CN116861337A - Electric power engineering label draws and discernment platform based on fuse LSTM - Google Patents
Electric power engineering label draws and discernment platform based on fuse LSTM Download PDFInfo
- Publication number
- CN116861337A CN116861337A CN202310679376.9A CN202310679376A CN116861337A CN 116861337 A CN116861337 A CN 116861337A CN 202310679376 A CN202310679376 A CN 202310679376A CN 116861337 A CN116861337 A CN 116861337A
- Authority
- CN
- China
- Prior art keywords
- electrically connected
- item
- output end
- lstm
- call
- 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.)
- Pending
Links
- 238000000605 extraction Methods 0.000 claims abstract description 61
- 238000007726 management method Methods 0.000 claims abstract description 40
- 238000000034 method Methods 0.000 claims abstract description 37
- 238000009825 accumulation Methods 0.000 claims abstract description 15
- 230000001174 ascending effect Effects 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 22
- 238000004140 cleaning Methods 0.000 claims description 9
- 238000012790 confirmation Methods 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 7
- 238000013519 translation Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 5
- 238000007792 addition Methods 0.000 claims description 4
- 238000012217 deletion Methods 0.000 claims description 4
- 230000037430 deletion Effects 0.000 claims description 4
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Computing Systems (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computational Mathematics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Public Health (AREA)
- Strategic Management (AREA)
- Water Supply & Treatment (AREA)
- Algebra (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Databases & Information Systems (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of power engineering label extraction, in particular to a power engineering label extraction and identification platform based on a fused LSTM. In the invention, the library files are classified in the model library based on classification basis logic, the storage address sequence is updated through a library management program, the multi-level tag tree and the identification items are managed in the LSTM reference model, the frequency comprehensive operation is executed according to the classification accumulation operation and the proportion operation through the hierarchical call management method, the order of ascending and descending is carried out, and the storage address sequence is managed, so that in the call process, the call items with high levels are preferentially called, the call speed is improved, and the method is suitable for the large-batch call requirements in the electric power engineering.
Description
Technical Field
The invention relates to the technical field of power engineering label extraction, in particular to a power engineering label extraction and identification platform based on fusion LSTM.
Background
In the practical method of extracting the power engineering label, the existing power engineering label extracting means is often based on a big data joint check and search mode to realize the retrieval and comparison of the complete label content, so that the label extracting efficiency is low, the identification progress is slow, and the method is difficult to adapt to the large-batch data requirement of the power engineering and needs to be improved.
Disclosure of Invention
The invention aims to solve the defects in the prior art, and provides an electric power engineering label extraction and identification platform based on fusion LSTM.
In order to achieve the above purpose, the present invention adopts the following technical scheme: the utility model provides a power engineering label draws and discernment platform based on fuse LSTM comprises power engineering operation system, data acquisition system, log database, feature extraction module, model training module, model storehouse, power engineering operation system's output electric connection has the data stream, data acquisition system includes along with flowing the record subassembly, along with flowing the output of record subassembly and the input electric connection of data stream, data acquisition system's output electric connection has the operation log file, the output of operation log file and the input electric connection of log database, the output of log database and the input electric connection of feature extraction module, the output of feature extraction module and the input electric connection of model training module.
As a further aspect of the invention: the output end of the characteristic extraction module is electrically connected with a hierarchical calling management method, the output end of the hierarchical calling management method is electrically connected with the input end of a model library, the model library comprises an LSTM reference model, and the output end of the LSTM reference model is electrically connected with an electric power engineering tag library and an adding library.
As a further aspect of the invention: the model training module comprises a word sense identification method and a word sense association method, wherein the output ends of the word sense identification method and the word sense association method are electrically connected with accuracy verification, and the output end of the accuracy verification is electrically connected with the input end of the LSTM reference model.
As a further aspect of the invention: the model library comprises a classified library file, the classified library file comprises a library management program, the library management program comprises storage address management and addition and deletion management, the output end of the library management program is electrically connected with an LSTM reference model, the LSTM reference model comprises a multi-stage label tree and an identification item, the multi-stage label tree comprises an inner nested index, and the multi-stage label tree is electrically connected with the identification item through the associated index.
As a further aspect of the invention: the output end of the classification library file is electrically connected with classification basis logic, the classification basis logic comprises a coordinate system, data bottom layer logic, a belonging management and control area, a time period, item class and item refinement, and the output end of the coordinate system, the data bottom layer logic, the belonging management and control area, the time period, the item class and the item refinement is electrically connected with a tag keyword.
As a further aspect of the invention: the hierarchical calling management method comprises calling hierarchical preset and calling item cleaning, wherein the output end of the calling hierarchical preset is electrically connected with a hierarchical setting, and the hierarchical setting comprises a first-level calling item and a second-level calling item … … N-level calling item.
As a further aspect of the invention: the output end of the level setting is electrically connected with a time axis section setting, the output end of the time axis section setting is electrically connected with a label item call, the output end of the label item call is electrically connected with a classified accumulation operation and a proportion operation, the output end of the classified accumulation operation and the proportion operation is electrically connected with a frequency comprehensive operation, the output end of the frequency comprehensive operation is electrically connected with a level ascending and descending order and a cleaning task release, the output end of the level ascending and descending order is electrically connected with the input end of the level setting, and the output end of the cleaning task release is electrically connected with the input end of the call item cleaning;
the classified accumulation operation adopts a secondary accumulation formula:
C X1 ＝C X1 +1
wherein C is X1 Cumulative amount of call times for call item x 1;
the proportional operation adopts a proportional formula:
wherein P is X1 C is the percentage of the total call quantity of the call item x1 in the time axis section X2 To accumulate the number of calls of the call item x2, C Xn The number of calls for call item xn is accumulated.
As a further aspect of the invention: the output end of the feature extraction module is electrically connected with a single log extraction, the output end of the single log extraction is electrically connected with a log data item comparison unit, the output end of the log data item comparison unit is electrically connected with log entry, and the log entry comprises short sentence processing, keyword extraction and keyword translation.
As a further aspect of the invention: the output end of the word sense identification method and the output end of the word sense association method are electrically connected with fuzzy word retrieval, the output end of the fuzzy word retrieval is electrically connected with a tag retrieval item list, the output end of the tag retrieval item list is electrically connected with retrieval item judgment, the output end of the retrieval item judgment is electrically connected with semantic judgment and keyword corresponding item judgment, the output end of the semantic judgment and keyword corresponding item judgment is electrically connected with tag confirmation, the output end of the tag confirmation is electrically connected with LSTM reference model call, the output end of the LSTM reference model call is electrically connected with tag pre-entry and identifier pre-entry, and the output end of the tag pre-entry and identifier pre-entry is electrically connected with a data identification file.
As a further aspect of the invention: the output of data identification file and the input electric connection of accuracy check, the output electric connection of accuracy check has data identification file to be typeeed, the output electric connection that data identification file was typeeed has label item contrast, sign content contrast, the output electric connection of label item contrast, sign content contrast has log file to translate, the output electric connection that log file translated has the accuracy to judge, the output electric connection of accuracy judgement has the entry, the output electric connection of entry has the entry search, the output electric connection of entry search has the model training module to invoke.
Compared with the prior art, the invention has the advantages and positive effects that:
in the invention, the library files are classified in the model library based on classification basis logic, the storage address sequence is updated through a library management program, the multi-level tag tree and the identification items are managed in the LSTM reference model, the frequency comprehensive operation is executed according to the classification accumulation operation and the proportion operation through the hierarchical call management method, the order of ascending and descending is carried out, and the storage address sequence is managed, so that in the call process, the call items with high levels are preferentially called, the call speed is improved, and the method is suitable for the large-batch call requirements in the electric power engineering.
Drawings
Fig. 1 is a schematic diagram of a system main frame of a power engineering label extraction and identification platform based on a fusion LSTM;
FIG. 2 is a flowchart of model library refinement of a power engineering label extraction and identification platform based on a fused LSTM;
FIG. 3 is a detailed flow chart of a hierarchical call management method based on a fused LSTM (least squares) power engineering label extraction and identification platform;
FIG. 4 is a detailed flow chart of a feature extraction module of an electric power engineering label extraction and identification platform based on a fusion LSTM;
FIG. 5 is a detailed flow chart of a model training module of an electric power engineering label extraction and recognition platform based on a fusion LSTM;
fig. 6 is a flowchart for verifying and refining the accuracy of the power engineering label extraction and recognition platform based on the fusion LSTM.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
In the description of the present invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate orientations or positional relationships based on the orientation or positional relationships shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, in the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
Example 1
Referring to fig. 1, an electric power engineering label extracting and identifying platform based on an integrated LSTM is composed of an electric power engineering operation system, a data acquisition system, a log database, a feature extraction module, a model training module and a model library, wherein an output end of the electric power engineering operation system is electrically connected with a data stream, the data acquisition system comprises a stream following recording assembly, an output end of the stream following recording assembly is electrically connected with an input end of the data stream, an output end of the data acquisition system is electrically connected with an operation log file, an output end of the operation log file is electrically connected with an input end of the log database, an output end of the log database is electrically connected with an input end of the feature extraction module, and an output end of the feature extraction module is electrically connected with an input end of the model training module.
Specifically, in the power engineering label extraction and identification platform integrated with the LSTM, in the operation process of a power engineering operation system, a data stream is generated based on operation data, a data acquisition system acquires the data stream through a stream following recording component and generates an operation log file, the operation log file is stored in a log database, and a feature extraction module extracts the operation data file in the log database and transmits the operation data file to a model training module to participate in training.
Referring to fig. 1, an output end of the feature extraction module is electrically connected with a hierarchical call management method, an output end of the hierarchical call management method is electrically connected with an input end of a model library, the model library comprises an LSTM reference model, and an output end of the LSTM reference model is electrically connected with an electric power engineering tag library and an addition library.
Specifically, the feature extraction module specifically adopts a hierarchical call management method to perform extraction work, stores an LSTM reference model through a model library, and uses the LSTM reference model as an extraction basis, and uses an electric power engineering tag library and an addition library as data sources.
Referring to fig. 1, the model training module includes a word sense recognition method and a word sense association method, wherein the output ends of the word sense recognition method and the word sense association method are electrically connected with a precision check, and the output end of the precision check is electrically connected with the input end of the LSTM reference model.
Specifically, in the training process of the model training module, accuracy verification is performed based on a word sense recognition method, recognition and association results of the word sense association method, and based on the accuracy verification result and recognition and association contents, the information is stored in the LSTM reference model.
Referring to fig. 2, the model library includes a classification library file, the classification library file includes a library management program, the library management program includes storage address management and addition and deletion management, an output end of the library management program is electrically connected with an LSTM reference model, the LSTM reference model includes a multi-level tag tree and an identifier, the multi-level tag tree includes an inner nested index, the multi-level tag tree is electrically connected with the identifier through an association index, an output end of the classification library file is electrically connected with a classification basis logic, the classification basis logic includes a coordinate system, a data bottom logic, a belonging management and control area, a time period, a project class and a project refinement, and an output end of the coordinate system, the data bottom logic, the belonging management and control area, the time period, the project class and the project refinement is electrically connected with a tag keyword.
Specifically, in the model library, the classification library file performs classification storage work based on classification basis logic, the classification basis logic comprises a coordinate system, data bottom layer logic, a belonging management and control area, a time period, item class, item refinement and other contents, the content comprises corresponding tag keywords, a basis is provided for classification, the classification library file is loaded with a library management program, wherein the library management program comprises storage address management and addition and deletion management, an LSTM reference model stores the existing tags based on a multi-level tag tree and an identification item, and manages the association relation between the tags based on an inner nested index, and manages the association relation between the multi-level tag tree and the identification item based on the association index.
Referring to fig. 3, the hierarchical call management method includes calling a hierarchical preset, and clearing a call item, wherein an output end of the hierarchical preset is electrically connected with a level setting, the level setting includes a level call item and a level call item … … N, an output end of the level setting is electrically connected with a time axis section setting, an output end of the time axis section setting is electrically connected with a tag item call, an output end of the tag item call is electrically connected with a classified accumulation operation and a proportional operation, an output end of the classified accumulation operation and the proportional operation is electrically connected with a frequency comprehensive operation, an output end of the frequency comprehensive operation is electrically connected with a level ascending and descending order and a clearing task release, an output end of the level ascending and descending order is electrically connected with an input end of the level setting, and an output end of the clearing task release is electrically connected with an input end of the call item clearing;
the classified accumulation operation adopts a secondary accumulation formula:
C X1 ＝C X1 +1
wherein C is X1 Cumulative amount of call times for call item x 1;
the proportional operation adopts a proportional formula:
wherein P is X1 C is the percentage of the total call quantity of the call item x1 in the time axis section X2 To accumulate the number of calls of the call item x2, C Xn The number of calls for call item xn is accumulated.
Specifically, in the hierarchical call management method, call hierarchical preset is set up, and the storage addresses of the first-level call item and the second-level call item … … N-level call item are managed, so that in the call process, call items with high levels are preferentially called, call speed is improved, in the time axis section setting, based on a label item call result, frequency comprehensive operation is executed according to classified accumulation operation and proportional operation, and in this way, the order of ascending and descending is conducted, the storage addresses are managed sequentially, or in a cleaning task release mode, the unusual call items are cleaned.
Referring to fig. 4, an output end of the feature extraction module is electrically connected with a single log extraction, an output end of the single log extraction is electrically connected with a log data item comparison unit, and an output end of the log data item comparison unit is electrically connected with a log entry, wherein the log entry comprises a short sentence processing, a keyword extraction and a keyword translation.
Specifically, in the feature extraction module, a single log is firstly extracted, log entry work is carried out through a log data item comparison unit, and the log entry work comprises short sentence processing, keyword extraction and keyword translation.
Referring to fig. 5, the output end of the word sense recognition method and the output end of the word sense association method are electrically connected with a fuzzy word search, the output end of the fuzzy word search is electrically connected with a tag search item list, the output end of the tag search item list is electrically connected with a search item judgment, the output end of the search item judgment is electrically connected with a semantic judgment and a keyword corresponding item judgment, the output end of the semantic judgment and the keyword corresponding item judgment is electrically connected with a tag confirmation, the output end of the tag confirmation is electrically connected with an LSTM reference model call, the output end of the LSTM reference model call is electrically connected with a tag pre-entry and an identification item pre-entry, and the output end of the tag pre-entry and the identification item pre-entry is electrically connected with a data identification file.
Specifically, the word sense identification method and the word sense association method are based on a fuzzy word retrieval mode, the keywords obtained by the feature extraction module are retrieved, a tag retrieval item list is generated, subsequent retrieval item judgment work is carried out, the retrieval item judgment work specifically comprises semantic judgment and keyword corresponding item judgment, tag confirmation is carried out according to semantic judgment and keyword corresponding item judgment content, after an LSTM reference model is called, tags and identification items are pre-recorded, and a data identification file is generated.
