CN110362798B - Method, apparatus, computer device and storage medium for judging information retrieval analysis - Google Patents
Method, apparatus, computer device and storage medium for judging information retrieval analysis Download PDFInfo
- Publication number
- CN110362798B CN110362798B CN201910520201.7A CN201910520201A CN110362798B CN 110362798 B CN110362798 B CN 110362798B CN 201910520201 A CN201910520201 A CN 201910520201A CN 110362798 B CN110362798 B CN 110362798B
- Authority
- CN
- China
- Prior art keywords
- analysis
- case
- target
- vector
- statement
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 274
- 238000000034 method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 177
- 230000014509 gene expression Effects 0.000 claims abstract description 75
- 238000013145 classification model Methods 0.000 claims abstract description 20
- 230000015654 memory Effects 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 12
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 6
- 230000005484 gravity Effects 0.000 description 6
- 238000007619 statistical method Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 3
- 238000012790 confirmation Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000001427 coherent effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/186—Templates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/18—Legal services; Handling legal documents
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application relates to a machine learning-based arbitration information retrieval analysis method, a machine learning-based arbitration information retrieval analysis device, a machine learning-based arbitration information retrieval analysis computer device and a storage medium. The method comprises the following steps: receiving a judging and analyzing request sent by a terminal; the arbitration analysis request carries a retrieval analysis statement; acquiring a case statistical table and corresponding table information; generating a target vector according to the search analysis statement and the table information; inputting the target vector into a preset sequence model to obtain an analysis intention expression; inputting the target vector into a preset intention classification model to obtain a target SQL template; filling the analysis intention expression into the target SQL template to obtain an SQL query statement; inquiring related cases in the case statistics table based on the SQL inquiry statement, carrying out statistics analysis on case information of the related cases, and returning an analysis result to the terminal. The method can improve the efficiency of the retrieval and analysis of the judging information.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a method and apparatus for determining information retrieval and analysis, a computer device, and a storage medium.
Background
Currently, in the case approval process, in order to more accurately and efficiently judge the current case, judges and judges are expected to search for case processing conditions of the related cases in the past, such as the judgment opinion tendency of a court. However, the conventional approach only supports the user to index the documents based on the case and the criminal name specified in the national legal system, and generally feeds back the arbitrated document for each relevant case to the user. The method requires the user to manually check the judging documents one by one, and is difficult to quickly know the case processing conditions of the related cases, so that the efficiency of the retrieval and analysis of the judging information is reduced.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a method, an apparatus, a computer device, and a storage medium for searching and analyzing information, which can automatically count the case processing conditions of related cases in the past by a court and further improve the efficiency of searching and analyzing the information.
A method of arbitrating information retrieval analysis, the method comprising: receiving a judging and analyzing request sent by a terminal; the arbitration analysis request carries a retrieval analysis statement; acquiring a case statistical table and corresponding table information; generating a target vector according to the search analysis statement and the table information; inputting the target vector into a preset sequence model to obtain an analysis intention expression; inputting the target vector into a preset intention classification model to obtain a target SQL template; filling the analysis intention expression into the target SQL template to obtain an SQL query statement; inquiring related cases in the case statistics table based on the SQL inquiry statement, carrying out statistics analysis on case information of the related cases, and returning an analysis result to the terminal.
In one embodiment, before the obtaining the case statistics table and the corresponding table information, the method further includes: acquiring case files of a plurality of historical cases; extracting case identifications and one or more factor description sentences of corresponding historical cases from the case file through regular matching; inputting the factor description statement into a preset semantic understanding model to obtain a plurality of case factors; and constructing a case statistical table based on the case identifications and the case factors corresponding to the case identifications.
In one embodiment, the table information includes a plurality of field enumeration values; the generating a target vector according to the search analysis statement and the table information comprises the following steps: performing word segmentation on the search analysis statement, calculating word vectors of each word segmentation, and recording the word vectors as first vectors; calculating word vectors corresponding to each field enumeration value, and recording the word vectors as second vectors; calculating the similarity of the first vector and different second vectors; and splicing the first vector with the second vector with the highest similarity to obtain the target vector.
In one embodiment, the sequence model includes a dimensional sequence model and a conditional sequence model; the analysis intent expression includes an analysis dimension expression; inputting the target vector into a preset sequence model to obtain an analysis intention expression, wherein the analysis intention expression comprises the following steps: calling the dimension sequence model to forget a local vector containing analysis condition information in the target vector, so as to obtain one or more analysis dimension expressions; and calling the conditional sequence model to forget the local vector containing the analysis dimension information in the target vector, so as to obtain one or more analysis dimension expressions.
In one embodiment, the dimension sequence model includes an encoder, a decoder, and an attention module; the step of calling the dimension sequence model to perform forgetting processing on a local vector containing analysis condition information in the target vector to obtain one or more analysis dimension expressions, wherein the step of calling the dimension sequence model comprises the following steps: calling the encoder to forget a local vector containing analysis condition information in the target vector to obtain a compressed vector; invoking the decoder to decode the compressed vector pair to obtain an initial matching probability corresponding to each field enumeration value; invoking the attention module to perform attention training on the compressed vector to obtain similarity weighting corresponding to each field enumeration value; adjusting the initial matching probability of the corresponding field enumeration value according to the similarity weighting to obtain a target matching probability corresponding to each field enumeration value; and generating an analysis dimension expression according to the field enumeration value with the highest target matching probability.
In one embodiment, the returning the analysis result to the terminal includes: acquiring a chart template associated with the target SQL template; determining a plurality of basic coordinates and coordinate elements according to the coordinate extraction rules recorded by the chart template; extracting coordinate values corresponding to each coordinate element from the analysis result; constructing a target chart based on a plurality of basic coordinates and coordinate elements and coordinate values corresponding to each basic coordinate, and generating a judging result analysis page according to the target chart; determining a corresponding alternative chart type according to the basic coordinates and the coordinate elements; and adding an alternative option corresponding to each alternative chart type in the arbitration result analysis page, and returning the arbitration result analysis page added with the alternative option to the terminal.
An apparatus for arbitrating information retrieval analysis, the apparatus comprising: the analysis intention recognition module is used for receiving an arbitration analysis request sent by the terminal; the arbitration analysis request carries a retrieval analysis statement; acquiring a case statistical table and corresponding table information; generating a target vector according to the search analysis statement and the table information; inputting the target vector into a preset sequence model to obtain an analysis intention expression; the retrieval analysis statement generation module is used for inputting the target vector into a preset intention classification model to obtain a target SQL template; filling the analysis intention expression into the target SQL template to obtain an SQL query statement; and the query statistics analysis module is used for querying relevant cases in the case statistics table based on the SQL query statement, carrying out statistics analysis on the case information of the relevant cases, and returning analysis results to the terminal.
In one embodiment, the device further comprises a case statistics table construction module, configured to obtain case files of a plurality of historical cases; extracting case identifications and one or more factor description sentences of corresponding historical cases from the case file through regular matching; inputting the factor description statement into a preset semantic understanding model to obtain a plurality of case factors; and constructing a case statistical table based on the case identifications and the case factors corresponding to the case identifications.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the arbitration information retrieval analysis method provided in any one of the embodiments of the present application when the computer program is executed.
A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps of the arbitration information retrieval analysis method provided in any of the embodiments of the present application.
The method, the device, the computer equipment and the storage medium for judging information retrieval and analysis can accurately identify the retrieval and analysis intention expressed by a user based on retrieval and analysis sentences based on the sequence model and the intention classification model; by combining a case statistical table which is pre-deconstructed, the retrieval and analysis efficiency of the judging information can be improved, and different retrieval and analysis intentions of a user can be responded quickly.
Drawings
FIG. 1 is an application scenario diagram of a method of arbitration information retrieval analysis in one embodiment;
FIG. 2 is a flow diagram of a method of arbitration information retrieval analysis in one embodiment;
FIG. 3 is a flow diagram of the steps for analyzing intent expression determination in one embodiment;
FIG. 4 is a block diagram of an apparatus for arbitrating information retrieval analysis in one embodiment;
fig. 5 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
The method for analyzing the judging information retrieval can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smartphones, tablet computers, and portable wearable devices, and the server 104 may be implemented by a stand-alone server or a server cluster composed of a plurality of servers. When the user needs to perform a search analysis based on the current case, a resolution request may be sent to the server 104 through the terminal 102. The resolution request carries a retrieval resolution statement. The server 104 stores a pre-built case statistics table. The case statistics table records field enumeration values of a plurality of historical cases, such as case factors of a case, a region and the like. The server 104 acquires table information of the case statistics table, and generates a target vector for characterizing the user's search intention according to the table information and the search analysis statement. The server 104 trains the sequence model and the intention classification model in advance based on case information of the history cases. The server inputs the target vector into the sequence model, outputs a plurality of analysis intention expressions, inputs the target vector into the intention classification model, and outputs a target SQL template. The server 104 sequentially fills a plurality of analysis intention expressions into a target SQL template according to the appearance sequence of the analysis intention expressions in the corresponding search analysis statement, so as to obtain an SQL query statement. The server 104 queries related cases in the case statistics table based on the SQL query statement, performs statistical analysis on case information of the related cases, and returns an analysis result to the terminal 102, so that a user makes a decision on the current case after knowing that the decision of the related cases is inclined according to the analysis result. The judging information retrieval and analysis process can accurately identify the retrieval and analysis intention of the user based on the retrieval and analysis statement expression based on the sequence model and the intention classification model; and combining a case statistical table which is pre-solved, so that the retrieval and analysis efficiency of the judging information can be improved.
In one embodiment, as shown in fig. 2, a method for determining information retrieval and analysis is provided, and the method is applied to the server in fig. 1 for illustration, and includes the following steps:
step 202, receiving a arbitration analysis request sent by a terminal; the resolution request carries a retrieval resolution statement.
The search analysis statement may be one or more phrases formed in natural language. For example, "support rate of financial borrowing cases in Guangdong area", "specific gravity of dispute cases in 2018 Guangdong area contract release", "where cases related to lending disputes are generally distributed", and the like. The search analysis statement may be a statement that has a grammatical error and is semantically non-coherent. For example, "case interpretation trend of overdue repayment of loan in the last five years", such as Guangdong court ", and the like. For the search analysis statement with grammar errors and incoherent semantics, the server performs semantic analysis on the search analysis statement to generate one or more corresponding search intention statements with coherent semantics, generates a search intention confirmation prompt based on the search intention statement, and returns the search intention confirmation prompt to the terminal. The user can select one of the search intention sentences based on the search intention confirmation prompt, and the terminal transmits the selected information to the server. The server performs search analysis based on the search intention sentence selected by the user according to the method provided by the embodiment. In another embodiment, the search analysis statement may be a plurality of search fields, which is not limited thereto.
Step 204, obtaining a case statistical table and corresponding table information.
The case statistics table records case information of a plurality of history cases. The case information may include a case identification and a plurality of case factors deconstructed from the case file. The case file may be a litigation request book, a resolution document, etc. for the historical case. The case factors may be the result of arbitration, annual rate, subject, territory, referee time, court level or case law, etc. for the historical case. The case statistics table may be as shown in table 1 below.
TABLE 1
The server obtains the table information of the case statistics table. The table information comprises a table name, a plurality of table heads and a plurality of field enumeration values corresponding to each table head. Each header may be a case factor. For example, in table 1, each field of the first row is a header, and the fields of the other rows in each column are field enumeration values corresponding to the corresponding header, such as "Guangdong Shenzhen" and "Shanghai" are field enumeration values of the header "region" respectively.
Step 206, generating a target vector according to the search analysis statement and the table information.
The server calculates the characterization vectors corresponding to the retrieval analysis statement and the table information respectively, and splices the characterization vectors corresponding to the retrieval analysis statement and the characterization vectors corresponding to the table information to obtain the target vector.
And step 208, inputting the target vector into a preset sequence model to obtain an analysis intention expression.
The server pre-trains the sequence model based on case information of a large number of real historical cases. The sequence model is used for identifying the retrieval analysis intention of the user, namely mining potential information of analysis dimensions, analysis conditions and the like which can reflect the expectations of the user in the retrieval analysis statement. The analysis intention expression may be in the form of Key-value Key value pair, for example, the analysis intention expression corresponding to the search analysis statement "support rate of financial borrowing cases in guangdong region" may be "case by=financial borrowing disputes", "region=guangdong", "judge result=support", "intention=support ratio". For another example, the corresponding analysis intent expressions of the search analysis statement "where cases related to lending disputes are generally distributed" may be "case by= lending disputes OR × borrowing disputes", "intent=region", respectively.
Step 210, inputting the target vector into a preset intention classification model to obtain a target SQL template.
The server presets a variety of SQL templates. Different SQL templates are used to satisfy the user's intent to analyze based on different dimensions and conditions. The server training trains the intent classification model. The intention classification model is used for determining which SQL template to choose according to the current user retrieval analysis intention.
The intention classification model can be obtained by performing supervised training on the basic classification model based on a large number of simulated search analysis sentences and target SQL templates correspondingly marked by each search analysis sentence. The basic training model may be an RNN model (Recurrent neural network, recurrent neural network model). And the server sequentially inputs a plurality of target vectors corresponding to the retrieval analysis sentences into the intention classification model to obtain a target SQL template.
And step 212, filling the analysis intention expression into a target SQL template to obtain an SQL query statement.
The manner in which different SQL templates are populated may be different. And the server fills the analysis intention expression into the target SQL template according to the filling mode of the target SQL template, so that the SQL query statement can be obtained.
In another embodiment, the sequence model further outputs an intent strength corresponding to each analysis intent expression, sorts the analysis intent expressions according to the intent strength, and fills the analysis intent expressions into the target SQL template according to the sorting to obtain the SQL query statement.
In yet another embodiment, each analysis intent expression has a corresponding intent expression word in the search analysis statement, for example, in the search analysis statement "where cases related to lending disputes are generally distributed", the intent expression "case by = lending disputes OR = lending disputes" corresponding intent expression word may be "lending disputes", and the intent = region "corresponding intent expression word may be" where ". The server sequentially fills a plurality of analysis intention expressions into the target SQL template according to the sequence of the intention expression word corresponding to the intention expression word in the retrieval analysis statement.
Step 214, inquiring the related cases in the case statistics table based on the SQL inquiry statement, carrying out statistical analysis on the case information of the related cases, and returning the analysis result to the terminal.
Based on different SQL query sentences, the data query with different dimensionalities such as judge time, region, case, and the like can be realized; and the data statistics of different conditions such as case specific gravity, case number, support rate and the like can be realized. And the server performs data query and statistical analysis in the case statistics table based on the SQL query statement, and returns an analysis result to the terminal.
For the case processing information query of the related cases in the past, the traditional mode directly judges the search intention of the user through the mode of presetting word lists, and does not support the user to search in a natural language mode. In addition, the preset vocabulary not only requires a lot of labor, but also has difficulty in ensuring the coverage rate of the vocabulary information, and once a certain search keyword input by a user is not covered in the vocabulary, search analysis fails. And the application supports the retrieval of the user in a natural language manner. It is easy to understand that natural language can express the search intention of the user more accurately than the single search keyword, so that the search analysis intention of the user can be mined more accurately based on the search analysis statement. The method and the device can further quickly and accurately identify the search analysis intention of the user through the machine learning pre-training sequence model and the intention classification model, and compared with a preset word list, the method and the device can reduce manual participation and realize end-to-end judgment information search analysis in a true sense.
In this embodiment, according to the arbitration analysis request sent by the terminal, a search analysis statement may be obtained; according to the search analysis statement and the table information of the preset case statistics table, a target vector can be generated; inputting the target vector into a preset sequence model to obtain a plurality of analysis intention expressions; inputting the target vector into a preset intention classification model to obtain a target SQL template; filling the analysis intention expression into a target SQL template to obtain an SQL query statement; the related cases are inquired in the case statistics table based on the SQL inquiry sentences, and the case information of the related cases is subjected to statistical analysis, so that an analysis result can be obtained, the analysis result is returned to the terminal, and different retrieval analysis intentions of a user can be met. Based on the sequence model and the intention classification model, the search analysis intention expressed by the user based on the search analysis statement can be accurately identified; by combining a case statistical table which is pre-deconstructed, the retrieval and analysis efficiency of the judging information can be improved, and different retrieval and analysis intentions of a user can be responded quickly.
In one embodiment, before acquiring the case statistics table and the corresponding table information, the method further includes: acquiring case files of a plurality of historical cases; extracting case identifications and one or more factor description sentences of corresponding historical cases from the case file through regular matching; inputting the factor description statement into a preset semantic understanding model to obtain a plurality of case factors; and constructing a case statistics table based on the case factors corresponding to the case identifications.
The extraction mode of different case factors can be different. For the information content directly recorded in the plaintext in the case file, the factor value of the corresponding case factor, such as the judge time, can be obtained by utilizing keyword matching or regular matching. For the factor value which is not recorded in the case factor in the plaintext in the case file, the factor value needs to be refined based on a pre-trained semantic understanding model.
The semantic understanding model can be obtained based on case file training of a large number of sample cases and is used for extracting factor values of target case factors and recording the factor values as target factor values. Specifically, the server screens the description sentences related to the target case factors in the case files of the historical cases through regular matching, and records the description sentences as factor description sentences. And the server marks the target factor value of the screened factor description statement. Different factor description sentences and corresponding target factor values respectively form different samples. Training the initial model to be trained based on a large number of samples to obtain a semantic understanding model. The initial model to be trained may be an X-GBoost model or the like.
To ensure accuracy of the search analysis, the case statistics may be dynamically updated. For example, the case files of newly added cases are crawled on a designated website according to the preset time frequency, the case files are deconstructed according to the mode, and case information obtained through deconstructing is recorded into a case statistics table.
In the embodiment, the case files of a large number of historical cases are deconstructed in advance, the deconstructed case information is utilized to construct the case statistical table, different retrieval analysis intentions of users can be responded quickly based on the case statistical table, the users can know the case processing conditions of the related cases in the past quickly from different angles, and then the efficiency of judging information retrieval analysis is improved.
In one embodiment, the table information includes a plurality of field enumeration values; the sequence model comprises a dimension sequence model and a condition sequence model; the dimension sequence model comprises an encoder, a decoder and an attention module; the analysis intent expression includes an analysis dimension expression and an analysis condition expression. As shown in fig. 3, according to the search analysis statement and table information, a target vector is generated, and the target vector is input into a preset sequence model to obtain an analysis intention expression, that is, a step of determining the analysis intention expression, including:
step 302, word segmentation is performed on the search analysis statement, word vectors of each word segmentation are calculated, and the word vectors are recorded as first vectors.
The server performs word segmentation on the search analysis sentences, and performs optimization processing such as stop word replacement, synonym replacement and the like on the obtained multiple word segments. For example, the term corresponding to the search analysis sentence "specific gravity of contract dispute case in Guangdong area in 2018" in the above example may be "2018", "Guangdong", "area", "contract dispute", "case", "and" specific gravity ". Wherein, the word "region", "case" can be removed as stop words; the term "specific gravity" may be replaced with the synonym "proportion".
And the server performs One-hot independent encoding on each word after optimization processing to obtain a first vector corresponding to each word. It is readily understood that a search analysis statement may correspond to a plurality of first vectors. The first vector may be calculated in other ways, without limitation.
Step 304, a word vector corresponding to each field enumeration value is calculated and recorded as a second vector.
The server calculates the second vector of each field enumeration value in the case statistics table as described above. In another embodiment, the second vector may be pre-calculated by the server and recorded in the case statistics table, so that the vector matching efficiency may be improved, and further, the search analysis efficiency may be improved.
Step 306, calculating the similarity between the first vector and the different second vectors; and splicing the first vector with the second vector with the highest similarity to obtain the target vector.
The server obtains the similarity between the first vector and each second vector by calculating the Euclidean distance between the first vector and the second vector and the like. The server compares whether the highest similarity reaches a threshold. If yes, the server splices the first vector with the second vector with the highest similarity to obtain the target vector. The first vector is spliced with the second vector with high similarity, so that the retrieval intention characteristics of the user are more obvious, and the model identification precision is improved.
If there are multiple second vectors with the highest similarity to the first vector, and the similarity reaches the threshold value, the server in this embodiment randomly splices the first vector with the second vector with the highest similarity. In other words, each word after optimization processing in this embodiment corresponds to a field enumeration value last. For example, if the field enumeration values of the word "guangdong" in the case statistics table are not completely matched and have the same vector similarity as the field enumeration values of "guangdong shenzhen", "guangdong guangzhou", etc., the server randomly concatenates the second vector of one of the field enumeration values having the highest similarity and the same vector of the corresponding word with the first vector of the corresponding word. It is easy to use immediately, other vector splicing methods can be used, and this is not a limitation.
And 308, calling a conditional sequence model to forget a local vector containing analysis dimension information in the target vector, and obtaining one or more analysis conditional expressions.
The sequence model is used to identify a search analysis intent of the user. The sequence model includes a dimensional sequence model and a conditional sequence model. The dimension sequence model and the condition sequence model may be different RNN models, such as LSTM (Long Short-Term Memory network), etc. The search analysis intention refers to statistical analysis of which aspect of case information of which dimensions in the case statistics table the user desires to perform, and includes analysis of dimension intention and analysis of condition intention. The dimension sequence model is used for identifying analysis dimension intention of the user; the condition sequence model is used to identify the user's analysis condition intent.
The same search analysis statement may correspond to multiple target vectors. Forgetting analysis dimension information contained in each target vector through LSTM, and screening to obtain analysis condition field values. The server generates an analysis condition expression from the analysis condition field value. For example, a preset "intention" field may be used as a Key Value, an analysis condition field Value or a Value after conversion of the analysis condition field Value, and a Key-Value Key Value pair formed may be used as an analysis condition expression.
Step 3102, calling the encoder to forget the local vector containing the analysis condition information in the target vector, and obtaining the compressed vector.
Step 3104, the decoder is invoked to decode the compressed vector pair, and obtain an initial matching probability corresponding to each field enumeration value.
In step 3106, the attention module is invoked to perform attention training on the compressed vector, so as to obtain a similarity weight corresponding to each field enumeration value.
Step 3108, adjusting the initial matching probability of the corresponding field enumeration value according to the similarity weighting, to obtain a target matching probability corresponding to each field enumeration value.
In step 3110, an analysis dimension expression is generated from the field enumeration value with highest target matching probability.
The dimensional sequence model includes an encoder, a decoder, and an attention module. The encoder, decoder and attention module may be different RNN models. The encoder is used for encoding the search analysis statement, namely forgetting the local vector which corresponds to the analysis condition information in the plurality of target vectors to obtain a compressed vector. The compressed vector contains the sentence meaning of the search analysis sentence. The decoder is used for carrying out dimension reduction on the compressed vector and calculating the initial matching probability of the target vector and each field enumeration value based on the dimension reduced compressed vector mapping.
The attention module is used for performing attention training on the compressed vector after the dimension reduction, and calculating similarity weighting corresponding to each field enumeration value of the target vector. The decoder is further configured to adjust an initial matching probability of the target vector and the corresponding field enumeration value according to the similarity weighting, so as to obtain target matching probabilities of each target vector and different field enumeration values.
And the server generates an analysis dimension expression corresponding to the corresponding target vector based on the field enumeration value with the highest target matching probability. For example, a header corresponding to a field enumeration Value with the highest target matching probability is used as a Key Value, the field enumeration Value with the highest target matching probability or a field enumeration Value with the highest target matching probability is converted and then used as a Value, and a formed Key-Value Key Value pair can be used as an analysis dimension expression. The conversion processing of the field enumeration value with the highest target matching probability may be to replace a part of fields in the field enumeration value with specified characters such as a sign. For example, the user may perform search analysis on the search analysis statement "specific gravity of disputed cases in Guangdong region contract in 2018" to count only cases in Guangdong region, and if none of the field enumeration values in the case statistics table is "Guangdong", then the field enumeration value "Guangdong Shenzhen" with the highest target matching probability may be converted into "Guangdong x".
In this embodiment, since the information included in the analysis intent expression generated in the above manner is all from the case statistics table, the search analysis result can be ensured, and failure of the statistics result caused by the fact that the search keyword directly expressed by the search analysis statement does not exist in the case statistics table can be avoided, so that the user can be prevented from repeatedly modifying the input search analysis information, and the search analysis efficiency is improved.
In one embodiment, returning the analysis results to the terminal includes: acquiring a chart template associated with a target SQL template; determining a plurality of basic coordinates and coordinate elements according to the coordinate extraction rules recorded by the chart template; extracting coordinate values corresponding to each coordinate element from the analysis result; constructing a target chart based on a plurality of basic coordinates and coordinate elements and coordinate values corresponding to each basic coordinate, and generating a judging result analysis page according to the target chart; determining a corresponding alternative chart type according to the basic coordinates and the coordinate elements; and adding an alternative option corresponding to each alternative chart type in the arbitration result analysis page, and returning the arbitration result analysis page added with the alternative option to the terminal.
Based on the intention classification model, the search analysis intention of the user can be identified, and different chart templates can be selected according to different search analysis intents. The chart type of the chart template may be a line graph, a bar graph, a radar graph, etc. Each graph template is associated with a corresponding coordinate extraction rule. According to the coordinate extraction rule, a plurality of basic coordinates and coordinate elements and coordinate values corresponding to each basic coordinate can be extracted from the analysis result. For example, the target chart type corresponding to the search analysis sentence "proportion of contract dispute resolution cases in the Guangdong region 2018" may be a histogram. Wherein, the abscissa is the contract release result, and the coordinate elements are two discrete values of release and non-release; the ordinate is the case proportion, and the coordinate elements are continuous values of 0-100%.
The alternative chart types corresponding to each target chart are preset, and one-key change of the chart types can be supported. In another embodiment, the user is also supported to make changes to chart elements in the target chart. Specifically, the server receives an adjustment request for the target chart sent by the terminal. The adjustment request carries change information for one or more primitives in the target graph. And the server performs processing such as adding, deleting or changing the display position of the corresponding graphic element in the target graph according to the change information. By adjusting the target chart, the user can conveniently conduct further deep retrieval on the basis of the original retrieval result.
In the embodiment, the analysis result is visually displayed in a chart mode, so that the analysis result is more visual; and simultaneously, a plurality of options of alternative chart types are provided, so that the personalized requirements of the user can be greatly met while the feedback efficiency of the analysis result is ensured.
It should be understood that, although the steps in the flowcharts of fig. 2 to 3 are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in fig. 2-3 may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed sequentially, but may be performed alternately or alternately with at least a portion of the other steps or sub-steps of other steps.
In one embodiment, as shown in fig. 4, there is provided an arbitrated information retrieval analysis device, comprising: an analysis intent recognition module 402, a search analysis statement generation module 404, and a query statistics analysis module 406, wherein:
An analysis intention recognition module 402, configured to receive an arbitration analysis request sent by a terminal; the resolution analysis request carries a retrieval analysis statement; acquiring a case statistical table and corresponding table information; generating a target vector according to the search analysis statement and the table information; and inputting the target vector into a preset sequence model to obtain an analysis intention expression.
The search analysis statement generation module 404 is configured to input a target vector into a preset intention classification model to obtain a target SQL template; and filling the analysis intention expression into a target SQL template to obtain an SQL query statement.
The query statistics analysis module 406 is configured to query related cases in the case statistics table based on the SQL query statement, perform statistics analysis on case information of the related cases, and return an analysis result to the terminal.
In one embodiment, the apparatus further includes a case statistics table construction module 408, configured to obtain case files of a plurality of historical cases; extracting case identifications and one or more factor description sentences of corresponding historical cases from the case file through regular matching; inputting the factor description statement into a preset semantic understanding model to obtain a plurality of case factors; and constructing a case statistics table based on the case factors corresponding to the case identifications.
In one embodiment, the table information includes a plurality of field enumeration values; the analysis intention recognition module 402 is further configured to segment the search analysis sentence, calculate a word vector of each segment, and record the word vector as a first vector; calculating word vectors corresponding to each field enumeration value, and recording the word vectors as second vectors; calculating the similarity between the first vector and different second vectors; and splicing the first vector with the second vector with the highest similarity to obtain the target vector.
In one embodiment, the sequence model includes a dimensional sequence model and a conditional sequence model; analyzing the intent expression includes analyzing the dimension expression and analyzing the condition expression; the analysis intention recognition module 402 is further configured to invoke a dimension sequence model to perform forgetting processing on a local vector containing analysis condition information in the target vector, so as to obtain one or more analysis dimension expressions; and calling a conditional sequence model to forget a local vector containing analysis dimension information in the target vector, so as to obtain one or more analysis dimension expressions.
In one embodiment, the dimension sequence model includes an encoder, a decoder, and an attention module; the analysis intention recognition module 402 is further configured to invoke an encoder to forget a local vector containing analysis condition information in the target vector, so as to obtain a compressed vector; invoking a decoder to decode the compressed vector pair to obtain initial matching probability corresponding to each field enumeration value; invoking an attention module to perform attention training on the compressed vector to obtain similarity weighting corresponding to each field enumeration value; adjusting the initial matching probability of the corresponding field enumeration value according to the similarity weighting to obtain a target matching probability corresponding to each field enumeration value; and generating an analysis dimension expression according to the field enumeration value with the highest target matching probability.
In one embodiment, the query statistics analysis module 406 is further configured to obtain a chart template associated with the target SQL template; determining a plurality of basic coordinates and coordinate elements according to the coordinate extraction rules recorded by the chart template; extracting coordinate values corresponding to each coordinate element from the analysis result; constructing a target chart based on a plurality of basic coordinates and coordinate elements and coordinate values corresponding to each basic coordinate, and generating a judging result analysis page according to the target chart; determining a corresponding alternative chart type according to the basic coordinates and the coordinate elements; and adding an alternative option corresponding to each alternative chart type in the arbitration result analysis page, and returning the arbitration result analysis page added with the alternative option to the terminal.
The specific limitation regarding the apparatus for the analysis of the information retrieval is referred to above as limitation of the method for the analysis of the information retrieval, and will not be described in detail herein. The respective modules in the above-described arbitration information retrieval analysis means may be implemented in whole or in part by software, hardware, or a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 5. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured 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, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used for storing a case statistics table, an SQL template and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of arbitration information retrieval analysis.
It will be appreciated by those skilled in the art that the structure shown in fig. 5 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the arbitration information retrieval analysis method provided in any of the embodiments of the present application.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples represent only a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (10)
1. A method of arbitrating information retrieval analysis, the method comprising:
receiving a judging and analyzing request sent by a terminal; the arbitration analysis request carries a retrieval analysis statement;
acquiring a case statistical table and corresponding table information; the step of determining the case statistics table comprises the following steps: acquiring case files of a plurality of historical cases; extracting case identifications and one or more factor description sentences of corresponding historical cases from the case file through regular matching; inputting the factor description statement into a preset semantic understanding model to obtain a plurality of case factors; constructing a case statistics table based on a plurality of case identifications and case factors corresponding to each case identification;
Generating a target vector according to the search analysis statement and the table information;
inputting the target vector into a preset sequence model to obtain an analysis intention expression;
inputting the target vector into a preset intention classification model to obtain a target SQL template;
filling the analysis intention expression into the target SQL template to obtain an SQL query statement;
inquiring related cases in the case statistics table based on the SQL inquiry statement, and carrying out statistics analysis on case information of the related cases;
acquiring a chart template associated with the target SQL template;
determining a plurality of basic coordinates and coordinate elements according to the coordinate extraction rules recorded by the chart template;
extracting coordinate values corresponding to each coordinate element from the analysis result;
constructing a target chart based on a plurality of basic coordinates and coordinate elements and coordinate values corresponding to each basic coordinate, and generating a judging result analysis page according to the target chart;
determining a corresponding alternative chart type according to the basic coordinates and the coordinate elements;
and adding an alternative option corresponding to each alternative chart type in the arbitration result analysis page, and returning the arbitration result analysis page added with the alternative option to the terminal.
2. The method of claim 1, wherein the table information includes a plurality of field enumeration values; the generating a target vector according to the search analysis statement and the table information comprises the following steps:
performing word segmentation on the search analysis statement, calculating word vectors of each word segmentation, and recording the word vectors as first vectors;
calculating word vectors corresponding to each field enumeration value, and recording the word vectors as second vectors;
calculating the similarity of the first vector and different second vectors;
and splicing the first vector with the second vector with the highest similarity to obtain the target vector.
3. The method of claim 2, wherein the sequence model comprises a dimensional sequence model and a conditional sequence model; the analysis intent expression includes an analysis dimension expression; inputting the target vector into a preset sequence model to obtain an analysis intention expression, wherein the analysis intention expression comprises the following steps:
calling the dimension sequence model to forget a local vector containing analysis condition information in the target vector, so as to obtain one or more analysis dimension expressions;
and calling the conditional sequence model to forget the local vector containing the analysis dimension information in the target vector, so as to obtain one or more analysis dimension expressions.
4. The method of claim 3, wherein the dimensional sequence model comprises an encoder, a decoder, and an attention module; the step of calling the dimension sequence model to perform forgetting processing on a local vector containing analysis condition information in the target vector to obtain one or more analysis dimension expressions, wherein the step of calling the dimension sequence model comprises the following steps:
calling the encoder to forget a local vector containing analysis condition information in the target vector to obtain a compressed vector;
invoking the decoder to decode the compressed vector to obtain initial matching probability corresponding to each field enumeration value;
invoking the attention module to perform attention training on the compressed vector to obtain similarity weighting corresponding to each field enumeration value;
adjusting the initial matching probability of the corresponding field enumeration value according to the similarity weighting to obtain a target matching probability corresponding to each field enumeration value;
and generating an analysis dimension expression according to the field enumeration value with the highest target matching probability.
5. An apparatus for arbitrating information retrieval analysis, the apparatus comprising:
the analysis intention recognition module is used for receiving an arbitration analysis request sent by the terminal; the arbitration analysis request carries a retrieval analysis statement; acquiring a case statistical table and corresponding table information; generating a target vector according to the search analysis statement and the table information; inputting the target vector into a preset sequence model to obtain an analysis intention expression;
The retrieval analysis statement generation module is used for inputting the target vector into a preset intention classification model to obtain a target SQL template; filling the analysis intention expression into the target SQL template to obtain an SQL query statement;
the query statistics analysis module is used for querying related cases in the case statistics table based on the SQL query statement and carrying out statistics analysis on the case information of the related cases; acquiring a chart template associated with the target SQL template; determining a plurality of basic coordinates and coordinate elements according to the coordinate extraction rules recorded by the chart template; extracting coordinate values corresponding to each coordinate element from the analysis result; constructing a target chart based on a plurality of basic coordinates and coordinate elements and coordinate values corresponding to each basic coordinate, and generating a judging result analysis page according to the target chart; determining a corresponding alternative chart type according to the basic coordinates and the coordinate elements; adding an alternative option corresponding to each alternative chart type in the arbitration result analysis page, and returning the arbitration result analysis page added with the alternative option to the terminal;
the apparatus further comprises:
the case statistics table construction module is used for acquiring case files of a plurality of historical cases; extracting case identifications and one or more factor description sentences of corresponding historical cases from the case file through regular matching; inputting the factor description statement into a preset semantic understanding model to obtain a plurality of case factors; and constructing a case statistical table based on the case identifications and the case factors corresponding to the case identifications.
6. The apparatus of claim 5, wherein the table information comprises a plurality of field enumeration values; the analysis intention recognition module is further to: performing word segmentation on the search analysis statement, calculating word vectors of each word segmentation, and recording the word vectors as first vectors; calculating word vectors corresponding to each field enumeration value, and recording the word vectors as second vectors; calculating the similarity of the first vector and different second vectors; and splicing the first vector with the second vector with the highest similarity to obtain the target vector.
7. The apparatus of claim 6, wherein the sequence model comprises a dimensional sequence model and a conditional sequence model; the analysis intent expression includes an analysis dimension expression; the analysis intention recognition module is further to: calling the dimension sequence model to forget a local vector containing analysis condition information in the target vector, so as to obtain one or more analysis dimension expressions; and calling the conditional sequence model to forget the local vector containing the analysis dimension information in the target vector, so as to obtain one or more analysis dimension expressions.
8. The apparatus of claim 7, wherein the dimensional sequence model comprises an encoder, a decoder, and an attention module; the analysis intention recognition module is further to: calling the encoder to forget a local vector containing analysis condition information in the target vector to obtain a compressed vector; invoking the decoder to decode the compressed vector to obtain initial matching probability corresponding to each field enumeration value; invoking the attention module to perform attention training on the compressed vector to obtain similarity weighting corresponding to each field enumeration value; adjusting the initial matching probability of the corresponding field enumeration value according to the similarity weighting to obtain a target matching probability corresponding to each field enumeration value; and generating an analysis dimension expression according to the field enumeration value with the highest target matching probability.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 4 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910520201.7A CN110362798B (en) | 2019-06-17 | 2019-06-17 | Method, apparatus, computer device and storage medium for judging information retrieval analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910520201.7A CN110362798B (en) | 2019-06-17 | 2019-06-17 | Method, apparatus, computer device and storage medium for judging information retrieval analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110362798A CN110362798A (en) | 2019-10-22 |
| CN110362798B true CN110362798B (en) | 2023-12-19 |
Family
ID=68217358
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910520201.7A Active CN110362798B (en) | 2019-06-17 | 2019-06-17 | Method, apparatus, computer device and storage medium for judging information retrieval analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110362798B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110888897B (en) * | 2019-11-12 | 2020-07-14 | 杭州世平信息科技有限公司 | Method and device for generating SQ L statement according to natural language |
| CN111597205B (en) * | 2020-05-26 | 2024-02-13 | 北京金堤科技有限公司 | Template configuration method, information extraction device, electronic equipment and medium |
| CN113515621B (en) * | 2021-04-02 | 2024-03-29 | 中国科学院深圳先进技术研究院 | Data retrieval method, device, equipment and computer readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975488A (en) * | 2016-04-25 | 2016-09-28 | 哈尔滨工程大学 | Method for querying keyword based on topic cluster unit in relational database |
| CN107885874A (en) * | 2017-11-28 | 2018-04-06 | 上海智臻智能网络科技股份有限公司 | Data query method and apparatus, computer equipment and computer-readable recording medium |
| CN109241159A (en) * | 2018-08-07 | 2019-01-18 | 威富通科技有限公司 | A kind of subregion querying method, system and the terminal device of data cube |
| CN109522393A (en) * | 2018-10-11 | 2019-03-26 | 平安科技（深圳）有限公司 | Intelligent answer method, apparatus, computer equipment and storage medium |
| CN109542956A (en) * | 2018-10-17 | 2019-03-29 | 深圳壹账通智能科技有限公司 | Report form generation method, device, computer equipment and storage medium |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002140339A (en) * | 2000-10-31 | 2002-05-17 | Tonfuu:Kk | System, device and program for retrieving law and the like |
| CN107256267B (en) * | 2017-06-19 | 2020-07-24 | 北京百度网讯科技有限公司 | Query method and device |
-
2019
- 2019-06-17 CN CN201910520201.7A patent/CN110362798B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105975488A (en) * | 2016-04-25 | 2016-09-28 | 哈尔滨工程大学 | Method for querying keyword based on topic cluster unit in relational database |
| CN107885874A (en) * | 2017-11-28 | 2018-04-06 | 上海智臻智能网络科技股份有限公司 | Data query method and apparatus, computer equipment and computer-readable recording medium |
| CN109241159A (en) * | 2018-08-07 | 2019-01-18 | 威富通科技有限公司 | A kind of subregion querying method, system and the terminal device of data cube |
| CN109522393A (en) * | 2018-10-11 | 2019-03-26 | 平安科技（深圳）有限公司 | Intelligent answer method, apparatus, computer equipment and storage medium |
| CN109542956A (en) * | 2018-10-17 | 2019-03-29 | 深圳壹账通智能科技有限公司 | Report form generation method, device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110362798A (en) | 2019-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111160017B (en) | Keyword extraction method, phonetics scoring method and phonetics recommendation method | |
| CN110765763B (en) | Error correction method and device for voice recognition text, computer equipment and storage medium | |
| CN109829629B (en) | Risk analysis report generation method, apparatus, computer device and storage medium | |
| WO2021169111A1 (en) | Resume screening method and apparatus, computer device and storage medium | |
| WO2020077895A1 (en) | Signing intention determining method and apparatus, computer device, and storage medium | |
| EP3855324A1 (en) | Associative recommendation method and apparatus, computer device, and storage medium | |
| CN110377632B (en) | Litigation result prediction method, litigation result prediction device, litigation result prediction computer device and litigation result prediction storage medium | |
| CN111444723B (en) | Information extraction method, computer device, and storage medium | |
| CN109858010B (en) | Method and device for recognizing new words in field, computer equipment and storage medium | |
| CN111666401B (en) | Document recommendation method, device, computer equipment and medium based on graph structure | |
| CN110377558B (en) | Document query method, device, computer equipment and storage medium | |
| CN110674319A (en) | Label determination method and device, computer equipment and storage medium | |
| CN110377631B (en) | Case information processing method, device, computer equipment and storage medium | |
| CN110362798B (en) | Method, apparatus, computer device and storage medium for judging information retrieval analysis | |
| CN110458324B (en) | Method and device for calculating risk probability and computer equipment | |
| CN106708929B (en) | Video program searching method and device | |
| CN110377618B (en) | Method, device, computer equipment and storage medium for analyzing decision result | |
| CN110674131A (en) | Financial statement data processing method and device, computer equipment and storage medium | |
| CN111985228A (en) | Text keyword extraction method and device, computer equipment and storage medium | |
| CN112651236B (en) | Method and device for extracting text information, computer equipment and storage medium | |
| CN114218958A (en) | Work order processing method, device, equipment and storage medium | |
| CN110532229B (en) | Evidence file retrieval method, device, computer equipment and storage medium | |
| CN106570196B (en) | Video program searching method and device | |
| CN110362592B (en) | Method, device, computer equipment and storage medium for pushing arbitration guide information | |
| US11880798B2 (en) | Determining section conformity and providing recommendations |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |