Spectral database query model construction method and spectral database query method

Abstract

The present application relates to a spectral database query model construction method and a spectral database query method, which obtains a molecular spectral database; based on the molecular spectral database, constructs a natural language query template and an SQL query statement template corresponding to the natural language query template; constructs a query data set according to the natural language query template and the SQL query statement template; obtains a fine-tuned vertical field big model; based on the query data set, trains the fine-tuned vertical field big model to obtain a spectral database query model; based on this, the spectral database query model can better understand and process the language features of the astrochemical field, thereby laying a foundation for improving the professionalism and accuracy of the query results.

Description

Technical Field

[0001] The present application relates to the technical field of database query, and in particular to a spectral database query model construction method and a spectral database query method. Background Art

[0002] Molecular spectral databases play a vital role in the study of interstellar media. By analyzing the line width, line intensity and profile of molecular spectra, we can infer the physical parameters of the interstellar medium, such as temperature, density, momentum distribution, etc., and then study the spectral characteristics of molecules in the interstellar medium, which helps to reveal the chemical composition, physical state and dynamic process of the interstellar medium. Traditionally, the spectral indexing method is usually used to query the molecular spectral database.

[0003] However, traditional spectral indexing methods suffer from the problem of low query result accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a spectral database query model construction method and a spectral database query method to address the above technical problems.

[0005] In a first aspect, the present application provides a method for constructing a spectral database query model, wherein the spectral database query model is applied to spectral database query, and the method comprises:

[0006] Access to molecular spectroscopy databases;

[0007] Based on the molecular spectroscopy database, construct a natural language query template and an SQL query statement template corresponding to the natural language query template;

[0008] Constructing a query data set according to the natural language query template and the SQL query statement template;

[0009] Obtain a fine-tuned vertical domain model;

[0010] Based on the query data set, the fine-tuned vertical domain large model is trained to obtain a spectral database query model.

[0011] In one embodiment, obtaining a fine-tuned vertical domain macro model includes:

[0012] Get astrochemical text data;

[0013] Based on the astrochemical text data, the vertical field big model is fine-tuned to obtain a fine-tuned vertical field big model.

[0014] In one embodiment, the molecular spectroscopy database includes spectral parameter data; and constructing a query data set according to the natural language query template and the SQL query statement template includes:

[0015] Based on the spectral parameter data, construct a random parameter generation function;

[0016] Generate a random variable according to the random parameter generation function;

[0017] Substituting the random variables into the natural language query template and the SQL query statement template respectively to generate multiple natural language query conditions and corresponding multiple SQL query statements;

[0018] A query data set is constructed according to the multiple natural language query conditions and the corresponding multiple SQL query statements.

[0019] In one embodiment, the step of acquiring a molecular spectroscopy database comprises:

[0020] Obtaining raw spectral data from multiple different spectral databases;

[0021] Performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data;

[0022] A molecular spectrum database is constructed according to the molecular spectrum data.

[0023] In one embodiment, performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data includes:

[0024] Performing data deduplication processing on repeated data corresponding to the same molecule in the original spectral data to obtain deduplication spectral data;

[0025] Determine parameter column information based on multiple different spectral databases;

[0026] Determine data missing parameter columns and shared parameter columns according to the deduplicated spectral data and the parameter column information;

[0027] Based on the common parameter column, performing data filling processing on the data missing parameter column to obtain reconstructed spectral data;

[0028] The deduplicated spectral data and the reconstructed spectral data are integrated to obtain molecular spectral data.

[0029] In one embodiment, the fine-tuned vertical domain large model is trained based on the query data set to obtain a spectral database query model, including:

[0030] Preprocessing the natural language query conditions in the query data set to obtain corresponding multiple index sequences;

[0031] According to the plurality of index sequences, the fine-tuned vertical domain large model is trained until the target loss function meets the preset convergence condition, thereby obtaining a spectral database query model.

[0032] In one embodiment, the method further comprises:

[0033] Obtaining multiple test natural language query conditions and multiple correct query results corresponding to the multiple test natural language query conditions;

[0034] Generate a plurality of corresponding test SQL query statements according to a plurality of test natural language query conditions through the spectrum database query model;

[0035] Based on the molecular spectroscopy database and the plurality of test SQL query statements, obtaining a plurality of corresponding test query results;

[0036] Determining an exact match degree according to the plurality of correct query results and the plurality of test query results;

[0037] If the degree of exact match does not reach the preset accuracy threshold, the spectrum database query model is optimized until the degree of exact match reaches the preset accuracy threshold.

[0038] In a second aspect, the present application also provides a spectrum database query method, the method comprising:

[0039] Inputting the target natural language input by the user into the spectral database query model to obtain a target query statement corresponding to the target natural language;

[0040] Based on the molecular spectrum database, generating corresponding target query results by executing the target query statement;

[0041] The method further comprises:

[0042] Generate a molecular spectrum diagram according to the target query result, and display it on the spectrum database query display interface;

[0043] The method further comprises:

[0044] Based on the spectral database query model and the target query result, molecular interpretation information corresponding to the target query result is generated and displayed on the spectral database query display interface.