Referring to fig. 6, an output end of the data identification file is electrically connected with an input end of the accuracy check, an output end of the accuracy check is electrically connected with a data identification file input, an output end of the data identification file input is electrically connected with a tag item comparison and an identification content comparison, an output end of the tag item comparison and the identification content comparison is electrically connected with a log file translation, an output end of the log file translation is electrically connected with an accuracy judgment, an output end of the accuracy judgment is electrically connected with an access item, an output end of the access item is electrically connected with an access item search, and an output end of the access item search is electrically connected with a model training module for calling.
Specifically, the input data identification file is checked and input accurately, tag item comparison and identification content comparison are carried out according to input results, the log file is translated, the input and output items are locked based on accuracy judgment results, and the model training module is recalled based on the input and output items to carry out subdivision search work.
Working principle: in the electric power engineering label extraction and identification platform of the integrated LSTM, in the operation process of the electric power engineering operation system, based on operation data, data flow is generated, a data acquisition system acquires the data flow through a flow record component and generates an operation log file, the operation log file is stored in a log database, a feature extraction module extracts the operation data file in the log database and is trained by a model training module, the feature extraction module specifically adopts a hierarchical calling management method to carry out extraction work and stores an LSTM reference model through a model library as extraction basis, the LSTM reference model is based on the electric power engineering label library and an addition library as a data source, in the training process of the model training module, based on a word sense identification method, the identification and the association result of the word sense association method, the accuracy check result is carried out, based on the accuracy check result and the identification association content, the identification is stored in the LSTM reference model, the model library is stored in the model library, the classification library file is classified and stored based on a classification basis logic, the classification basis of the contents such as a coordinate system, a data bottom layer logic, a multi-level management area, a time period, a project class, a project refinement project and the like are called by a corresponding label keyword are included, the classification library is provided with a management item, wherein the classification library comprises a memory management item, the memory management method, the memory is called by a hierarchical calling management item is based on the priority label, the relation is set up by a hierarchical relation, the relation is based on the prior label, the relation is set up by setting up in the hierarchical relation, and the hierarchical relation is based on the relation between the relative label management item, and is stored in the hierarchical relation, by setting, and the relation is based on the relation between the hierarchical label management item, and relative item relation is based on the relative item and relative label and relative item is stored by the relative item, and relative item is stored by a higher by a hierarchical relation in a hierarchical relation, and by a relative relation is based on a relative relation between a relative item and a relative item is stored in a relative management item and a relative management item is in a relative management item and a is used, in the feature extraction module, firstly, a single log is extracted, log entry work is carried out through a log data item comparison unit, the log entry work comprises short sentence processing, keyword extraction and keyword translation, a word meaning identification method and a word meaning association method are based on a fuzzy word retrieval mode, keywords obtained by a feature extraction module are retrieved, a tag retrieval item list is generated, subsequent retrieval item judgment work is carried out, the retrieval item judgment work specifically comprises semantic judgment and keyword corresponding item judgment, tag confirmation is carried out according to semantic judgment and keyword corresponding item judgment content, after an LSTM reference model is called, a data identification file is generated, a tag item comparison and identification content comparison are carried out according to an entry result, the log file is carried out, and the entry and exit are accurately judged, the entry and exit are locked, and the entry and exit are subdivided based on a training module.
The present invention is not limited to the above embodiments, and any equivalent embodiments which can be changed or modified by the technical disclosure described above can be applied to other fields, but any simple modification, equivalent changes and modification made to the above embodiments according to the technical matter of the present invention will still fall within the scope of the technical disclosure.
Claims (10)
1. Electric power engineering label draws and discernment platform based on fuse LSTM, its characterized in that: the utility model discloses a power engineering label draws and discernment platform based on fusion LSTM comprises power engineering operation system, data acquisition system, log database, feature extraction module, model training module, model storehouse, power engineering operation system's output electric connection has the data stream, data acquisition system includes along with flowing the record subassembly, along with flowing the output of record subassembly and the input electric connection of data stream, data acquisition system's output electric connection has the operation log file, the output of operation log file and the input electric connection of log database, the output of log database and the input electric connection of feature extraction module, the output of feature extraction module and the input electric connection of model training module.
2. The fused LSTM based power engineering tag extraction and identification platform of claim 1, wherein: the output end of the characteristic extraction module is electrically connected with a hierarchical calling management method, the output end of the hierarchical calling management method is electrically connected with the input end of a model library, the model library comprises an LSTM reference model, and the output end of the LSTM reference model is electrically connected with an electric power engineering tag library and an adding library.
3. The fused LSTM based power engineering tag extraction and identification platform of claim 1, wherein: the model training module comprises a word sense identification method and a word sense association method, wherein the output ends of the word sense identification method and the word sense association method are electrically connected with accuracy verification, and the output end of the accuracy verification is electrically connected with the input end of the LSTM reference model.
4. The fused LSTM based power engineering tag extraction and identification platform of claim 1, wherein: the model library comprises a classified library file, the classified library file comprises a library management program, the library management program comprises storage address management and addition and deletion management, the output end of the library management program is electrically connected with an LSTM reference model, the LSTM reference model comprises a multi-stage label tree and an identification item, the multi-stage label tree comprises an inner nested index, and the multi-stage label tree is electrically connected with the identification item through the associated index.
5. The fused LSTM based power engineering tag extraction and identification platform of claim 4, wherein: the output end of the classification library file is electrically connected with classification basis logic, the classification basis logic comprises a coordinate system, data bottom layer logic, a belonging management and control area, a time period, item class and item refinement, and the output end of the coordinate system, the data bottom layer logic, the belonging management and control area, the time period, the item class and the item refinement is electrically connected with a tag keyword.
6. The fused LSTM based power engineering tag extraction and identification platform of claim 2, wherein: the hierarchical calling management method comprises calling hierarchical preset and calling item cleaning, wherein the output end of the calling hierarchical preset is electrically connected with a hierarchical setting, and the hierarchical setting comprises a first-level calling item and a second-level calling item … … N-level calling item.
7. The fused LSTM based power engineering tag extraction and identification platform of claim 6, wherein: the output end of the level setting is electrically connected with a time axis section setting, the output end of the time axis section setting is electrically connected with a label item call, the output end of the label item call is electrically connected with a classified accumulation operation and a proportion operation, the output end of the classified accumulation operation and the proportion operation is electrically connected with a frequency comprehensive operation, the output end of the frequency comprehensive operation is electrically connected with a level ascending and descending order and a cleaning task release, the output end of the level ascending and descending order is electrically connected with the input end of the level setting, and the output end of the cleaning task release is electrically connected with the input end of the call item cleaning;
the classified accumulation operation adopts a secondary accumulation formula:
C X1 ＝C X1 +1
wherein C is X1 Cumulative amount of call times for call item x 1;
the proportional operation adopts a proportional formula:
wherein P is X1 C is the percentage of the total call quantity of the call item x1 in the time axis section X2 To accumulate the number of calls of the call item x2, C Xn The number of calls for call item xn is accumulated.
8. The fused LSTM based power engineering tag extraction and identification platform of claim 1, wherein: the output end of the feature extraction module is electrically connected with a single log extraction, the output end of the single log extraction is electrically connected with a log data item comparison unit, the output end of the log data item comparison unit is electrically connected with log entry, and the log entry comprises short sentence processing, keyword extraction and keyword translation.
9. The fused LSTM based power engineering label extraction and identification platform of claim 3, wherein: the output end of the word sense identification method and the output end of the word sense association method are electrically connected with fuzzy word retrieval, the output end of the fuzzy word retrieval is electrically connected with a tag retrieval item list, the output end of the tag retrieval item list is electrically connected with retrieval item judgment, the output end of the retrieval item judgment is electrically connected with semantic judgment and keyword corresponding item judgment, the output end of the semantic judgment and keyword corresponding item judgment is electrically connected with tag confirmation, the output end of the tag confirmation is electrically connected with LSTM reference model call, the output end of the LSTM reference model call is electrically connected with tag pre-entry and identifier pre-entry, and the output end of the tag pre-entry and identifier pre-entry is electrically connected with a data identification file.
10. The fused LSTM based power engineering tag extraction and identification platform of claim 9, wherein: the output of data identification file and the input electric connection of accuracy check, the output electric connection of accuracy check has data identification file to be typeeed, the output electric connection that data identification file was typeeed has label item contrast, sign content contrast, the output electric connection of label item contrast, sign content contrast has log file to translate, the output electric connection that log file translated has the accuracy to judge, the output electric connection of accuracy judgement has the entry, the output electric connection of entry has the entry search, the output electric connection of entry search has the model training module to invoke.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310679376.9A CN116861337A (en) | 2023-06-08 | 2023-06-08 | Electric power engineering label draws and discernment platform based on fuse LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310679376.9A CN116861337A (en) | 2023-06-08 | 2023-06-08 | Electric power engineering label draws and discernment platform based on fuse LSTM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116861337A true CN116861337A (en) | 2023-10-10 |
Family
ID=88218094
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310679376.9A Pending CN116861337A (en) | 2023-06-08 | 2023-06-08 | Electric power engineering label draws and discernment platform based on fuse LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116861337A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117743261A (en) * | 2023-11-30 | 2024-03-22 | 广西壮族自治区自然资源信息中心 | Authentication method and system for natural resource government data tag |
-
2023
- 2023-06-08 CN CN202310679376.9A patent/CN116861337A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117743261A (en) * | 2023-11-30 | 2024-03-22 | 广西壮族自治区自然资源信息中心 | Authentication method and system for natural resource government data tag |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111723215B (en) | Device and method for establishing biotechnological information knowledge graph based on text mining | |
| CN101957816B (en) | Webpage metadata automatic extraction method and system based on multi-page comparison | |
| CN107992633A (en) | Electronic document automatic classification method and system based on keyword feature | |
| CN104679867B (en) | Address method of knowledge processing and device based on figure | |
| US20060184516A1 (en) | Search engine | |
| CN112463971B (en) | E-commerce commodity classification method and system based on hierarchical combination model | |
| CN103593410A (en) | System for search recommendation by means of replacing conceptual terms | |
| CN110633365A (en) | Word vector-based hierarchical multi-label text classification method and system | |
| CN107180075A (en) | The label automatic generation method of text classification integrated level clustering | |
| CN110442728A (en) | Sentiment dictionary construction method based on word2vec automobile product field | |
| CN103324700A (en) | Noumenon concept attribute learning method based on Web information | |
| CN110889003B (en) | Vehicle image fine-grained retrieval system based on text | |
| CN116861337A (en) | Electric power engineering label draws and discernment platform based on fuse LSTM | |
| PT1288792E (en) | A method for automatically indexing documents | |
| CN113033198B (en) | Similar text pushing method and device, electronic equipment and computer storage medium | |
| CN109871473A (en) | A kind of method of pair of project file and Database full-text search document | |
| CN113919366A (en) | Semantic matching method and device for power transformer knowledge question answering | |
| CN110188189A (en) | A kind of method that Knowledge based engineering adaptive event index cognitive model extracts documentation summary | |
| KR20080007740A (en) | System and method for retrieving/classifying web ontology | |
| CN113343701B (en) | Extraction method and device for text named entities of power equipment fault defects | |
| CN110688461B (en) | Online text education resource label generation method integrating multi-source knowledge | |
| CN110210033B (en) | Chinese basic chapter unit identification method based on main bit theory | |
| CN115982316A (en) | Multi-mode-based text retrieval method, system and medium | |
| CN110020034A (en) | A kind of information citation analysis method and system | |
| CN114610744A (en) | Data query method and device and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |