction accuracy of target entity vocabulary under different segmentation modes includes: in, Indicates The extraction accuracy of target entity vocabulary under the segmentation method; Indicates the number of times the target entity vocabulary is extracted in different segmentation modes at the corresponding position; It represents the sum of the frequency of occurrence of the target entity vocabulary when it is extracted in different segmentation modes at the corresponding position; Indicates the maximum frequency of occurrence of all entity words extracted at the current position; Indicates The frequency of occurrence of the target entity vocabulary in the target segmented sentence under the segmentation method; Indicates The frequency of occurrence of the target entity vocabulary in the target segmented sentence under the segmentation method; Indicates the preset second adjustment parameter, The value range is .

[0012] Further, obtaining the best triple for each segmented sentence includes: The entity vocabulary corresponding to the maximum extraction accuracy obtained at the same position of each segmented sentence under different segmentation methods is taken as the best entity vocabulary in each segmented sentence. Entity relationship extraction is performed on each segmented sentence to obtain the best triplet of each segmented sentence.

[0013] Furthermore, the information verification of the experimental operation records of the laboratory full-process real-time input system based on all the best triples includes: Import all the best triplets into the GPT model, generate corresponding sentences, and sort the corresponding sentences according to the sequence order of the triplets to generate new text; Perform synonym replacement operation on the two new texts, perform word-by-word text matching on the two texts obtained after the operation, and provide feedback on inconsistent results during text matching; When the experimental operation record obtained during the previous operation meets the actual requirements, the current operator continues to perform the next step of the experiment, otherwise the experiment is terminated.

[0014] The present invention has the following beneficial effects: The present invention segments sentences by setting different segmentation intervals, and extracts entity words and entity relationships in segmented sentences through existing methods, and judges the rationality of entity words through the number of occurrences of entity words in different segmented sentences and the distribution of entity words with other relationships, thereby effectively reducing the shortcomings of slow dictionary update or inability to accurately identify target words in a variety of different word order expressions, making the recognition of corresponding sentence semantics more accurate and the verification result more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flow chart of an information verification method for a laboratory full-process real-time input system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the information verification method of a laboratory full-process real-time input system proposed by the present invention, its specific implementation method, structure, characteristics and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0018] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0019] The following is a detailed description of a specific scheme of an information verification method for a laboratory full-process real-time input system provided by the present invention in conjunction with the accompanying drawings.

[0020] See also Figure 1 , which shows a flow chart of an information verification method for a laboratory full-process real-time input system provided by an embodiment of the present invention, the method comprising: Step S1: Obtain the experimental operation records of the whole laboratory process into the real-time input system.

[0021] The importance of laboratory full-process input and verification of experimental operations lies in ensuring the safety of experimental operations, improving medical quality, reducing medical errors, and protecting the lives of patients. Through a strict verification process, the possibility of errors in experimental operations can be minimized, ensuring the smooth progress of experimental operations and the recovery of patients.

[0022] The embodiment of the present invention first uses RFID technology to store the record text of the operator's actual operation through the RFID chip, and obtains the experimental operation record of the laboratory's full-process real-time input system, wherein the experimental operation record is presented in text form, and the content included specifically includes the various operation steps of the operator during the experimental operation; at the same time, when performing the next step of operation, another operator of the corresponding operation will read the corresponding record text to the computer through the RFID chip, and then compare the actual operation steps of the previous operator with the corresponding operation steps pre-stored in the computer. When the comparison is successful, the current operator continues to perform the next step of operation, otherwise the experiment is terminated. The overall record verification system supports multiple operating systems, including Windows, IOS, Android, etc.

[0023] Among them, RFID technology is a technical means well known to those skilled in the art and will not be elaborated here.

[0024] Step S2: Segment the content in the experimental operation record to obtain multiple segmented sentences of the experimental operation record; perform entity relationship extraction on each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence, wherein the triple contains two entity words, and any one of the entity words is used as the target entity word, and papers containing the target entity word are selected from the database, and the commonness of the target entity word is obtained according to the number of papers containing the target entity word and the publication time of the papers.

[0025] Different operators will have their own writing habits when writing experimental operation records, and these habits will be reflected in aspects such as word order expression. In the experimental operation records, due to the complexity of the operation and the detailed requirements of the records, triples related to the currently extracted vocabulary often appear in multiple areas of the experimental operation records, and the word order is similar, and the front and back steps in the records are similar. Therefore, the embodiment of the present invention first segments the content in the experimental operation records to obtain multiple segmented sentences of the experimental operation records, and then entity relationships can be extracted for different segmented sentences, so as to analyze the segmented sentences.

[0026] Preferably, in one embodiment of the present invention, the experimental operation record is divided using each period in the experimental operation record as an interval to obtain a plurality of segmented sentences of the experimental operation record.

[0027] The most important thing in the record checking system is to capture entity words and entity relationships in the record text. Therefore, the embodiment of the present invention extracts entity relationships for each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence, where entity words are words or phrases with specific meanings and identifiability in the text. They can be specific nouns such as names of people, places, organizations, and products, or they can be abstract concepts or attributes. Entity relationships refer to a certain connection or interaction between two or more entities. This relationship can be spatial, temporal, causal, or inclusive. Depending on the nature of the entity and the context, the entity relationship can be represented in the form of triples, where a triple is composed of three words: two entity words and one relationship word, namely (entity word 1, relationship word, entity word 2). For example, for the sentence "Using Coomassie Brilliant Blue G-250 as a standard, the Coomassie Brilliant Blue method was used to determine the protein content of the pumpkin seed polysaccharide component." Entity relationship extraction is performed, and the obtained triplets are (Coomassie Brilliant Blue G-250, as, standard), (Coomassie Brilliant Blue method, determination, pumpkin seed polysaccharide component protein content).

[0028] Preferably, in one embodiment of the present invention, the method for obtaining multiple entity words and multiple triples of each segmented sentence specifically includes: Based on the rule and dictionary method, entity relations are extracted for each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence. In the entity relation extraction process, two medical dictionaries, CHV and SNOMED-CT, are used as references. The rules between entities in actual experimental operations are used to define the relationship types between entities, and the grammar and semantic rules are constructed as references for the rule-based method when capturing entity relations. The rule-based and dictionary-based method is a technical means well known to those skilled in the art and will not be elaborated here.

[0029] When using rule-based and dictionary-based methods to extract entity relationships from segmented sentences, errors in entity vocabulary extraction may often occur due to insufficient updating of medical dictionaries and incomplete coverage of academic terms or drug names involved in current experimental operations, which in turn affects subsequent verification work. For example, "Fogliptin benzoate tablets" may be segmented and extracted as "benzoic acid" because there is no record in the dictionary, resulting in a mismatch between the extracted entity and the entity it represents. In order to determine whether the current extracted vocabulary is accurate, the extracted vocabulary can be searched in papers related to the current experimental operation records. The more papers that use the extracted vocabulary as keywords and the more recent the writing time of the papers containing the extracted vocabulary, the greater the possibility that the current extracted vocabulary is the actual vocabulary used. Therefore, any entity vocabulary can be first used as the target entity vocabulary, and papers containing the target entity vocabulary can be selected from the database. The commonness of the target entity vocabulary can be obtained based on the number of papers containing the target entity vocabulary and the publication time of the papers. The greater the commonness, the more likely the target entity vocabulary is to be a frequently used vocabulary, providing a data basis for subsequent calculations to improve the accuracy of information verification.

[0030] Preferably, in one embodiment of the present invention, the method for obtaining the commonness of the target entity vocabulary specifically includes: First, the papers containing the target entity vocabulary are taken as the reference papers of the target entity vocabulary, and the absolute value of the difference between the publication time of each reference paper of the target entity vocabulary and the current time is taken as the publication time of each reference paper of the target entity vocabulary, and then the common degree expression is used for calculation: in, Indicates the commonness of the target entity vocabulary; The number of reference papers representing the target entity vocabulary; The maximum number of reference papers representing all entity words; The first The publication time of the referenced papers; Represents the cumulative value of the publication time of all reference papers of the target entity vocabulary.

[0031] in, The larger the value is, the more the target entity vocabulary is used in the reference papers. The smaller it is, the newer the current extracted vocabulary is, and the more it conforms to the characteristics of the continuous updating of medical knowledge.

[0032] At this point, the commonness of the target entity vocabulary is obtained.

[0033] Step S3: Take the segmented sentence where the target entity vocabulary is located as the target segmented sentence, take any triple in the target segmented sentence containing the target entity vocabulary as the target triple, take other segmented sentences except the target segmented sentence and containing the same triple as the target triple as the reference segmented sentences of the target triple, and take the triple in each reference segmented sentence that is the same as the target triple as the reference triple, and obtain the rationality of the target triple according to the difference in the number of words between the two entity vocabulary of the target triple of the target segmented sentence and the number of words between the two entity vocabulary of the reference triple of each reference segmented sentence, and the number of entity vocabulary that is the same between the two entity vocabulary of the target triple of the target segmented sentence and the two entity vocabulary of the reference triple of each reference segmented sentence; and obtain the frequency of occurrence of the target entity vocabulary of the target segmented sentence according to the rationality of each triple containing the target entity vocabulary in the target segmented sentence and the number of reference segmented sentences.

[0034] When extracting entity relations from a segmented sentence, an entity word may appear in different triples. For example, the triples for the sentence "Studies have shown that quercetin can inhibit a variety of tumor signal transduction pathways, inhibit cancer cell proliferation, improve tumor microenvironment, and effectively reduce the toxicity of certain chemotherapy drugs." include (quercetin, inhibit, tumor signal transduction pathway), (quercetin, inhibit, cancer cell proliferation), (quercetin, improve, tumor microenvironment), (quercetin, reduce, toxicity), etc., where the entity word "quercetin" appears in different triples. At the same time, when entity relations are extracted from different segmented sentences, the same triples will appear in different segmented sentences. For example, the triple (quercetin, inhibition, cancer cell proliferation) may appear in different segmented sentences. In a certain segmented sentence, the more times a triple containing a certain entity word appears in other segmented sentences, and the smaller the difference in the number of words between two entity words in the same triple, the more the number of identical entity words between two entity words, which proves that the entity word itself is more likely to be used in the experimental operation record.

[0035] Therefore, the embodiment of the present invention first takes the segmented sentence where the target entity vocabulary is located as the target segmented sentence, takes any triple of the target segmented sentence containing the target entity vocabulary as the target triple, takes other segmented sentences except the target segmented sentence and containing the same triple as the target triple as the reference segmented sentence of the target triple, and takes the triple that is the same as the target triple in each reference segmented sentence as the reference triple, and then obtains the rationality of the target triple by calculating the difference in the number of words between the two entity vocabulary of the target triple of the target segmented sentence and the number of words between the two entity vocabulary of the reference triple of each reference segmented sentence, as well as the number of the same entity vocabulary between the two entity vocabulary of the target triple of the target segmented sentence and the two entity vocabulary of the reference triple of each reference segmented sentence. The rationality of the extraction of the target entity vocabulary is judged and referenced by the rationality, thereby providing a data basis for subsequent calculations.

[0036] Preferably, in one embodiment of the present invention, the method for obtaining the rationality of the target triplet specifically includes: The number of characters between two entity words in the target triple of the target segmented sentence is used as the first interval distance of the target triple of the target segmented sentence; the number of characters between two entity words in the reference triple of each reference segmented sentence is used as the second interval distance of the reference triple of each reference segmented sentence.

[0037] The set consisting of all entity words between the two entity words of the target triple of the target segmented sentence is used as the first vocabulary set of the target triple of the target segmented sentence; the set consisting of all entity words between the two entity words of the reference triple of each reference segmented sentence is used as the second vocabulary set of the reference triple of each reference segmented sentence; the intersection of the first vocabulary set and the second vocabulary set of the reference triple of each reference segmented sentence is used as the third vocabulary set of the reference triple of each reference segmented sentence, and then the calculation is performed using the expression of reasonable degree: in, Indicates the rationality of the target triple; The first interval distance of the target triplet representing the target segmented sentence; Indicates The second interval distance of the reference triples of the reference segmented sentences; Indicates the number of entity words in the third word set of the reference triples of the reference segmented sentences; The number of reference segmentation sentences representing the target triples; Indicates the preset first adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0038] in, The smaller the The larger the value is, the more similar the distribution of the target triples in the target segmented sentence is to the distribution of the reference triples in the reference segmented sentence, which further indicates that the extraction of the target entity vocabulary is more reasonable. The larger it is, the greater the number of identical entity words between the two entity words in the target triple and the two entity words in the reference triple. This further indicates that the more similar sentences appear in the experimental operation records for the segmented sentences containing the target triple containing the target entity words, the more reasonable the extraction of the target entity words is.

[0039] Using the same method as above, we can obtain the rationality of each triple of the target entity vocabulary in the target segmented sentence. The more similar distributions of the multiple triples corresponding to the obtained entity vocabulary are contained in the experimental operation records and the greater the similarity, the more accurate the current extracted vocabulary is and the more times it appears. Therefore, we can analyze the rationality of each triple of the target entity vocabulary in the target segmented sentence and the number of reference segmented sentences to obtain the frequency of occurrence of the target entity vocabulary in the target segmented sentence. The greater the frequency of occurrence, the more triples related to the target entity vocabulary will appear in multiple areas of the experimental operation records, and the word order is similar, and the front and back steps in the records are similar, providing a data basis for subsequent calculations.

[0040] Preferably, in one embodiment of the present invention, the method for obtaining the occurrence frequency of the target entity vocabulary of the target segmented sentence specifically includes: Calculate using an expression of frequency of occurrence: in, Indicates the frequency of occurrence of the target entity vocabulary in the target segment sentence; Indicates that the target segment sentence contains the target entity vocabulary The rationality of the triples; Indicates that the target segment sentence contains the target entity vocabulary The number of reference segmentation sentences of triples; The maximum number of reference segment sentences that contain all triples of the target entity vocabulary in the target segment sentence; Indicates the number of triples containing the target entity vocabulary in the target segment sentence.

[0041] in, and The larger it is, the more times the target entity vocabulary appears in the experimental operation records, and the more reasonable it is.

[0042] Step S4: Segment each segmented sentence in different ways, and extract entity relations for each segmented sentence under different segmentation methods, and obtain the extraction accuracy of the target entity vocabulary under different segmentation methods according to the number of times the target entity vocabulary is extracted under different segmentation methods at the corresponding position, the frequency of occurrence of the target entity vocabulary when extracted under different segmentation methods at the corresponding position, the common degree of the target entity vocabulary, and the difference in the frequency of occurrence of the target entity vocabulary under different segmentation methods; based on the extraction accuracy of the entity vocabulary at the same position of each segmented sentence under different segmentation methods, obtain the best triplet of each segmented sentence.

[0043] When a sentence in the experimental operation record is divided into different sentence ranges, different entity words and different entity word relationships may be obtained. For example, for a special word, if the interception position is inside the word, the word will be divided into two parts, resulting in insufficient recognition of the current word or not being used as an extracted word, resulting in omissions in the recognition of the experimental operation steps.

