CN114611485A - Intelligent vehicle fault diagnosis method combining text analysis and machine learning method - Google Patents
Intelligent vehicle fault diagnosis method combining text analysis and machine learning method Download PDFInfo
- Publication number
- CN114611485A CN114611485A CN202210262718.2A CN202210262718A CN114611485A CN 114611485 A CN114611485 A CN 114611485A CN 202210262718 A CN202210262718 A CN 202210262718A CN 114611485 A CN114611485 A CN 114611485A
- Authority
- CN
- China
- Prior art keywords
- model
- machine learning
- text
- fault diagnosis
- data
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to the technical field of fault diagnosis, and particularly provides an intelligent vehicle fault diagnosis method combining text analysis and machine learning methods, which specifically comprises the following steps: training on text data and non-text characteristic data respectively by using a language model and a machine learning model to obtain two fault diagnosis models; training a multi-level integration model to integrate two fault diagnosis model results to achieve an optimal fault diagnosis result; according to the method, the text model and the machine learning model are fused by a model integration method, so that the implicit knowledge in the text data and the non-text data is fully learned, the model fusion effect is optimized, and the vehicle fault diagnosis accuracy rate is improved.
Description
Technical Field
The invention relates to the technical field of fault diagnosis, in particular to an intelligent vehicle fault diagnosis method combining text analysis and machine learning methods.
Background
The fault diagnosis is a systematic method for solving problems, the fault diagnosis aims to determine the reason that some functions do not work as expected and a method for solving the problems, and some common methods can help to complete fault diagnosis tasks, such as making expert rules to analyze faults, building a machine learning model by using data characteristics to analyze the faults and the like, and good results are obtained in some specific scenes.
At present, the method is widely applied to the fields of industrial production, instrument manufacturing, vehicle quality inspection and the like, and fault diagnosis is mainly performed on single-mode data such as texts, images, continuous or discrete characteristic data and the like at present, however, in the face of multi-mode data, a single-mode model cannot fully utilize data, important information is lost, the analysis dimension is narrow, and the precision and the effect are limited.
Today, mainstream fault diagnosis technology is widely applied in industrial production and vehicle quality inspection industries, but fault diagnosis technology is still a challenging task. First, most models today are fault-diagnosable, but are often single-modality models, and there are fewer methods of using multi-modality models. Secondly, in the case of multi-modal data, the diagnosis results of multiple independent models are often different or even opposite, and it is difficult to fully learn the implicit knowledge in the multi-modal data. Thirdly, the wrong multi-model fusion results improve the instability of the diagnosis effect and even reduce the diagnosis effect.
Based on the method, the invention provides an intelligent vehicle fault diagnosis method combining text analysis and machine learning methods.
Disclosure of Invention
The invention aims to provide an intelligent vehicle fault diagnosis method combining text analysis and machine learning methods, and aims to solve the problems that although the existing model can be used for fault diagnosis, the model is often used for single-mode model diagnosis, and a multi-mode model is less in utilization.
In order to achieve the purpose, the invention provides the following technical scheme:
the intelligent vehicle fault diagnosis method combining the text analysis method and the machine learning method specifically comprises the following steps:
step 1, training on text data and non-text characteristic data respectively by using a language model and a machine learning model to obtain two fault diagnosis models.
And 2, training the multi-stage integrated model to integrate two fault diagnosis model results to achieve the best fault diagnosis result.
Further, the intelligent vehicle fault diagnosis method combining the text analysis method and the machine learning method specifically further comprises the following steps of establishing multi-mode model training: and respectively training the models of the plurality of modes by using the labeled data.
Further, the specific steps of establishing the multi-modal model training include:
and (3) multi-modal data processing: and dividing the training data into text data and non-text characteristic data, and selecting a corresponding language model and a machine learning model. Training a model: and respectively training a language model and a machine learning model on the text data and the non-text data to obtain a model optimized on the verification set.
The language model adopts a BERT model, takes input fault texts as source data, takes corresponding multiple types of the fault texts as a verification set, and trains the BERT model by using two new unsupervised prediction tasks. Wherein the fault text comprises standard fault phrases and similar fault phrases which are masked with a probability of 15%. The masking steps of the similar fault phrases are as follows:
acquiring input texts with similar fault meanings;
generating a phrase using Autophrase;
the phrases are enhanced and refined using rules and fastText.
Further, the intelligent vehicle fault diagnosis method combining the text analysis method and the machine learning method specifically further comprises the step of fusing the models.
Further, fusing the models specifically includes:
s1, selecting mlp (multilayer perceptron) models to integrate language models and machine learning model results.
And S2, training by using the output values of the text model and the machine learning model as input features, and optimizing the integration effect by adjusting the model layer number, the model weight and the loss function option in the training process.
And S3, obtaining the integrated model with better effect on the verification set.
Further, the intelligent vehicle fault diagnosis method combining the text analysis method and the machine learning method specifically further comprises a model implementation step.
Further, the model implementation step includes: selecting mlp a model integration language model and a machine learning model; when the integrated model is used for prediction, firstly, input data is split into text data and numerical data, the text data and the numerical data enter a text model and a machine learning model respectively for reasoning to obtain output values, and then output probability values of the two models are input into an mlp model as input characteristics for reasoning to obtain a final probability value result.
In conclusion, compared with the prior art, the invention has the following beneficial effects:
the method respectively processes the text characteristic data and the non-text characteristic data through the language model and the machine learning model, thereby avoiding the problem that the effect of uniformly processing the text and numerical data by using a single model such as machine learning on different modal data is poor; the method has the advantages that the text model and the machine learning model are fused by a model integration method, implicit knowledge in text characteristic data and non-text characteristic data is fully learned by combining a text analysis technology and a multi-mode model of data characteristics, the model fusion effect is optimized, the method is used for a vehicle fault diagnosis scene with text characteristic data and non-text characteristic data, the multi-stage integration model is used for processing a language model and the machine learning model to output and determine a final result instead of simple voting or voting with weight to judge the fault type, the fault diagnosis accuracy rate is effectively improved, and the result is more accurate.
Drawings
Fig. 1 is a schematic flow chart of the implementation of the present invention.
Fig. 2 is a schematic flow chart of the implementation of model fusion according to the present invention.
FIG. 3 is a schematic diagram of the present invention for performing multi-modal data processing.
FIG. 4 is a schematic diagram of model training performed in the present invention.
FIG. 5 is a schematic diagram of model fusion performed in the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the protection scope of the present invention.
Example 1
As shown in fig. 1, the intelligent vehicle fault diagnosis method combining text analysis and machine learning according to an embodiment of the present invention specifically includes the following steps:
step 1, training on text characteristic data and non-text characteristic data respectively by using a language model and a machine learning model to obtain two fault diagnosis models.
And 2, training the multi-stage integrated model to integrate two fault diagnosis model results to achieve the best fault diagnosis result.
Step S2 specifically includes:
firstly, the method comprises the following steps: multi-modal model training: and respectively training the models of the plurality of modes by using the labeled data. The method comprises the following specific steps:
s21, multi-modal data processing: and dividing the training data into text characteristic data and non-text characteristic data, and selecting a corresponding language model and a machine learning model.
S22, training a model: and respectively training a language model and a machine learning model on the text characteristic data and the non-text characteristic data to obtain a model which is better on a verification set.
In the embodiment, a Bert language model is preferably adopted for text features, a network structure composed of multiple layers of neural networks is adopted, a section of text is input, the text is analyzed into word vectors, probability values are output by the last layer of neural network after successively passing through the multiple layers of neural networks, the network structure is a self-attention mechanism-based transform structure, the number of layers is 12, the transform structure comprises position codes, word vector codes and sentence vector codes, the Bert learns basic semantics and a grammar structure like Chinese in a mode of pre-training mass knowledge texts in a bidirectional shielding mode, and a good effect can be achieved on text classification tasks only by fine adjustment, so that the method can be used for intelligent diagnosis tasks.
In performing pre-training of the language model: and taking the input fault text as source data, taking the corresponding multiple types of the fault text as a verification set, and training the BERT model by using two new unsupervised prediction tasks.
Wherein the fault text comprises standard fault phrases and similar fault phrases which are masked with a probability of 15%. The standard fault phrase is based on a standard fault instruction manual or knowledge, such as: the 'coolant liquid non-filling' is a standard fault expression phrase, and also comprises 'video sensing failure', 'tire pressure alarm' and the like, and text similar to 'cooling liquid' is covered, so that the fault reason and the solution are explicitly injected into the model.