[0045] In a third aspect, the present application further provides a spectral database query display interface, wherein the spectral database query display interface is provided with a user input interface, a query result display interface, a spectral graph display interface, and a molecular interpretation information display interface;

[0046] The user input interface is used to provide an input port for the target natural language;

[0047] The query result display interface is used to display the target query result; the target query result is a target query statement corresponding to the target natural language obtained by inputting the target natural language input by the user into the spectrum database query model; the target query result is generated by executing the target query statement based on the molecular spectrum database;

[0048] The spectrum display interface is used to display a molecular spectrum; the molecular spectrum is a molecular spectrum generated according to the target query result;

[0049] The molecular interpretation information display interface is used to display molecular interpretation information; the molecular interpretation information is molecular interpretation information corresponding to the target query result, which is generated based on the spectral database query model and the target query result.

[0050] In a fourth aspect, the present application also provides a spectrum database query system, the system comprising: a user input module, a query result generation module, a spectrum graph generation module and a molecular interpretation information generation module;

[0051] The user input module is used to receive a target natural language input by a user;

[0052] The query result generation module is used to input the target natural language input by the user into the spectrum database query model to obtain a target query statement corresponding to the target natural language; based on the molecular spectrum database, generate a corresponding target query result by executing the target query statement;

[0053] The spectrum graph generation module is used to generate a molecular spectrum graph according to the target query result and display it on the spectrum database query display interface;

[0054] The molecular interpretation information generation module is used to generate molecular interpretation information corresponding to the target query result based on the spectrum database query model and the target query result, and display it on the spectrum database query display interface.

[0055] In a fifth aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in any one of the embodiments of the first and second aspects are implemented.

[0056] In a sixth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in any one of the embodiments of the first and second aspects above are implemented.

[0057] In a seventh aspect, the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the method described in any one of the embodiments of the first and second aspects above.

[0058] The above-mentioned spectral database query model construction method and spectral database query method; the spectral database query model is applied to spectral database query, and based on the molecular spectral database, a natural language query template and an SQL query statement template corresponding to the natural language query template are constructed; then, according to the natural language query template and the SQL query statement template, a query data set is constructed, which not only enhances the professionalism of the query data set, but also improves the quality of the query data set, laying a foundation for improving the accuracy of the spectral database query model; then, based on the query data set, the fine-tuned vertical field large model is trained to obtain the spectral database query model, which can enable the spectral database query model to better understand and process the language characteristics of the astrochemical field, thereby laying a foundation for improving the professionalism and accuracy of the query results. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 A schematic diagram of a process for constructing a spectral database query model in one embodiment;

[0061] Figure 2 A flowchart of a query data set construction step in one embodiment;

[0062] Figure 3 A schematic diagram of a process for constructing a molecular spectroscopy database in one embodiment;

[0063] Figure 4 A schematic diagram of a process for optimizing a spectrum database query model in one embodiment;

[0064] Figure 5 A schematic diagram of a spectrum database query display interface in one embodiment;

[0065] Figure 6 A schematic diagram of a spectrum database query method in a specific embodiment;

[0066] Figure 7 A structural block diagram of a spectrum database query system in one embodiment;

[0067] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0069] In an exemplary embodiment, Figure 1 As shown, a method for constructing a spectral database query model is provided. The spectral database query model is applied to spectral database query. The method for constructing a spectral database query model includes the following steps:

[0070] Step S101, obtaining a molecular spectrum database.

[0071] Among them, the molecular spectrum database includes a molecule table, a spectrum table and an associated attachment table; among them, the molecule table records the basic information of the molecule; the basic information of the molecule includes at least the molecule number, molecule ID, name, chemical formula, molecular weight, and molecule; the spectrum table records the spectral parameter data corresponding to each molecule; the spectral parameter data includes at least frequency, intensity, temperature, pressure, Einstein coefficient, vibration energy level, line width, linearity, high energy level state, low energy level state, and partition function; the associated attachment table is used to record related references, spectral analysis, spectral graphs and other link information.

[0072] Step S102: constructing a natural language query template and an SQL query statement template corresponding to the natural language query template based on the molecular spectrum database.

[0073] Step S103: construct a query data set according to the natural language query template and the SQL query statement template.

[0074] Among them, the natural language query template is an input template corresponding to various natural language query conditions; the SQL query statement template is a template corresponding to converting the natural language query into a language that can be understood and executed by the molecular spectrum database.

[0075] Exemplarily, a natural language query template is constructed according to the structure of a molecular spectrum database and the data contained in the molecular spectrum database. For example, the natural language query template is "query spectral data of {molecule} molecules with a temperature above {temp_min}K", "find the spectrum of {molecule} molecules with a pressure below {pressure_max} atm", "display the spectrum of {molecule} molecules with a frequency between {freq_min} Hz and {freq_max} Hz", "find the spectral data of {molecule} molecules with an intensity greater than {intensity_min}", "query spectral data of {molecule} molecules with a temperature between {temp_min}K and {temp_max}K", "query spectral data of {molecule} molecules with a temperature above {temp_min}K", "query spectral data of {molecule} molecules with an Einstein value above {A}", "query spectral data of {molecule} molecules with a low energy level state greater than {g0}"; wherein molecule is a molecular variable; temp_min and temp_max are both temperature variables; pressure_max is a pressure variable; freq_min and freq_max are both frequency variables; intensity_min is an intensity variable; A is an Einstein coefficient variable; and g0 is an energy level state variable. It is to be understood that the above variables can be assigned any value.

[0076] Further, according to the above natural language template, the corresponding SQL query statement template is constructed. The SQL query statement templates are respectively "SELECT FROM Spectra WHERE MoleculeID=(SELECT MoleculeID FROM MoleculesWHERE Name = '{molecule}') AND Temperature>{temp_min};","SELECT FROM SpectraWHERE MoleculeID=(SELECT MoleculeID FROM Molecules WHERE Name = '{molecule}')AND Pressure<{pressure_max};","SELECT FROM Spectra WHERE MoleculeID=(SELECTMoleculeID FROM Molecules WHERE Name = '{molecule}')AND Frequency BETWEEN{freq_min}AND {freq_max};","SELECT FROM Spectra WHERE MoleculeID=(SELECTMoleculeID FROM Molecules WHERE Name='{molecule}')AND Intensity>{intensity_min};","SELECT FROM Spectra WHERE MoleculeID=(SELECT MoleculeID FROMMolecules WHERE Name='{molecule}') AND Temperature BETWEEN{temp_min} AND{temp_max};","SELECT FROM Spectra WHERE MoleculeID=(SELECT MoleculeID FROMMolecules WHERE Name='{molecule}') AND A>{A};","SELECT FROM Spectra WHEREMoleculeID=(SELECT MoleculeID FROM Molecules WHERE Name='{molecule}')AND G>{g0}". Among them, Spectra represents the spectrum table in the molecular spectroscopy database; Molecules represents the molecule table in the molecular spectroscopy database; MoleculeID represents the molecule ID.

[0077] Furthermore, based on the natural language query template and the SQL query statement template, by assigning values ​​to each variable, multiple natural language query conditions and corresponding multiple SQL query statements can be obtained, and then based on the multiple natural language query conditions and corresponding multiple SQL query statements, a query data set can be obtained.

[0078] Step S104, obtaining the fine-tuned vertical domain macro model.

[0079] Among them, the fine-tuned vertical field large model refers to the model that is obtained by fine-tuning the pre-trained model with astrochemical data, which can better understand and handle astrochemical tasks; among them, the pre-trained model refers to a deep learning model that has been pre-trained on a large-scale dataset. The model has been trained to learn rich feature representations and can be used as a basic model in astrochemical tasks.

[0080] Step S105, based on the query data set, the fine-tuned vertical domain large model is trained to obtain a spectral database query model.

[0081] Exemplarily, the query data set needs to be formatted first to adapt to the input requirements of the fine-tuned vertical domain large model, such as converting the format of the query data set to JSON format; further, the query data set after format conversion is divided into a training set and a test set according to a preset division ratio; then based on the training set, the fine-tuned vertical domain large model is trained; based on the test set, the performance of the trained fine-tuned vertical domain large model is evaluated until the target loss function meets the preset convergence condition, and the spectral database query model is obtained. Among them, the preset division ratio needs to be set according to the model training requirements, and no specific limitation is made here; the target loss function can be but is not limited to a cross loss function; the preset convergence condition needs to be set according to the actual training requirements, and no specific limitation is made here.

[0082] In this embodiment, based on the molecular spectral database, a natural language query template and an SQL query statement template corresponding to the natural language query template are constructed; then, a query data set is constructed according to the natural language query template and the SQL query statement template, which not only enhances the professionalism of the query data set, but also improves the quality of the query data set, laying a foundation for improving the accuracy of the spectral database query model; then, based on the query data set, the fine-tuned vertical field large model is trained to obtain a spectral database query model, which enables the model to better understand and process the language features in the field of astrochemistry, thereby laying a foundation for improving the professionalism and accuracy of the query results.

[0083] In one embodiment, obtaining a fine-tuned vertical domain macro model includes the following steps:

[0084] Step 1: Obtain astrochemical text data.

[0085] Step 2: Based on the astrochemical text data, fine-tune the vertical field big model to obtain the fine-tuned vertical field big model.

[0086] The astrochemical text data may include, but are not limited to, astrochemical textbooks, papers, articles in the fields of astronomy and chemistry, web pages, and public data.

[0087] The vertical domain big model is a pre-trained basic model. The vertical domain big model can be but is not limited to the LLaMa3 base big model, which is not specifically limited here.

[0088] Exemplarily, astrochemical text data is acquired and preprocessed to obtain preprocessed astrochemical text data; based on the preprocessed astrochemical text data, a vertical field big model is fine-tuned to obtain a fine-tuned vertical field big model.

[0089] Among them, the method for preprocessing astrochemical text data can be: performing word segmentation processing on astrochemical text data; the method for fine-tuning the vertical field large model can be: fine-tuning the vertical field large model based on SFT (Supervised Fine-Tuning) technology.

[0090] In this embodiment, by collecting astrochemical text data and then fine-tuning the vertical field big model based on the astrochemical text data, a fine-tuned vertical field big model that can better understand and process the language features of the astrochemical field can be obtained, thereby laying the foundation for building a spectral database query model.

[0091] In one embodiment, Figure 2 As shown, Figure 2 The flowchart of the query data set construction step in one embodiment is as follows; the query data set is constructed according to the natural language query template and the SQL query statement template, including the following steps:

[0092] Step S201, constructing a random parameter generation function based on spectral parameter data.

[0093] Among them, the molecular spectrum database includes spectrum parameter data; wherein, the spectrum parameter data at least includes frequency, intensity, temperature, pressure, Einstein coefficient, vibration energy level, line width, linearity, high energy level state, low energy level state, and partition function.

[0094] The random parameter generation function is used to generate a random variable that meets the preset parameter range; the preset parameter range needs to be set according to actual needs and is not specifically limited here. For example, assuming that the preset parameter range is 0 to 100, the random parameter generation function is used to generate a random variable within the parameter range of 0 to 100.

[0095] Step S202, generating a random variable according to a random parameter generation function.

[0096] Step S203: Substitute the random variables into the natural language query template and the SQL query statement template respectively to generate multiple natural language query conditions and corresponding multiple SQL query statements.

[0097] Step S205: construct a query data set according to multiple natural language query conditions and corresponding multiple SQL query statements.

[0098] Exemplarily, when the spectral parameter data are temperature, pressure, frequency, intensity, and Einstein coefficient, a random parameter generation function is constructed to randomly generate a minimum temperature value temp_min in the range of 0 to 100, a maximum temperature value temp_max in the range of 100 to 10000, a maximum pressure value pressure_max in the range of 0 to 1000, a minimum frequency value freq_min in the range of 0 to 5000, a maximum frequency value freq_max in the range of 5001 to 30000, a minimum intensity value intensity_min in the range of 10 to 100, an Einstein coefficient A in the range of 0 to 1, an Einstein coefficient A in the range of 0 to 5, and randomly select a molecule molecule in a molecule table of a molecular spectral database; the random parameter generation function is defined as generate_random_params (); then the random parameter generation function is {"temp_min": random.randint(0, 100), "temp_max": random.randint(100, 10000), "pressure_max": random.uniform(0, 1000), "freq_min": random.randint(0, 5000), "freq_max": random.randint(5001, 30000), "intensity_min": random.randint(10, 100), "molecule": random.choice(molecules), "A": random.randint(0, 1), "A": random.randint(0, 5)}.

[0099] Furthermore, based on the random parameter generation function generate_random_params(), random variables corresponding to the spectral parameter data are generated; the random variables corresponding to each spectral parameter data are respectively substituted into the natural language query template and the SQL query statement template to generate multiple natural language query conditions and corresponding multiple SQL query statements; and a query data set is constructed according to the obtained multiple natural language query conditions and the corresponding multiple SQL query statements.

[0100] It should be noted that both natural language query conditions and SQL query statements are set based on the query requirements of real molecular spectral data. Therefore, they are highly professional and relevant, and can adapt to various types of query requirements, laying the foundation for improving the professionalism and accuracy of the spectral database query model.

[0101] In this embodiment, a random parameter generation function corresponding to the spectral parameter data can be constructed based on the spectral parameter data, and then based on the random parameter generation function, random variables corresponding to the spectral parameter data can be generated, which lays a foundation for obtaining multiple natural language query conditions and corresponding multiple SQL query statements, and then based on the multiple natural language query conditions and the corresponding multiple SQL query statements, a query data set is constructed, which can improve the professionalism and quality of the query data set while ensuring the data volume of the query data set.

[0102] In one embodiment, Figure 3 As shown, Figure 3 The figure is a flow chart of the steps of constructing a molecular spectroscopy database in one embodiment; the step of obtaining a molecular spectroscopy database includes:

[0103] Step S301: acquiring original spectral data from a plurality of different spectral databases.

[0104] The spectral database is an existing spectral database, including the CDMS database and the HITRAN database. The original spectral data covers the spectral parameter data corresponding to all existing molecules.

[0105] It should be noted that the existing spectral database has some shortcomings due to its long construction time. For example, the spectral coverage frequency ranges in different spectral databases are different, resulting in inconsistent data; the spectral data accumulation lacks classification information, and some data and quantum number information are missing, resulting in incomplete data; there is a lack of unified interpretation and description function for molecular spectra and parameter information, resulting in inaccurate data, etc. The above-mentioned deficiencies of the spectral database will lead to low accuracy of database query results; therefore, in order to improve the accuracy and comprehensiveness of spectral database query results, it is necessary to build a complete and comprehensive molecular spectral database.

[0106] Step S302, performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data.

[0107] It is understandable that since the original spectral data comes from multiple different spectral databases, there are problems such as data duplication and missing data. Therefore, it is necessary to perform deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain more comprehensive and accurate molecular spectral data.

[0108] Step S303: construct a molecular spectrum database according to the molecular spectrum data.

[0109] Preferably, step S302, performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data, includes the following steps:

[0110] Step 1: perform data deduplication processing on duplicate data corresponding to the same molecule in the original spectral data to obtain deduplication spectral data.

[0111] Among them, repeated data corresponding to the same molecule, for example, repeated data of the same molecule in different bands, repeated data of the same molecule measured by different devices.

[0112] Exemplarily, deduplication processing of repeated data corresponding to the same molecule in the original spectral data may be: based on the frequency information and intensity information of the molecule, deduplication processing is performed on overlapping data of the same frequency and the same intensity of the same molecule in the original spectral data.

[0113] Step 2: Determine parameter column information based on multiple different spectral databases.

[0114] Among them, the parameter column information includes all parameter type information corresponding to all molecules, such as frequency, intensity, temperature, pressure, Einstein coefficient, vibration energy level, line width, linearity, high energy level state, low energy level state, partition function, etc.

[0115] Step 3: Determine the missing parameter columns and the shared parameter columns based on the deduplicated spectral data and parameter column information.

[0116] The data missing parameter column refers to the parameter column corresponding to the missing related parameters in the spectral parameter data; the shared parameter column refers to the parameter column corresponding to the spectral parameter data shared by multiple different spectral databases.

[0117] It should be noted that due to the large differences in parameter quantities in different spectral databases, there are certain differences in the parameter columns corresponding to each molecule in the original spectral data. In order to ensure the comprehensiveness and integrity of the molecular spectral data, it is necessary to fill in the missing parameter columns while retaining the common parameter columns to achieve reconstruction of the missing parameter columns.

[0118] Step 4: Based on the common parameter columns, fill in the missing parameter columns to obtain reconstructed spectral data.

[0119] Optionally, based on the common parameter column, the method of performing data filling processing on the data missing parameter column to obtain the reconstructed spectral data may be: based on the common parameter column, using the quantitative calculation method, performing data filling processing on the data missing parameter column to obtain the reconstructed spectral data. Exemplarily, this embodiment takes the data missing parameter column as the Einstein coefficient parameter column as an example for explanation, based on the calculation formula corresponding to the Einstein coefficient A, the specific parameters involved in the calculation of the Einstein coefficient A can be determined, and these parameters can all be determined based on the common parameter column, and then based on the calculation formula corresponding to the common parameter column and the Einstein coefficient A, the corresponding Einstein coefficient A can be calculated, and the Einstein coefficient parameter column is subjected to data filling processing to obtain the reconstructed Einstein coefficient parameter column.

[0120] Step 5: Integrate the deduplicated spectral data with the reconstructed spectral data to obtain molecular spectral data.

[0121] In this embodiment, based on multiple existing different spectral databases, a large amount of raw spectral data is obtained, and the raw spectral data is preprocessed for deduplication and reconstruction to obtain deduplicated spectral data and reconstructed spectral data, and then the deduplicated spectral data and the reconstructed spectral data are integrated to obtain molecular spectral data, which can construct a more comprehensive and accurate molecular spectral database, laying a data foundation for improving the accuracy of spectral data query results.

[0122] In one embodiment, based on the query data set, the fine-tuned vertical domain large model is trained to obtain a spectral database query model, including the following steps:

[0123] Step 1: preprocess the natural language query conditions in the query data set to obtain corresponding multiple index sequences.

[0124] Among them, the preprocessing of natural language query conditions in the query data set may include but is not limited to: space and punctuation removal, simplified and traditional Chinese conversion, word segmentation, text standardization, and text entity recognition; for example, the preprocessing of natural language query conditions is implemented based on the QWEN2 function to obtain the corresponding index sequence.

[0125] In an exemplary embodiment, before processing the natural language query condition, it is necessary to first perform space and punctuation removal processing and simplified-traditional conversion processing on the natural language query condition; wherein, space and punctuation removal processing refers to removing redundant spaces and punctuation marks in the natural language query condition, or retaining key punctuation marks as needed; simplified-traditional conversion processing refers to the need to unify the language format of the natural language query condition into simplified Chinese for the Chinese natural language query condition, that is, when there are traditional Chinese characters, the traditional Chinese characters are converted into simplified Chinese. Further, based on a predefined Chinese dictionary, the word boundaries in the natural language query condition are identified, and based on the word boundaries, the natural language query condition is segmented to obtain a corresponding plurality of segmented words; part-of-speech tagging is performed on the plurality of segmented words, that is, a corresponding part-of-speech tag is assigned to each segmented word, such as a noun, a verb, etc., so as to provide richer semantic information for downstream tasks; further, based on a preset mapping vocabulary table, the index corresponding to each segmented word is determined, and the corresponding index sequence is obtained according to the index corresponding to each segmented word. It should be noted that the preset Chinese dictionary and the preset mapping vocabulary table need to be set according to actual needs, and no specific limitation is made here.

[0126] Step 2: According to multiple index sequences, the fine-tuned vertical domain large model is trained until the target loss function meets the preset convergence condition to obtain the spectral database query model.

[0127] Preferably, the target loss function is a cross loss function, and the target loss function is shown in formula (1).

[0128] (1)

[0129] in, , represents the actual value; , represents the predicted value, n, represents the number of samples.

[0130] Exemplarily, multiple index sequences are input into the fine-tuned vertical domain large model, and the fine-tuned vertical domain large model is trained until the target loss function meets the preset convergence condition, thereby obtaining a spectral database query model. The preset convergence condition needs to be set according to actual training requirements and is not specifically limited here.

[0131] In this embodiment, the natural language query conditions in the query data set are preprocessed to obtain multiple corresponding index sequences; then, based on the multiple index sequences, the fine-tuned vertical field large model is trained until the target loss function meets the preset convergence condition, and the spectral database query model is obtained. Based on this, the professionalism and accuracy of the spectral database query model are improved.

[0132] In one embodiment, Figure 4 As shown, Figure 4The figure is a flow chart of the optimization step of the spectral database query model in one embodiment; the spectral database query model construction method also includes the following steps:

[0133] Step S401, obtaining multiple test natural language query conditions and multiple correct query results corresponding to the multiple test natural language query conditions.

[0134] The test natural language query condition may be, but is not limited to, a preset spectral data query condition; and the normal query result refers to a correct database query result corresponding to the test natural language query condition.

[0135] Step S402: Generate a plurality of corresponding test SQL query statements according to a plurality of test natural language query conditions through a spectrum database query model.

[0136] Step S403, based on the molecular spectrum database and multiple test SQL query statements, obtain corresponding multiple test query results.

[0137] Exemplarily, the test natural language query conditions are preprocessed to obtain a corresponding index sequence, which is then used as the input of a spectral database query model. Based on the spectral database query model, a corresponding test SQL query statement is output, and then the molecular spectral database is accessed and the test SQL query statement is executed to obtain the corresponding test query result.

[0138] Step S404: determining the exact matching degree according to the multiple correct query results and the multiple test query results.

[0139] Exemplarily, the exact matching degree is calculated according to formula (2) based on multiple correct query results and multiple test query results.

[0140] (2)

[0141] Among them, EM represents the exact match; N represents the logarithm of the query results; , represents the i-th test query result; , represents the i-th correct query result corresponding to the i-th test query result; , represents the indicator function, when If true, the value is 1, otherwise it is 0.

[0142] Step S405: If the exact match degree does not reach the preset accuracy threshold, the spectrum database query model is optimized until the exact match degree reaches the preset accuracy threshold.

[0143] Among them, the preset accuracy threshold needs to be set according to actual needs and is not specifically limited here.

[0144] Exemplarily, according to multiple correct query results and multiple test query results, the exact match degree is calculated according to formula (2), and then it is determined whether the exact match degree reaches the preset accuracy threshold. If the exact match degree does not reach the preset accuracy threshold, the spectral database query model is optimized until the exact match degree reaches the preset accuracy threshold; if the exact match degree reaches the preset accuracy threshold, it indicates that the performance of the spectral database query model has reached the optimal level. It should be noted that the optimization of the spectral database query model can be, but is not limited to: optimizing the query data set and optimizing the parameters in the spectral database query model.

[0145] In this embodiment, based on the correct query results and test query results corresponding to multiple test natural language query conditions, the exact match degree can be accurately determined; based on the preset accuracy threshold, when the exact match degree does not reach the preset accuracy threshold, the spectral database query model is optimized to ensure the accuracy of the spectral database query model.

[0146] In an exemplary embodiment, a spectrum database query method is provided, comprising the following steps:

[0147] Step 1: Input the target natural language input by the user into the spectral database query model to obtain the target query statement corresponding to the target natural language.

[0148] Step 2: Based on the molecular spectroscopy database, generate corresponding target query results by executing the target query statement.

[0149] The target natural language is the query condition related to the spectral data used for input; the target query statement is the SQL query statement corresponding to the target natural language output by the spectral database query model.

[0150] Exemplarily, a target natural language input by a user is input into a spectral database query model to obtain a target query statement corresponding to the target natural language, including: obtaining the target natural language input by the user; preprocessing the target natural language input by the user to obtain a target index sequence; inputting the target index sequence into the spectral database query model to obtain a target query statement corresponding to the target natural language.

[0151] Among them, the method of preprocessing the target natural language input by the user to obtain the target index sequence can be: performing word segmentation processing, text normalization processing and text entity recognition on the target natural language input by the user to obtain the target index sequence. Among them, word segmentation processing refers to dividing the target natural language into multiple word segments; according to the multiple word segments, a corresponding word segmentation list is obtained. Text normalization processing refers to lowercase conversion, unit normalization and part-of-speech restoration of the word segmentation list to obtain a normalized word segmentation list; among them, lowercase conversion refers to converting all English texts to lowercase; unit normalization refers to unifying units; part-of-speech restoration refers to restoring verbs or nouns. Text entity recognition refers to identifying entities with specific meanings in the normalized word segmentation list to obtain corresponding target entities and target entity categories; entity categories can include but are not limited to chemical symbols, spectral parameter data, comparison operations, etc.

[0152] In an exemplary embodiment, assuming that the target natural language input by the user is "query the data results of CO temperature greater than 300K", the target natural language "query the data results of CO temperature greater than 300K" is segmented to obtain multiple corresponding segmented words; then, according to the multiple segmented words, a corresponding segmented word list is obtained as ["query", "CO", "greater than", "300K", "temperature", "of", "data", "result"]; further, the segmented word list is subjected to text normalization processing, i.e., lowercase conversion, unit normalization and part of speech restoration, and the normalized segmented word list obtained is ["query", "CO", "greater than", "300K", "temperature", "data", "result"]; further, text entity recognition is performed on the normalized segmented word list, and the corresponding target entity and target entity category are obtained as [("CO", "C HEMICAL_SYMBOL", ("300K", "TEMPERATURE"), ("greater than", "COMPARISON_OPERATOR"), ("temperature", "PROPERTY")], where "CO", "300K", "greater than", and "temperature" are all target entities; "CHEMICAL_SYMBOL", "TEMPERATURE", "COMPARISON_OPERATOR" and "PROPERTY" are all target entity categories corresponding to the target entity; where "CHEMICAL_SYMBOL" represents chemical symbol; "TEMPERATURE" represents temperature; "COMPARISON_OPERATOR" represents comparison operation; and "PROPERTY" represents property. Further, based on the preset mapping vocabulary, the index corresponding to each target entity is determined, and the corresponding target index sequence is obtained according to the index corresponding to each target entity. The target index sequence is input into the spectral database query model to obtain the target query statement corresponding to the target natural language, and the target query language is SELECT FROM Spectra WHERE chemical_symbol = 'CO' AND temperature>300. Further, based on the molecular spectrum database, the corresponding target query results are generated by executing the target query statement.

[0153] In this implementation, based on the spectral database query model, an accurate target query statement can be obtained according to the target natural language input by the user, and then based on the molecular spectral database, the corresponding target query result is generated by executing the target query statement. Based on this, the user's query intention can be accurately identified, thereby improving the accuracy of the spectral database query results.

[0154] In one embodiment, the spectrum database query method further includes:

[0155] A molecular spectrum is generated according to the target query results and displayed on the spectrum database query display interface.

[0156] The spectrum database query display interface is used to display information related to the spectrum database query.

