A battery positive electrode material evidence question and answer method and system for experimental condition comparability checking

Abstract

This invention discloses a question-and-answer method and system for battery cathode materials evidence, oriented towards verifying the comparability of experimental conditions, belonging to the field of materials informatics technology. To address the problems of existing material question-and-answer schemes failing to effectively handle differences in experimental conditions, easily mixing incomparable literature evidence for the same conclusion, and resulting in insufficient accuracy and verifiability of the question-and-answer results, this invention constructs a condition-enhanced evidence graph. It receives user questions and parses out the material object, performance indicators, and experimental condition constraints. After retrieving candidate evidence units, it normalizes the experimental conditions and verifies their comparability with the stated experimental condition constraints. The resulting set of comparable evidence undergoes consistency analysis, source credibility assessment, and conflict resolution to generate a structured adjudication result. Finally, it outputs the question-and-answer results and evidence explanations. This invention improves the accuracy, interpretability, and credibility of material question-and-answer results.

Description

Technical Field

[0001] This invention belongs to the field of materials informatics technology, and in particular relates to a question-and-answer method and system for evidence of battery cathode materials for comparability verification under experimental conditions. Background Technology

[0002] Against the backdrop of the continued advancement of the "dual-carbon" strategy and the rapid development of the new energy industry, electrochemical energy storage technologies such as lithium-ion batteries and sodium-ion batteries have received widespread attention. As a crucial component affecting specific capacity, cycle stability, rate performance, and thermal stability, the research and development of battery cathode materials increasingly relies on the efficient acquisition, accurate comparison, and comprehensive analysis of existing experimental data and literature evidence. With the deepening of materials research, information on the composition design, preparation processes, structural characteristics, performance indicators, and failure mechanisms of different cathode material systems is widely dispersed across heterogeneous data from multiple sources, including academic papers, patent documents, technical reports, and databases, forming a vast but structurally fragmented resources of materials knowledge. Unlike information retrieval in general fields, experimental conclusions in the field of battery cathode materials often exhibit strong conditional dependence. The same material often shows significantly different performance results under different test temperatures, voltage windows, rate capabilities, cycle counts, preparation processes, and characterization methods. Therefore, when searching for and utilizing relevant literature evidence, researchers not only need to focus on the material and performance indicators themselves but also need to carefully determine whether different pieces of evidence are comparable in terms of experimental conditions. If the differences in experimental conditions are ignored, and conclusions from literature from different sources are directly spliced, summarized, or compared, it is easy to lead to distorted conclusions and affect the reliability of material screening, mechanism analysis, and R&D decisions.

[0003] In existing technologies, data acquisition and question-answering for materials science literature mainly rely on manual retrieval, keyword matching, text fragment retrieval, or general structured information extraction methods. While these methods can improve the efficiency of materials information acquisition to some extent, they typically focus on identifying explicit information such as material names and performance values, lacking sufficient ability to structure experimental conditions, perform normalization processing, and align across different documents. They also struggle to effectively characterize differences in test conditions across different documents. When faced with questions with clear experimental constraints, such as "whether the cycling stability of a certain material is superior to that of another material under a specific voltage window," existing methods tend to incorporate evidence of inconsistent conditions into the answer generation process, thereby reducing the accuracy and auditability of the question-answering results. Furthermore, while some existing materials knowledge organization methods introduce knowledge graphs, databases, or other structured storage methods, they focus more on the static representation of entity relationships, lacking an evidence-based organization centered on "material object - performance index - experimental value - experimental conditions - source information." This makes it difficult to directly support evidence retrieval, comparability verification, and conflict resolution oriented towards experimental condition constraints. When the same material and the same indicator have numerical differences or even conflicting conclusions in different literature, existing technologies usually lack effective evidence screening and adjudication mechanisms. They cannot distinguish which conclusions are valid evidence formed under similar experimental conditions and which conclusions are data that cannot be directly compared due to inconsistent conditions. Therefore, it is difficult to output Q&A results with clear applicable conditions and source basis to users. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a question-and-answer method and system for battery cathode materials to verify the comparability of experimental conditions, thereby resolving the issues existing in the prior art.

[0005] Firstly, to achieve the above objectives, this invention provides a question-and-answer method for battery cathode materials to verify the comparability of experimental conditions, comprising the following steps:

[0006] Acquire multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, and construct a condition-enhanced evidence map. The condition-enhanced evidence map includes evidence units organized by material object, performance index, experimental value, experimental conditions and source information.

[0007] The system receives a natural language question input by the user and parses the natural language question to extract the target material object, target performance index, and experimental condition constraints.

[0008] Based on the target material object, the target performance index, and the experimental condition constraints, candidate evidence units are retrieved from the condition-enhanced evidence map.

[0009] The experimental conditions in the candidate evidence units are normalized, and the normalized experimental conditions are compared with the experimental condition constraints to obtain a set of comparable evidence.

[0010] The comparable evidence set is subjected to consistency analysis, source credibility assessment and conflict resolution to generate a structured adjudication result;

[0011] Based on the structured adjudication results, a question-and-answer result and evidence description are generated and sent to the user.

[0012] Secondly, the present invention also provides a battery cathode material evidence question-and-answer system for comparability verification of experimental conditions, used to implement a battery cathode material evidence question-and-answer method for comparability verification of experimental conditions, the system comprising:

[0013] The evidence graph construction module is used to acquire multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, construct evidence units containing material objects, performance indicators, experimental values, experimental conditions and source information, and construct condition-enhanced evidence graphs based on the evidence units.

[0014] The problem parsing module is used to receive natural language questions input by users, parse the natural language questions to extract target material objects, target performance indicators and experimental condition constraints;

[0015] The evidence retrieval module is used to retrieve candidate evidence units from the condition-enhanced evidence map based on the target material object, the target performance index and the experimental condition constraints.

[0016] The condition normalization module is used to normalize the experimental conditions in the candidate evidence units;

[0017] The comparability verification module is used to perform comparability verification between the normalized experimental conditions and the experimental condition constraints to obtain a set of comparable evidence.

[0018] The evidence adjudication module is used to perform consistency analysis, source credibility assessment and conflict resolution on the comparable evidence set, and generate structured adjudication results;

[0019] The question-and-answer output module is used to generate question-and-answer results and evidence descriptions based on the structured adjudication results and send them to the user.

[0020] Thirdly, the present invention also provides a computer terminal device, comprising:

[0021] One or more processors;

[0022] A memory, coupled to the processor, for storing one or more programs;

[0023] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the battery cathode material evidence question-and-answer method for comparability verification of experimental conditions in the first aspect described above.

[0024] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the battery cathode material evidence question-and-answer method for comparability verification of experimental conditions described in the first aspect above.

[0025] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the battery cathode material evidence question-and-answer method for comparability verification of experimental conditions described in the first aspect above.

[0026] Compared with the prior art, the present invention has the following advantages and technical effects:

[0027] This invention provides a question-and-answer method and system for battery cathode materials evidence, oriented towards verifying the comparability of experimental conditions. By introducing experimental condition normalization and comparability verification mechanisms, this invention structurally identifies and filters differences in experimental conditions from different sources, avoiding the mixing of inconsistent evidence in the same question-and-answer conclusion, thereby significantly improving the accuracy and reliability of the material question-and-answer results. Addressing the high sensitivity of battery cathode material performance conclusions to test temperature, voltage window, rate capability, cycle number, and preparation process, this invention introduces experimental condition constraints during evidence retrieval and screening, ensuring that the question-and-answer results correspond to specific experimental contexts, thus enhancing the adaptability of the material question-and-answer system to condition-sensitive issues. Simultaneously with outputting the question-and-answer results, this invention provides applicable experimental conditions, source evidence, conflict explanations, and credibility levels, enabling researchers to verify and trace the basis for the answers, thereby improving the transparency and auditability of the question-and-answer results. For numerical differences or conflicting conclusions of the same material and performance index in different literature, this invention can form a structured adjudication result through consistency analysis, source credibility assessment and conflict resolution mechanism. At the same time, it can update the evidence unit and rule parameters by combining manual verification feedback, thereby improving the system's continuous optimization capability and adaptability to new materials and new data scenarios. Attached Figure Description

[0028] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0029] Figure 1This is an overall flowchart of a battery cathode material evidence question-and-answer method for verifying the comparability of experimental conditions, according to an embodiment of the present invention.

[0030] Figure 2 This is a flowchart illustrating the construction process of a conditionally enhanced evidence graph according to an embodiment of the present invention.

[0031] Figure 3 This is a flowchart illustrating the normalization and comparability verification of experimental conditions in an embodiment of the present invention.

[0032] Figure 4 This is a flowchart illustrating the evidence conflict resolution and structured adjudication process according to an embodiment of the present invention.

[0033] Figure 5 This is a block diagram of a battery cathode material evidence question-and-answer system for verifying the comparability of experimental conditions, according to an embodiment of the present invention. Detailed Implementation

[0034] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0035] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0036] Example 1

[0037] This embodiment provides a question-and-answer method for evidence of battery cathode materials for comparability verification under experimental conditions, including:

[0038] Acquire multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, and construct a condition-enhanced evidence map. The condition-enhanced evidence map includes evidence units organized by material object, performance index, experimental value, experimental conditions and source information.

[0039] The system receives a natural language question input by the user and parses the natural language question to extract the target material object, target performance index, and experimental condition constraints.

[0040] Based on the target material object, the target performance index, and the experimental condition constraints, candidate evidence units are retrieved from the condition-enhanced evidence map.

[0041] The experimental conditions in the candidate evidence units are normalized, and the normalized experimental conditions are compared with the experimental condition constraints to obtain a set of comparable evidence.

[0042] The comparable evidence set is subjected to consistency analysis, source credibility assessment and conflict resolution to generate a structured adjudication result;

[0043] Based on the structured adjudication results, a question-and-answer result and evidence description are generated and sent to the user.

[0044] As one implementation method in this embodiment, the process of constructing the condition-enhanced evidence map includes:

[0045] Text cleaning was performed on the obtained literature data on multi-source battery cathode materials to obtain preprocessed corpus;

[0046] Material names, structural properties, performance indicators, experimental values, test conditions, preparation processes, characterization methods, and source literature identifiers are extracted from the preprocessed corpus to construct evidence units;

[0047] The entity names, unit expressions, and condition descriptions in the evidence unit are disambiguated, standardized, and aligned to form a unified structured expression;

[0048] The standardized evidence units and their relationships with material objects, performance indicators, experimental conditions, and source information are stored in a graph database to complete the construction of the condition-enhanced evidence graph.

[0049] As one implementation method in this embodiment, the process of parsing the natural language problem to extract the target material object, target performance indicators, and experimental condition constraints includes:

[0050] Identify the material objects, target performance indicators, comparative relationships, numerical constraints, and experimental condition descriptions in the natural language problem.

[0051] The experimental conditions are described and analyzed as including constraints in temperature, voltage window, multiplier, number of cycles, preparation process, and characterization method.

[0052] Based on the parsing results, a structured query instruction containing the target object, target indicators, and conditional constraints is generated.

[0053] As one implementation method in this embodiment, the process of normalizing the experimental conditions in the candidate evidence units includes:

[0054] A unified dimensional conversion was performed on the temperature, voltage window, multiplier, number of cycles, capacity unit, and energy unit in the candidate evidence unit.

[0055] Synonymous mapping and standardized replacement are performed on heterogeneous expressions representing the same experimental semantics in different documents;

[0056] Experimental conditions containing interval expressions, threshold expressions, or default expressions are padded, split, or standardized to obtain a normalized set of experimental conditions.

[0057] As one implementation method in this embodiment, the process of verifying the comparability of the normalized experimental conditions with the experimental condition constraints includes:

[0058] Determine whether the experimental conditions corresponding to the candidate evidence unit satisfy the following conditions: complete consistency, range inclusion, threshold compatibility, or incomparability with the experimental conditions constraints in the user question;

[0059] Candidate evidence units that satisfy the conditions of complete consistency or range inclusion relationship are retained as high-priority evidence;

[0060] For candidate evidence units that meet the threshold compatibility relationship, weight reduction processing is performed according to preset rules;

[0061] For candidate evidence units that are not comparable, a removal process is performed to obtain a set of comparable evidence.

[0062] As one implementation method in this embodiment, the process of performing consistency analysis, source credibility assessment, and conflict resolution on the comparable evidence set includes:

[0063] For multiple evidence units pointing to the same material object and the same performance index, extract the literature source type identifier, experimental condition field completeness, test method field completeness, cross-corroboration degree and time decay factor, and calculate the source credibility value of each evidence unit.

[0064] For numerical performance indicators, the comprehensive weight of each evidence unit is calculated based on the condition matching degree, source credibility, test method consistency and time priority, and the deviation degree is calculated based on the weighted reference value of the experimental results corresponding to each evidence unit to determine whether each evidence unit belongs to consistent evidence or conflicting evidence.

[0065] Based on the number of valid evidence, the proportion of consistent evidence, the proportion of conflicting evidence, and the completeness of key experimental condition fields, a supporting ruling, a disputed ruling, or a ruling due to insufficient evidence is generated.

[0066] The purpose of this invention is to provide a question-and-answer method and system for battery cathode materials evidence oriented towards the comparability verification of experimental conditions. By organizing experimental facts from material literature into evidence, and combining experimental condition normalization, comparability verification, evidence screening, conflict resolution, and structured adjudication mechanisms, it solves the problems in existing technologies where evidence from different sources is difficult to directly compare, and question-and-answer results are easily detached from the specific experimental context. For example... Figure 1 As shown, the specific steps include the following:

[0067] S1. Obtain multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, construct evidence units containing material objects, performance indicators, experimental values, experimental conditions and source information, and construct condition-enhanced evidence maps based on the evidence units.

[0068] S2. Receive the natural language question input by the user, perform question parsing and constraint extraction on the natural language question to identify the target material object, target performance index and experimental condition constraints, and generate a structured query instruction;

[0069] S3. Send the structured query instruction to the evidence retrieval module to retrieve candidate evidence units related to the problem in the pre-constructed condition-enhanced evidence graph;

[0070] S4. Normalize the experimental conditions in the candidate evidence unit, and verify the comparability of the normalized experimental conditions with the experimental condition constraints to obtain a set of comparable evidence.

[0071] S5. Perform consistency analysis, source credibility assessment and conflict resolution on the comparable evidence set to generate a structured adjudication result;

[0072] S6. Generate a question-and-answer result and evidence description based on the structured adjudication result and send them to the user;

[0073] S7. Receive feedback information from users based on manual verification, and update the evidence unit, experimental condition mapping rules, or credibility evaluation parameters according to the feedback information to complete the iterative optimization of the condition-enhanced evidence map.

[0074] Furthermore, the condition-enhanced evidence graph in this invention serves as the foundational data structure for verifying the comparability of experimental conditions and answering evidence questions. Unlike static graphs that merely express entity relationships, it organizes experimental facts from literature into evidence units with experimental context and source information, explicitly linking material objects, performance indicators, experimental values, experimental conditions, and source literature at the graph level. This provides structured support for subsequent evidence screening and conclusion adjudication. Figure 2 As shown, the construction process of the conditional enhancement evidence map includes the following steps:

[0075] S11. Multi-source heterogeneous data aggregation and preprocessing:

[0076] First, raw literature data is obtained from multi-source data resources in the field of battery cathode materials. This raw literature data includes, but is not limited to, journal articles, patent texts, conference papers, publicly available experimental datasets, and technical reports. The raw literature data undergoes page layout parsing, text extraction, noise removal, and text cleaning processes, removing references, headers and footers, author information, and irrelevant descriptions to form a preprocessed corpus.

[0077] In one embodiment, for PDF documents, layout recognition and paragraph recombination can be used to extract the main text; for patent texts, abstracts, embodiments, and experimental data paragraphs can be extracted; for tabular data, material names, performance values, and experimental conditions fields can be retained for subsequent structured extraction.

[0078] S12. Evidence Unit Extraction and Structured Modeling:

[0079] Information such as material name, structural properties, performance indicators, experimental values, testing conditions, preparation process, characterization method, and source literature identification is extracted from the preprocessed corpus and organized into evidence units. Each evidence unit is used to characterize a citationable experimental fact from the material literature and includes at least: the material object, performance indicators, experimental values, experimental conditions, and source information.

[0080] In one embodiment, if a document states that "the capacity retention of NCM811 material is 91.3% after 100 cycles within a voltage window of 2.5-4.2V", then this can be constructed as an evidence unit, wherein the material object is NCM811, the performance index is the capacity retention, the experimental value is 91.3%, the experimental conditions include a voltage window of 2.5-4.2V and 100 cycles, and the source information is the corresponding document identifier.

[0081] S13. Standardization of Evidence Units:

[0082] The entity names, unit expressions, and condition descriptions in the aforementioned evidence units undergo disambiguation, normalization, and alignment processing to reduce interference from synonymous and heterogeneous expressions in multi-source documents on subsequent evidence comparisons. The normalization processing includes, but is not limited to: material name normalization, unit unification, condition field standardization, and synonym mapping.

[0083] For example, “LiNi0.8Co0.1Mn0.1O2” and “NCM811” can be mapped to the same material object; “roomtemperature” and “25℃” can be mapped to a unified temperature representation; and “0.5 C” and “0.5C ratio” can be mapped to a unified ratio field expression.

[0084] S14. Construction of Conditional Enhancement Evidence Map:

[0085] The standardized evidence units and their relationships with material objects, performance indicators, experimental conditions, and source information are stored in a graph database to form a condition-enhanced evidence graph. This condition-enhanced evidence graph includes material object nodes, performance indicator nodes, experimental condition nodes, and source nodes, as well as evidence unit nodes that carry specific experimental facts. The nodes are connected through preset relationships to support subsequent evidence retrieval and adjudication analysis based on experimental condition constraints.

[0086] Furthermore, when using the method provided by this invention, the user can input a natural language question. For example, the user inputs: "Please compare the cycle stability of NCM811 and NCM622 in the 2.5-4.2V voltage window." After receiving this input, the system enters stage S2 to parse the question and extract constraints, as follows:

[0087] S21. Target object recognition:

[0088] Identify material objects, performance indicators, and comparative relationships in natural language questions. In the example above, the material objects can be identified as NCM811 and NCM622, the performance indicator as cyclic stability, and the question type as a comparative question-and-answer session.

[0089] S22. Extraction of experimental constraints:

[0090] Experimental constraints are extracted from natural language questions and converted into structured condition fields. In the example above, the voltage window constraint of 2.5-4.2V can be extracted. If the question also includes descriptions such as temperature, multiplier, number of cycles, fabrication process, or characterization method, further extraction can be performed to form composite constraints.

[0091] S23. Generation of structured query instructions:

[0092] Structured query instructions are generated based on the target object, performance indicators, and experimental constraints. These structured query instructions can take the form of graph database query statements, structured search templates, or internal search expressions, and are used to retrieve candidate evidence units relevant to the problem within the conditionally enhanced evidence graph.

[0093] Furthermore, the evidence retrieval process is performed in S3. Unlike fragment-based recall of unstructured text, this invention performs structured retrieval around material objects, performance indicators, and experimental conditions within a conditionally enhanced evidence graph to obtain a set of candidate evidence units directly related to the user's question.

[0094] In one embodiment, for the question of “comparing the cycling stability of NCM811 and NCM622 in the 2.5-4.2V voltage window”, the system can retrieve evidence units related to NCM811 and NCM622 and whose performance indicators are related to cycling stability, and summarize the evidence units that meet the preliminary matching conditions into a candidate evidence set.

[0095] Furthermore, experimental condition normalization and comparability checks are performed in S4. For example... Figure 3 As shown, this step is a crucial aspect that distinguishes this invention from general question-and-answer methods. Its core lies in determining whether candidate evidence units can be used to answer the current question in the same or similar experimental context. The specific process includes the following:

[0096] S41. Normalization of experimental conditions:

[0097] The temperature, voltage window, multiplier, number of cycles, capacity unit, and energy unit in the candidate evidence unit are uniformly converted to a unified dimension; heterogeneous expressions representing the same experimental semantics in different literature are synonymized and standardized and replaced; interval expressions, threshold expressions, or default expressions are completed, split, or standardized and encoded to obtain a normalized set of experimental conditions.

[0098] For example, “2.50–4.20 V” and “2.5~4.2V” can be uniformly encoded into the same voltage window expression; “100 cycles” and “100 cycles” can be uniformly represented by the same field.

[0099] S42. Determination of comparability:

[0100] The normalized experimental conditions are compared with the experimental condition constraints in the user problem to determine whether they satisfy the conditions of complete consistency, range inclusion, threshold compatibility, or incomparability.

[0101] Among them, complete consistency means that the key experimental conditions of the candidate evidence unit are consistent with the problem constraints; range inclusion means that the condition range of the candidate evidence unit covers or is covered by the problem constraints, and can be regarded as a valid comparison under the preset rules; threshold compatibility means that although the candidate evidence units are not completely consistent, the condition differences are within the allowable tolerance range; incomparable means that the condition differences exceed the allowable range, or the key condition fields are missing, and cannot support a valid comparison.

[0102] S43. Generation of a comparable evidence set:

[0103] Candidate evidence units that meet the criteria of complete consistency or range inclusion are retained as high-priority evidence; candidate evidence units that meet the criteria of threshold compatibility are downgraded according to preset rules; and candidate evidence units that are incomparable are eliminated. The final set of comparable evidence is obtained.

[0104] For example, when the problem constraint is a voltage window of 2.5-4.2V, if the test condition corresponding to a certain evidence unit is 2.5-4.2V, it can be directly retained; if its test condition is 2.6-4.2V and the preset rules allow a certain voltage deviation, it can be retained with reduced weight; if its test condition is 3.0-4.5V, it can be determined to be incomparable and eliminated due to the large difference.

[0105] Furthermore, in S5, consistency analysis, source credibility assessment, and conflict resolution are performed on the comparable evidence set. For example... Figure 4 As shown, this step is used to generate a structured, outputtable adjudication result among multiple comparable pieces of evidence. The specific process is as follows:

[0106] S51. Calculation of conditional matching degree:

[0107] For multiple evidence units pointing to the same material object and the same performance index, the condition matching degree between each evidence unit and the problem constraints is calculated. The condition matching degree can be comprehensively calculated based on the matching degree of temperature, voltage window, multiplier, number of cycles, and other key condition fields.

[0108] S52. Source Credibility Assessment:

[0109] The source credibility of each evidence unit is assessed based on the literature source type identifier, the completeness of the experimental conditions field, the completeness of the test method field, the cross-corroboration degree, and the time decay factor.

[0110] Specifically, for the i-th evidence unit, the system extracts the document source type identifier from its source metadata. Completeness of experimental condition fields Test method field completeness Cross-corroboration and time decay factor The various indicators were then normalized to the range of 0 to 1 to obtain the source credibility value. :

[0111]

[0112] in, .

[0113] Among them, document source type identifier Determined based on the source rule table; completeness of the experimental condition field. The completeness of the test method fields is obtained by comparing the number of key experimental condition fields extracted from the evidence unit with the total number of pre-defined required fields. Determined by the completeness of the test method fields; cross-validation degree Calculated from the number of independent sources of evidence supporting the same conclusion; time decay factor It is calculated from the time difference between the publication time of the document and the current system time.

[0114] In one embodiment, the time decay factor It can be represented as:

[0115]

[0116] in, This represents the time difference between the document corresponding to the i-th evidence unit and the current time. This is the time decay coefficient.

[0117] S53. Consistency Analysis and Conflict Identification:

[0118] Consistency analysis is performed on multiple evidence units corresponding to the same material object and the same performance index.

[0119] For numerical performance indicators, suppose there are n valid evidence units after comparability verification, and the corresponding experimental results are as follows: The system first calculates the comprehensive weight of each evidence unit based on condition matching degree, source credibility, consistency of testing methods, and time priority. Then calculate the weighted reference value. :

[0120]

[0121] Wherein, the deviation of the i-th evidence unit relative to the weighted reference value Represented as:

[0122]

[0123] in, To prevent extremely small positive numbers with a denominator of zero.

[0124] when When this point is reached, the evidence unit is deemed consistent with the overall conclusion.

[0125] when At that time, the evidence unit was determined to be weak conflict evidence;

[0126] when In such cases, the evidence unit is determined to be conflicting evidence.

[0127] For comparative problems, if the performance results of the two sets of comparison objects corresponding to the i-th evidence unit are respectively and Then define the direction discrimination value. for:

[0128]

[0129] When the majority of valid evidence units correspond to The values ​​are the same, and the weighted percentage corresponding to the value is not lower than the preset threshold. When determining that the experimental conclusions of multiple evidence units show a consistent trend;

[0130] When in different directions When a large number of conflicting pieces of evidence appear simultaneously, and the weighted proportion of opposing evidence exceeds a preset threshold, conflicting evidence is determined to exist.

[0131] S54. Generation of Structured Decision Results:

[0132] Based on condition matching degree, source credibility, test method consistency and time priority, multiple evidence units are comprehensively scored and ranked.

[0133] For the i-th unit of evidence, define the comprehensive adjudication score. for:

[0134]

[0135] in, Indicates the degree of condition matching. Indicates the credibility of the source. This indicates the consistency score of the test method. This indicates the time priority score, and .

[0136] After completing the comprehensive scoring, the system generates a structured ruling based on the number of valid evidence, the proportion of consistent evidence, the proportion of conflicting evidence, and the completeness of key condition fields:

[0137] When the number of valid evidence units is not less than the preset lower limit, the weighted proportion of consistent evidence is not less than the first threshold, and there is no significant conflict among the high-priority evidence, a supportive ruling is generated.

[0138] When the number of valid evidence units is not less than the preset lower limit, the proportion of consistent evidence reaches the second threshold, but only some of the high-priority evidence meets the complete consistency condition, and the remaining evidence is mainly threshold-compatible or scope-inclusive evidence, a condition-supporting ruling result is generated.

[0139] When the weighted proportion of conflicting evidence is higher than the third threshold, or when opposing evidence exists simultaneously in a comparative question and both reach the preset lower weight limit, a disputed ruling is generated, along with an explanation of the source of the conflict.

[0140] When the number of valid evidence units is lower than a preset lower limit, or the missing rate of key experimental condition fields is higher than a preset threshold, resulting in the inability to form a stable ruling, an insufficient evidence ruling result is generated.

[0141] In one embodiment, the structured adjudication result includes at least one or more of the following: conclusion content, applicable experimental conditions, source of evidence, confidence level, conflict explanation, and insufficient evidence warning.

[0142] Furthermore, in S6, question-and-answer results and evidence descriptions are generated based on the structured adjudication results and sent to the user. The question-and-answer results may be presented in the form of natural language responses, table comparison results, conditional matching summaries, or conclusion classification descriptions. The question-and-answer results may be categorized into one of the following: explicitly supported, conditionally supported, disputed, or insufficient evidence.

[0143] For example, regarding the aforementioned issue, if a large number of highly credible evidences are found under the 2.5-4.2V condition indicating that NCM811 has a higher capacity retention rate than NCM622 under the same number of cycles, the system can output "The existing comparable evidence as a whole supports that NCM811 has better cycling stability than NCM622 under the 2.5-4.2V voltage window"; if only some high-priority evidence meets the complete consistency condition, and the remaining evidence mainly meets the range inclusion relationship or threshold compatibility relationship, the system can output "The existing evidence provides conditional support for the superior cycling stability of NCM811 compared to NCM622 under the given experimental conditions"; if there are significant differences in the relevant evidence in different literatures, the system can output "The existing evidence is controversial, and the controversy mainly stems from the differences in sample preparation processes and cycling test conditions in different literatures"; if the number of valid evidences is insufficient or many key experimental condition fields are missing, the system can output "The existing evidence is insufficient, and it is not yet possible to form a stable decision under the given experimental conditions".

[0144] It should be noted that, in this invention, the generation of question-and-answer results should be based on the aforementioned structured adjudication results and expressed in a limited manner. That is, the question-and-answer output should be based on the comparable evidence set and the adjudication results, and should not be freely expanded beyond the scope of the verified evidence, so as to ensure that the output content has a clear correspondence with the screened evidence.

[0145] Furthermore, the feedback optimization process is completed in S7. After the structured question-and-answer results and evidence descriptions are displayed to the user, the user can provide feedback on the question-and-answer content, experimental condition matching relationships, evidence sources, and conflict identification results based on the manual verification results. After receiving the feedback information, the system updates the evidence unit labels, experimental condition mapping rules, comparability verification thresholds, or credibility assessment parameters to improve the evidence screening and adjudication capabilities in subsequent question-and-answer processes.

[0146] For example, when a user confirms that "room temperature" in a certain document should be uniformly mapped to 25℃ in the current application scenario, or confirms that a certain type of conference abstract should not be used as a high-credibility source, the system can update the conditional mapping rules or source credibility parameters accordingly, thereby completing the iterative optimization of the conditional enhancement evidence map.

[0147] Based on this, the present invention provides a battery cathode material evidence question-and-answer method for verifying the comparability of experimental conditions, which discloses the following technical effects:

[0148] (1) Improve the accuracy of evidence use and the reliability of conclusion output: This invention introduces experimental condition normalization and comparability verification mechanism to structurally identify and screen the differences in experimental conditions in literature from different sources, avoids mixing evidence with inconsistent conditions for the same question and answer conclusion, thereby significantly improving the accuracy and reliability of the material question and answer results.

[0149] (2) Enhance the ability to constrain conditions in the material question and answer process: In view of the high sensitivity of the performance conclusions of battery cathode materials to test temperature, voltage window, rate, number of cycles and preparation process, the present invention introduces experimental condition constraints in the evidence retrieval and screening process, so that the question and answer results can correspond to the specific experimental context, thereby improving the adaptability of material question and answer to condition-sensitive questions.

[0150] (3) Improve the interpretability and auditability of the results: While outputting the question and answer results, the present invention can simultaneously provide the applicable experimental conditions, source evidence, conflict explanations and credibility levels, so that researchers can verify and trace the basis for the formation of the answers, thereby improving the transparency and auditability of the question and answer results.

[0151] (4) Possesses strong evidence conflict handling and continuous optimization capabilities: For numerical differences or conclusion conflicts of the same material object and the same performance index in different literature, the present invention can form a structured adjudication result through consistency analysis, source credibility assessment and conflict resolution mechanism; at the same time, it can update the evidence unit and rule parameters by combining manual verification feedback, thereby improving the system's continuous optimization capability and adaptability to new materials and new data scenarios.

[0152] Example 2

[0153] In this embodiment, a computer terminal device is provided, including:

[0154] One or more processors;

[0155] A memory, coupled to the processor, for storing one or more programs;

[0156] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the above-described battery cathode material evidence question-and-answer method for comparability verification under experimental conditions.

[0157] In this embodiment, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the above-described battery cathode material evidence question-and-answer method for comparability verification under experimental conditions.

[0158] In this embodiment, an electronic device is also provided, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the steps of the above-described battery cathode material evidence question-and-answer method for comparability verification under experimental conditions.

[0159] In this embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the above-described battery cathode material evidence question-and-answer method for comparability verification under experimental conditions.

[0160] The aforementioned program can run on a processor or be stored in memory (or a computer-readable medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0161] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented by different modules for different steps.

[0162] like Figure 5 As shown, this embodiment provides such a device or system. This system, referred to as a battery cathode material evidence question-and-answer system for comparability verification under experimental conditions, includes:

[0163] The evidence graph construction module is used to acquire multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, construct evidence units containing material objects, performance indicators, experimental values, experimental conditions and source information, and construct condition-enhanced evidence graphs based on the evidence units.

[0164] The problem parsing module is used to receive natural language questions input by users, parse the natural language questions to extract target material objects, target performance indicators and experimental condition constraints;

[0165] The evidence retrieval module is used to retrieve candidate evidence units from the condition-enhanced evidence map based on the target material object, the target performance index and the experimental condition constraints.

[0166] The condition normalization module is used to normalize the experimental conditions in the candidate evidence units;

[0167] The comparability verification module is used to perform comparability verification between the normalized experimental conditions and the experimental condition constraints to obtain a set of comparable evidence.

[0168] The evidence adjudication module is used to perform consistency analysis, source credibility assessment and conflict resolution on the comparable evidence set, and generate structured adjudication results;

[0169] The question-and-answer output module is used to generate question-and-answer results and evidence descriptions based on the structured adjudication results and send them to the user.

[0170] As one implementation method in this embodiment, the evidence map construction module includes:

[0171] The text cleaning unit is used to perform text cleaning on the acquired literature data of multi-source battery cathode materials to obtain preprocessed corpus.

[0172] The evidence extraction unit is used to extract material names, structural properties, performance indicators, experimental values, test conditions, preparation processes, characterization methods, and source document identifiers from the preprocessed corpus to construct evidence units.

[0173] The standardization processing unit is used to disambiguate, standardize, and align the entity names, unit expressions, and condition descriptions in the evidence unit to form a unified structured expression.

[0174] The graph storage unit is used to store the standardized evidence units and their relationships with the material object, performance indicators, experimental conditions and source information into the graph database, thereby completing the construction of the condition-enhanced evidence graph.

[0175] As one implementation method in this embodiment, the problem analysis module includes:

[0176] The problem identification unit is used to identify the material objects, target performance indicators, comparison relationships, numerical constraints, and experimental condition descriptions in the natural language problem.

[0177] The constraint resolution unit is used to resolve the experimental conditions description into one or more constraint terms including temperature, voltage window, multiplier, number of cycles, preparation process, and characterization method.

[0178] The query generation unit is used to generate structured query instructions containing target objects, target indicators, and conditional constraints based on the parsing results.

[0179] As one implementation method in this embodiment, the condition normalization module includes:

[0180] The dimension conversion unit is used to perform unified dimension conversion on temperature, voltage window, multiplier, number of cycles, capacity unit and energy unit in the candidate evidence unit;

[0181] Synonym mapping unit, used to perform synonym mapping and standardized replacement on heterogeneous expressions representing the same experimental semantics in different documents;

[0182] The standardized coding unit is used to complete, split, or standardize the experimental conditions that contain interval expressions, threshold expressions, or default expressions, so as to obtain a normalized set of experimental conditions.

[0183] As one implementation method in this embodiment, the comparability verification module includes:

[0184] The relationship determination unit is used to determine whether the experimental conditions corresponding to the candidate evidence unit and the experimental condition constraints in the user question satisfy the relationships of complete consistency, range inclusion, threshold compatibility, or incomparability.

[0185] The priority processing unit is used to retain candidate evidence units that meet the complete consistency or range inclusion relationship as high-priority evidence, perform downweighting processing on candidate evidence units that meet the threshold compatibility relationship according to preset rules, and perform elimination processing on candidate evidence units that belong to the incomparable relationship to obtain a set of comparable evidence.

[0186] As one implementation method in this embodiment, the evidence adjudication module includes:

[0187] The credibility assessment unit is used to extract the literature source type identifier, experimental condition field completeness, test method field completeness, cross-corroboration degree, and time decay factor for multiple evidence units pointing to the same material object and the same performance index, and calculate the source credibility value of each evidence unit.

[0188] The consistency analysis unit is used to calculate the comprehensive weight of each evidence unit for numerical performance indicators based on condition matching degree, source credibility, test method consistency and time priority, and to calculate the deviation degree based on the weighted reference value of the corresponding experimental results of each evidence unit, so as to determine whether each evidence unit belongs to consistent evidence or conflicting evidence.

[0189] The adjudication result generation unit is used to generate supportive adjudication results, disputed adjudication results, or insufficient evidence adjudication results based on the number of valid evidence, the proportion of consistent evidence, the proportion of conflicting evidence, and the completeness of key experimental condition fields.

[0190] The system or apparatus is used to implement the functions of the methods in the above embodiments. Each module in the system or apparatus corresponds to each step in the method, as has been described in the method and will not be repeated here.

[0191] The above implementation method solves the problem of evidence questioning for battery cathode materials in the related technology for comparability verification of experimental conditions, thereby ensuring that the problems existing in the prior art are resolved.

[0192] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A question-and-answer method for evidence of battery cathode materials for verifying comparability under experimental conditions, characterized in that, Includes the following steps: Acquire multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, and construct a condition-enhanced evidence map. The condition-enhanced evidence map includes evidence units organized by material object, performance index, experimental value, experimental conditions and source information. The system receives a natural language question input by the user and parses the natural language question to extract the target material object, target performance index, and experimental condition constraints. Based on the target material object, the target performance index, and the experimental condition constraints, candidate evidence units are retrieved from the condition-enhanced evidence map. The experimental conditions in the candidate evidence units are normalized, and the normalized experimental conditions are compared with the experimental condition constraints to obtain a set of comparable evidence. The comparable evidence set is subjected to consistency analysis, source credibility assessment and conflict resolution to generate a structured adjudication result; Based on the structured adjudication results, a question-and-answer result and evidence description are generated and sent to the user.

2. The method according to claim 1, characterized in that, The process of constructing the condition-enhanced evidence map includes: Text cleaning was performed on the obtained literature data on multi-source battery cathode materials to obtain preprocessed corpus; Material names, structural properties, performance indicators, experimental values, test conditions, preparation processes, characterization methods, and source literature identifiers are extracted from the preprocessed corpus to construct evidence units; The entity names, unit expressions, and condition descriptions in the evidence unit are disambiguated, standardized, and aligned to form a unified structured expression; The standardized evidence units and their relationships with material objects, performance indicators, experimental conditions, and source information are stored in a graph database to complete the construction of the condition-enhanced evidence graph.

3. The method according to claim 1, characterized in that, The process of parsing natural language questions to extract target material objects, target performance indicators, and experimental constraints includes: Identify the material objects, target performance indicators, comparative relationships, numerical constraints, and experimental condition descriptions in the natural language problem. The experimental conditions are described and analyzed as including constraints in temperature, voltage window, multiplier, number of cycles, preparation process, and characterization method. Based on the parsing results, a structured query instruction containing the target object, target indicators, and conditional constraints is generated.

4. The method according to claim 1, characterized in that, The process of normalizing the experimental conditions in the candidate evidence units includes: A unified dimensional conversion was performed on the temperature, voltage window, multiplier, number of cycles, capacity unit, and energy unit in the candidate evidence unit. Synonymous mapping and standardized replacement are performed on heterogeneous expressions representing the same experimental semantics in different documents; Experimental conditions containing interval expressions, threshold expressions, or default expressions are padded, split, or standardized to obtain a normalized set of experimental conditions.

5. The method according to claim 1, characterized in that, The process of comparing the normalized experimental conditions with the experimental condition constraints includes: Determine whether the experimental conditions corresponding to the candidate evidence unit satisfy the following conditions: complete consistency, range inclusion, threshold compatibility, or incomparability with the experimental conditions constraints in the user question; Candidate evidence units that satisfy the conditions of complete consistency or range inclusion relationship are retained as high-priority evidence; For candidate evidence units that meet the threshold compatibility relationship, weight reduction processing is performed according to preset rules; For candidate evidence units that are not comparable, a removal process is performed to obtain a set of comparable evidence.

6. The method according to claim 1, characterized in that, The process of performing consistency analysis, source credibility assessment, and conflict resolution on comparable evidence sets includes: For multiple evidence units pointing to the same material object and the same performance index, extract the literature source type identifier, experimental condition field completeness, test method field completeness, cross-corroboration degree and time decay factor, and calculate the source credibility value of each evidence unit. For numerical performance indicators, the comprehensive weight of each evidence unit is calculated based on the condition matching degree, source credibility, test method consistency and time priority, and the deviation degree is calculated based on the weighted reference value of the experimental results corresponding to each evidence unit to determine whether each evidence unit belongs to consistent evidence or conflicting evidence. Based on the number of valid evidence, the proportion of consistent evidence, the proportion of conflicting evidence, and the completeness of key experimental condition fields, a supporting ruling, a disputed ruling, or a ruling due to insufficient evidence is generated.

7. A question-and-answer system for evidence of battery cathode materials for verifying comparability under experimental conditions, characterized in that, The system for implementing the method of any one of claims 1-6 comprises: The evidence graph construction module is used to acquire multi-source heterogeneous literature data related to battery cathode materials, extract and structure the information from the multi-source heterogeneous literature data, construct evidence units containing material objects, performance indicators, experimental values, experimental conditions and source information, and construct condition-enhanced evidence graphs based on the evidence units. The problem parsing module is used to receive natural language questions input by users, parse the natural language questions to extract target material objects, target performance indicators and experimental condition constraints; The evidence retrieval module is used to retrieve candidate evidence units from the condition-enhanced evidence map based on the target material object, the target performance index and the experimental condition constraints. The condition normalization module is used to normalize the experimental conditions in the candidate evidence units; The comparability verification module is used to perform comparability verification between the normalized experimental conditions and the experimental condition constraints to obtain a set of comparable evidence. The evidence adjudication module is used to perform consistency analysis, source credibility assessment and conflict resolution on the comparable evidence set, and generate structured adjudication results; The question-and-answer output module is used to generate question-and-answer results and evidence descriptions based on the structured adjudication results and send them to the user.

8. A computer terminal device, characterized in that, include: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images