[0044] Therefore, the embodiment of the present invention first segments each segmented sentence in different ways, and extracts entity relationships for each segmented sentence under different segmentation methods. The entity relationship extraction method used here is the same as the entity relationship extraction method used in step S2. The specific method of segmentation is: the number of characters between two entity words in each triple in each segmented sentence is used as the span range of each triple in each segmented sentence; each segmented sentence is segmented by increasing the span range from the minimum span range to the maximum span range in each segmented sentence with a step of 1 character; when segmenting the segmented sentences, the position of the initial segment is moved backward by one character each time, starting from the first character of each segmented sentence, and the traversal is stopped when the last character of the span range is the last character of the segmented sentence; when the segment span ranges between the segmented sentences are consistent or closest, they are divided into one segmentation method.

[0045] The more reasonable and realistic the entity vocabulary is when extracting entity words from a sentence, the more times the extracted entity vocabulary appears when the sentence is segmented into multiple segments, and the more frequent the entity vocabulary corresponding to the current segmentation appears in the experimental operation records. Since the specified segmentation processing range changes from small to large, the entity vocabulary obtained at the same position of the analyzed sentence under different segmentation processing may gradually change from part of the real vocabulary to all, and when the complete real vocabulary is obtained, the possibility of increasing the number of words in the current range of vocabulary will suddenly decrease. Therefore, it can be judged by the changes in the frequency of occurrence of the entity vocabulary corresponding to the same position during multiple changes. The smaller the change in the frequency of occurrence, the more realistic the entity vocabulary obtained under the current segmentation processing.

[0046] It should be noted that, under different segmentation methods, in the same segmented sentence, when two entity words have the largest number of words in the same position, they are the corresponding entity words under the two different segmentation methods.

[0047] Then, according to the number of times the target entity vocabulary is extracted in different segmentation methods at the corresponding position, the frequency of occurrence of the target entity vocabulary when it is extracted in different segmentation methods at the corresponding position, the commonness of the target entity vocabulary, and the difference in the frequency of occurrence of the target entity vocabulary in different segmentation methods, the extraction accuracy of the target entity vocabulary in different segmentation methods is obtained, and then the best triplet can be obtained based on the extraction accuracy to improve the accuracy of subsequent information verification of the experimental operation records.

[0048] Preferably, in one embodiment of the present invention, the method for obtaining the extraction accuracy of target entity vocabulary under different segmentation methods specifically includes: Calculate by extracting the expression for accuracy: in, Indicates The extraction accuracy of target entity vocabulary under the segmentation method; Indicates the number of times the target entity vocabulary is extracted in different segmentation modes at the corresponding position; It represents the sum of the frequency of occurrence of the target entity vocabulary when it is extracted in different segmentation modes at the corresponding position; Indicates the maximum frequency of occurrence of all entity words extracted at the current position; Indicates The frequency of occurrence of the target entity vocabulary in the target segmented sentence under the segmentation method; Indicates The frequency of occurrence of the target entity vocabulary in the target segmented sentence under the segmentation method; Indicates the preset second adjustment parameter, which is used to prevent the denominator from being 0. The value range is In one embodiment of the present invention, Set to 0.01, The specific value of can also be set by the implementer according to the specific implementation scenario and is not limited here.

[0049] in, , and The larger it is, the more likely it is that the target entity vocabulary is an actual vocabulary. The smaller it is, the smaller the change in the current frequency of occurrence is, and the more appropriate the entity vocabulary is under the current segmentation method.

[0050] Furthermore, based on the extraction accuracy of entity words at the same position of each segmented sentence under different segmentation methods, the best triplet of each segmented sentence can be obtained.

[0051] Preferably, in one embodiment of the present invention, the method for obtaining the best triple of each segmented sentence specifically includes: The entity vocabulary corresponding to the maximum value of the extraction accuracy obtained at the same position of each segmented sentence under different segmentation methods is taken as the best entity vocabulary in each segmented sentence, and entity relationship extraction is performed on each segmented sentence to obtain the best triplet of each segmented sentence, wherein the method used for entity relationship extraction here is the same as the method used for entity relationship in step S2.

[0052] At this point, the best triples for each segmented sentence in the experimental operation record are obtained.

[0053] Step S5: Based on all the best triples, the experimental operation records of the whole process of the laboratory are input into the system in real time for information verification.

[0054] After obtaining all the best triplets in the experimental operation records, the experimental operation records of the laboratory full-process real-time entry system can be more accurately verified based on all the best triplets, thereby improving the accuracy of the information verification of the laboratory full-process real-time entry system.

[0055] Preferably, in one embodiment of the present invention, the method for verifying the information of the experimental operation records of the laboratory full-process real-time input system specifically includes: First, all the best triples are imported into the GPT model to generate corresponding sentences, and the corresponding sentences are sorted according to the sequence order of the triples to generate new texts. Among them, the GPT model is a technical means well known to technicians in this field and will not be elaborated here.

[0056] Then, a synonym replacement operation is performed on the two new texts, and word-by-word text matching is performed on the two texts obtained after the operation, and feedback is given to the inconsistent results during text matching. Among them, the synonym replacement operation and word-by-word text matching are technical means well known to those skilled in the art, and will not be elaborated here.

[0057] When the experimental operation record obtained during the previous operation meets the actual requirements, the current operator continues to perform the next step of the experiment, otherwise the experiment is terminated.

[0058] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0059] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for checking information of a laboratory full-process real-time input system, characterized in that: The method comprises: Obtain experimental operation records for the real-time input system of the entire laboratory process; Segment the contents of the experimental operation record to obtain multiple segmented sentences of the experimental operation record; extract entity relations for each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence, wherein the triples contain two entity words, any one of the entity words is used as a target entity word, and papers containing the target entity words are selected from a database, and the commonness of the target entity words is obtained according to the number of papers containing the target entity words and the publication time of the papers; The segmented sentence where the target entity vocabulary is located is taken as the target segmented sentence, any triple of the target segmented sentence containing the target entity vocabulary is taken as the target triple, other segmented sentences except the target segmented sentence and containing the same triple as the target triple are taken as reference segmented sentences of the target triple, and the triple in each reference segmented sentence that is the same as the target triple is taken as the reference triple, and the rationality of the target triple is obtained according to the difference in the number of words between the two entity vocabulary of the target triple of the target segmented sentence and the number of words between the two entity vocabulary of the reference triple of each reference segmented sentence, and the number of the same entity vocabulary between the two entity vocabulary of the target triple of the target segmented sentence and the two entity vocabulary of the reference triple of each reference segmented sentence; the frequency of occurrence of the target entity vocabulary of the target segmented sentence is obtained according to the rationality of each triple containing the target entity vocabulary in the target segmented sentence and the number of the reference segmented sentences; Segment each segmented sentence in different ways, and extract entity relations for each segmented sentence in different segmentation ways, and obtain the extraction accuracy of target entity vocabulary in different segmentation ways according to the number of times the target entity vocabulary is extracted in different segmentation ways at the corresponding position, the frequency of occurrence of the target entity vocabulary when it is extracted in different segmentation ways at the corresponding position, the common degree of the target entity vocabulary, and the difference in the frequency of occurrence of the target entity vocabulary in different segmentation ways; based on the extraction accuracy of the entity vocabulary at the same position of each segmented sentence in different segmentation ways, obtain the best triplet of each segmented sentence; Based on all the best triples, the experimental operation records entered into the real-time system of the entire laboratory process are checked for information.

2. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The multiple segmented statements for obtaining the experimental operation records include: The experimental operation record is divided by using periods as intervals to obtain multiple segmented sentences of the experimental operation record.

3. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The obtaining of multiple entity words and multiple triples of each segmented sentence includes: Based on the rule and dictionary method, entity relations are extracted for each segmented sentence to obtain multiple entity words and multiple triples for each segmented sentence.

4. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The obtaining of the commonness of the target entity vocabulary includes: The papers containing the target entity vocabulary are used as reference papers of the target entity vocabulary, and the absolute value of the difference between the publication time of each reference paper of the target entity vocabulary and the current time is used as the publication time of each reference paper of the target entity vocabulary; in, Indicates the commonness of the target entity vocabulary; The number of reference papers representing the target entity vocabulary; The maximum number of reference papers representing all entity words; The first The publication time of the referenced paper.

5. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The rationality of obtaining the target triplet includes: The number of characters between two entity words in the target triple of the target segmented sentence is used as the first interval distance of the target triple of the target segmented sentence; the number of characters between two entity words in the reference triple of each reference segmented sentence is used as the second interval distance of the reference triple of each reference segmented sentence; The set consisting of all entity words between two entity words of the target triple of the target segmented sentence is used as the first vocabulary set of the target triple of the target segmented sentence; the set consisting of all entity words between two entity words of the reference triple of each reference segmented sentence is used as the second vocabulary set of the reference triple of each reference segmented sentence; the intersection of the first vocabulary set and the second vocabulary set of the reference triple of each reference segmented sentence is used as the third vocabulary set of the reference triple of each reference segmented sentence; in, Indicates the rationality of the target triple; The first interval distance of the target triplet representing the target segment sentence; Indicates The second interval distance of the reference triples of the reference segmented sentences; Indicates the number of entity words in the third word set of the reference triples of the reference segmented sentences; The number of reference segmentation sentences representing the target triples; Indicates the preset first adjustment parameter, The value range is .

6. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The frequency of occurrence of the target entity vocabulary of the target segmented sentence is obtained as follows: in, Indicates the frequency of occurrence of the target entity vocabulary in the target segment sentence; Indicates that the target segment sentence contains the target entity vocabulary The rationality of the triples; Indicates that the target segment sentence contains the target entity vocabulary The number of reference segmentation sentences of triples; The maximum number of reference segment sentences that contain all triples of the target entity vocabulary in the target segment sentence; Indicates the number of triples containing the target entity vocabulary in the target segment sentence.

7. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The different ways of segmenting each segmented sentence include: The number of words between two entity words in each triple in each segmented sentence is taken as the span range of each triple in each segmented sentence; Each segmented statement is segmented by increasing the span from the minimum span range to the maximum span range in each segmented statement with a step length of 1 character; When each span range segments a segmented statement, the position of the initial segment moves backward one character at a time, starting with the first character of each segmented statement, and stops traversing when the last character of the span range is the last character of the segmented statement; When the span ranges between segmented statements are consistent or closest, they are divided into one segmentation method.

8. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The extraction accuracy of the target entity vocabulary under different segmentation methods includes: in, Indicates The extraction accuracy of target entity vocabulary under the segmentation method; Indicates the number of times the target entity vocabulary is extracted in different segmentation modes at the corresponding position; It represents the sum of the frequency of occurrence of the target entity vocabulary when it is extracted in different segmentation modes at the corresponding position; Indicates the maximum frequency of occurrence of all entity words extracted at the current position; Indicates The frequency of occurrence of the target entity vocabulary in the target segmented sentence under the segmentation method; Indicates The frequency of occurrence of the target entity vocabulary in the target segmented sentence under the segmentation method; Indicates the preset second adjustment parameter, The value range is .

9. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The method of obtaining the best triple for each segmented sentence includes: The entity vocabulary corresponding to the maximum extraction accuracy obtained at the same position of each segmented sentence under different segmentation methods is taken as the best entity vocabulary in each segmented sentence. Entity relationship extraction is performed on each segmented sentence to obtain the best triplet of each segmented sentence.

10. The information verification method of a laboratory full-process real-time input system according to claim 1 is characterized in that: The information verification of the experimental operation records of the laboratory full-process real-time input system based on all the best triples includes: Import all the best triplets into the GPT model, generate corresponding sentences, and sort the corresponding sentences according to the sequence order of the triplets to generate new text; Perform synonym replacement operation on the two new texts, perform word-by-word text matching on the two texts obtained after the operation, and provide feedback on inconsistent results during text matching; When the experimental operation record obtained during the previous operation meets the actual requirements, the current operator continues to perform the next step of the experiment, otherwise the experiment is terminated.

Images