[0157] Exemplarily, the method of generating a molecular spectrum map according to the target query result may be, but is not limited to: determining the frequency and intensity corresponding to the molecule according to the target query result; and generating the molecular spectrum map based on the frequency and intensity corresponding to the molecule.

[0158] In this embodiment, based on the target query result, a corresponding molecular spectrum graph is further generated and displayed, which can improve the user experience and user-friendliness.

[0159] In one embodiment, the spectrum database query method further includes:

[0160] Based on the spectral database query model and the target query result, the molecular interpretation information corresponding to the target query result is generated and displayed on the spectral database query display interface.

[0161] Among them, the molecular interpretation information is a detailed description of the molecules queried by the user, including molecular results, characteristics and common uses.

[0162] In this embodiment, based on the spectral database query model and the target query result, molecular interpretation information corresponding to the target query result is generated and displayed on the spectral database query display interface, which can effectively enhance the user's understanding of the target query result and improve the readability and professionalism of the target query result.

[0163] In one embodiment, Figure 5 As shown, a spectrum database query display interface is provided, and the spectrum database query display interface is provided with a user input interface, a query result display interface, a spectrum graph display interface and a molecular interpretation information display interface;

[0164] The user input interface is used to provide an input port for the target natural language;

[0165] The query result display interface is used to display the target query result; the target query result is the target query statement corresponding to the target natural language obtained by inputting the target natural language input by the user into the spectrum database query model; the target query result is generated by executing the target query statement based on the molecular spectrum database;

[0166] The spectrum display interface is used to display the molecular spectrum; the molecular spectrum is a molecular spectrum generated according to the target query result;

[0167] The molecular interpretation information display interface is used to display molecular interpretation information; the molecular interpretation information is molecular interpretation information corresponding to the target query result, generated based on the spectral database query model and the target query result.

[0168] This implementation realizes the visualization of spectral data query results through the spectral database query display interface, improves the user experience, and effectively enhances the user's understanding of the target query results through the spectral graph display interface and the molecular interpretation information display interface.

[0169] In a specific embodiment, see Figure 6 , the spectrum database query method includes the following steps:

[0170] Step S601: obtaining a target natural language input by a user in a user input interface.

[0171] Step S602: pre-process the target natural language to obtain a target index sequence.

[0172] Step S603: input the target index sequence into the spectral database query model to obtain the corresponding target query result.

[0173] Step S604: Based on the molecular spectrum database, by executing the target query statement, a corresponding target query result is generated and displayed on the query result display interface.

[0174] Step S605, generating a molecular spectrum graph according to the target query result, and displaying it on the spectrum graph display interface.

[0175] Step S506: Based on the spectral database query model and the target query result, generate molecular interpretation information corresponding to the target query result, and display it on the molecular interpretation information display interface.

[0176] The above-mentioned spectral database query model construction method, based on the molecular spectral database, constructs a natural language query template and an SQL query statement template corresponding to the natural language query template; then, based on the natural language query template and the SQL query statement template, a query data set is constructed, which not only enhances the professionalism of the query data set, but also improves the quality of the query data set, laying a foundation for improving the accuracy of the spectral database query model; then, based on the query data set, the fine-tuned vertical field large model is trained to obtain a spectral database query model, which enables the model to better understand and process the language features of the astrochemical field, thereby laying a foundation for improving the professionalism and accuracy of the query results. The above-mentioned spectral database query method, based on the spectral database query model, can obtain an accurate target query statement according to the target natural language input by the user, and then based on the molecular spectral database, by executing the target query statement, the corresponding target query result is generated, which can accurately identify the user's query intention and improve the accuracy of the spectral database query result.

[0177] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0178] Based on the same inventive concept, the embodiment of the present application also provides a spectral database query system for implementing the spectral database query method involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more spectral database query system embodiments provided below can refer to the limitations of the spectral database query method above, and will not be repeated here.

[0179] In an exemplary embodiment, Figure 7 As shown, a spectrum database query system is provided, the system includes: a user input module 701, a query result generation module 702, a spectrum graph generation module 703 and a molecule interpretation information generation module 704;

[0180] The user input module 701 is used to receive a target natural language input by a user;

[0181] The query result generation module 702 is used to input the target natural language input by the user into the spectrum database query model to obtain a target query statement corresponding to the target natural language; based on the molecular spectrum database, by executing the target query statement, generate a corresponding target query result;

[0182] The spectrum graph generation module 703 is used to generate a molecular spectrum graph according to the target query result and display it on the spectrum database query display interface;

[0183] The molecular interpretation information generating module 704 is used to generate molecular interpretation information corresponding to the target query result based on the spectrum database query model and the target query result, and display it on the spectrum database query display interface.

[0184] The above-mentioned spectral database query system, based on the spectral database query model, can obtain accurate target query statements according to the target natural language input by the user, and then generate corresponding target query results based on the molecular spectral database by executing the target query statements, thereby improving the accuracy of the spectral database query results; further, according to the target query results, the corresponding molecular spectrum diagram and molecular interpretation information are obtained, based on this, the accuracy and flexibility of natural language queries are improved, and it can more accurately identify user query intentions, extract complete information and explanations.

[0185] Each module in the above spectrum database query system can be implemented in whole or in part by software, hardware and their combination. Each module can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each module.

[0186] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store spectral database query related data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a spectral database query model construction method and a spectral database query method are implemented.

[0187] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0188] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0189] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0190] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0192] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0193] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0194] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing a spectral database query model, characterized in that: The spectrum database query model is applied to spectrum database query, and the method includes: Access to molecular spectroscopy databases; Based on the molecular spectroscopy database, construct a natural language query template and an SQL query statement template corresponding to the natural language query template; Constructing a query data set according to the natural language query template and the SQL query statement template; Obtain a fine-tuned vertical domain model; Based on the query data set, the fine-tuned vertical domain large model is trained to obtain a spectral database query model; The step of acquiring a molecular spectroscopy database comprises: Obtaining raw spectral data from multiple different spectral databases; Performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data; Constructing a molecular spectrum database according to the molecular spectrum data; The performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data includes: Performing data deduplication processing on repeated data corresponding to the same molecule in the original spectral data to obtain deduplication spectral data; Determine parameter column information based on multiple different spectral databases; Determine data missing parameter columns and shared parameter columns according to the deduplicated spectral data and the parameter column information; Based on the common parameter column, performing data filling processing on the data missing parameter column to obtain reconstructed spectral data; The deduplicated spectral data and the reconstructed spectral data are integrated to obtain molecular spectral data.

2. The method according to claim 1, characterized in that The step of obtaining a fine-tuned vertical domain macro model includes: Get astrochemical text data; Based on the astrochemical text data, the vertical field big model is fine-tuned to obtain a fine-tuned vertical field big model.

3. The method according to claim 1, characterized in that The molecular spectroscopy database includes spectral parameter data; the query data set is constructed according to the natural language query template and the SQL query statement template, including: Based on the spectral parameter data, construct a random parameter generation function; Generate a random variable according to the random parameter generation function; Substituting the random variables into the natural language query template and the SQL query statement template respectively to generate multiple natural language query conditions and corresponding multiple SQL query statements; A query data set is constructed according to the multiple natural language query conditions and the corresponding multiple SQL query statements.

4. The method according to claim 1, characterized in that: The method of training the fine-tuned vertical domain large model based on the query data set to obtain a spectral database query model includes: Preprocessing the natural language query conditions in the query data set to obtain corresponding multiple index sequences; According to the plurality of index sequences, the fine-tuned vertical domain large model is trained until the target loss function meets the preset convergence condition, thereby obtaining a spectral database query model.

5. The method according to claim 1, characterized in that The method further comprises: Obtaining multiple test natural language query conditions and multiple correct query results corresponding to the multiple test natural language query conditions; Generate a plurality of corresponding test SQL query statements according to a plurality of test natural language query conditions through the spectrum database query model; Based on the molecular spectroscopy database and the plurality of test SQL query statements, obtaining a plurality of corresponding test query results; Determining an exact match degree according to the plurality of correct query results and the plurality of test query results; If the degree of exact match does not reach the preset accuracy threshold, the spectrum database query model is optimized until the degree of exact match reaches the preset accuracy threshold.

6. A spectrum database query method, characterized in that: The method comprises: Inputting the target natural language input by the user into the spectral database query model to obtain a target query statement corresponding to the target natural language; Based on the molecular spectrum database, generating corresponding target query results by executing the target query statement; The method further comprises: Generate a molecular spectrum diagram according to the target query result, and display it on the spectrum database query display interface; The method further comprises: Based on the spectral database query model and the target query result, generate molecular interpretation information corresponding to the target query result, and display it on the spectral database query display interface; The process of constructing the molecular spectroscopy database includes: Obtaining raw spectral data from multiple different spectral databases; Performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data; Constructing a molecular spectrum database according to the molecular spectrum data; The performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data includes: Performing data deduplication processing on repeated data corresponding to the same molecule in the original spectral data to obtain deduplication spectral data; Determine parameter column information based on multiple different spectral databases; Determine data missing parameter columns and shared parameter columns according to the deduplicated spectral data and the parameter column information; Based on the common parameter column, performing data filling processing on the data missing parameter column to obtain reconstructed spectral data; The deduplicated spectral data and the reconstructed spectral data are integrated to obtain molecular spectral data.

7. A spectrum database query display interface, characterized in that: The spectrum database query display interface is provided with a user input interface, a query result display interface, a spectrum graph display interface and a molecular interpretation information display interface; The user input interface is used to provide an input port for the target natural language; The query result display interface is used to display the target query result; The target query result is a target query statement corresponding to the target natural language obtained by inputting the target natural language input by the user into the spectrum database query model; Based on the molecular spectrum database, a target query result is generated by executing the target query statement; The spectrum display interface is used to display a molecular spectrum; the molecular spectrum is a molecular spectrum generated according to the target query result; The molecular interpretation information display interface is used to display molecular interpretation information; The molecular interpretation information is molecular interpretation information corresponding to the target query result, generated based on the spectral database query model and the target query result; The process of constructing the molecular spectroscopy database includes: Obtaining raw spectral data from multiple different spectral databases; Performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data; Constructing a molecular spectrum database according to the molecular spectrum data; The performing deduplication preprocessing and reconstruction preprocessing on the original spectral data to obtain molecular spectral data includes: Performing data deduplication processing on repeated data corresponding to the same molecule in the original spectral data to obtain deduplication spectral data; Determine parameter column information based on multiple different spectral databases; Determine data missing parameter columns and shared parameter columns according to the deduplicated spectral data and the parameter column information; Based on the common parameter column, performing data filling processing on the data missing parameter column to obtain reconstructed spectral data; The deduplicated spectral data and the reconstructed spectral data are integrated to obtain molecular spectral data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

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

Images