For similar fault phrases, in normal fault diagnosis, an operator often cannot accurately input a standard fault phrase, but more nonstandard fault terms such as "coolant can not be filled", and the masking steps of the similar fault phrases are as follows:
acquiring input texts with similar fault meanings;
generating a phrase using Autophrase;
the phrases are enhanced and refined using rules and fastText.
During training, phrases finished in classification and words of the same unit can be fully masked, and an excellent fault diagnosis model is obtained by comprehensively collecting standard fault phrases and similar fault phrases, so that a better fault diagnosis result is obtained.
Specifically, in this implementation, the scheme of training the BERT model for the unsupervised prediction task and masking standard and similar fault phrases with a probability of 15% is as follows:
firstly, the training data generator randomly selects 15% tokens. For example, in the sentence "device display low voltage fault", the token it selects is "low voltage". Then, the following process is performed:
instead of always replacing the selected word with a [ mask ], the data generator will perform the following operations:
80% of the time: replacing words with [ masking ] marks, e.g. device shows Low Voltage Fault → My dog device shows [ masking ] Fault
10% of the time: replacing the word with a random word, e.g. device showing low voltage fault → device showing short circuit fault
10% of the time: keep the word unchanged, e.g., device shows low voltage fault → device shows low voltage fault. The purpose of this is to bias the representation towards the words that are actually observed.
Second, next sentence prediction, the previous masking language model is for word-level training, and many tasks in nlp are on sentence level, which requires the language model to understand the relations between sentences, which facilitates the downstream sentence relation determination task, while next sentence prediction in BERT is for this purpose.
The machine learning model in the embodiment adopts a random forest algorithm model, and the random forest is a relatively excellent model, has high efficiency for classifying the data sets of the multi-dimensional features, and can be used for selecting feature importance. The method has high operation efficiency and accuracy, is simple to realize, and is suitable for the information of less variables such as vehicle operation parameters.
Secondly, the method comprises the following steps: model fusion, as shown in fig. 2, includes the steps of:
s1, selecting mlp (multilayer perceptron) models to integrate the results of the language models and the machine learning models, wherein mlp network models of two layers of neural networks are selected in the implementation and are suitable for receiving output values of the language models and the machine learning models.
And S2, training by using the output values of the text model and the machine learning model as input features, and optimizing the integration effect by adjusting the model layer number, the model weight and the loss function option in the training process.
And S3, obtaining the integrated model with better effect on the verification set.
Specifically, the Mlp network: the artificial neural network consists of two layers of artificial neural networks, wherein the neural network mainly comprises three basic elements: weights, bias and activation functions, mlp the number of neurons per layer is manually specified. Wherein, the weight is: the connection strength between neurons is represented by a weight, the magnitude of the weight represents the magnitude of the probability, and the weight update uses a gradient descent method.
The model integration process is as follows:
a typical stacking model fusion method is selected, mlp networks of two layers of neural networks are selected, probability values output by a language model and a machine learning model are respectively used as features to be input into the mlp network, weight parameters of the two layers of neural networks are obtained by training a back propagation algorithm on a training set, and accordingly optimal weights of the two models are obtained, an integrated model comprises two sub models and weight value parameters of the two sub models on the two layers of neural networks, the accuracy on a verification set is optimal, the advantages of the two models are absorbed through the model integration method, and the prediction effect is optimal.
Thirdly, the method comprises the following steps: the model implementation steps include:
selecting mlp (multilayer perceptron) model integrated language model and machine learning model, training the model to have higher accuracy in actual expression, and when the integrated model is used for prediction: s31, splitting the collected input data into text data and numerical data, and respectively entering a text model and a machine learning model for reasoning to obtain an output value; and S32, inputting the output probability values of the two models as input features into the mlp model for reasoning to obtain a final probability value result.
The following explains and verifies the specific implementation of the present solution by specific examples:
firstly, dividing the plant intelligent diagnosis technical file into text features and non-text features according to text and non-text properties, wherein the text features are converted into word vectors and used as language model input, the non-text features are converted into discrete or continuous value feature input according to a machine learning model, and the two models are respectively trained to achieve the optimal effect on a plant intelligent diagnosis technical file training set, as shown in fig. 3: the fault in the text characteristic in the diagnosis technical document is described as that 'the cooling liquid cannot be filled, the equipment displays a low-pressure fault', and the non-text characteristic data comprises a voltage value of 187, a rotating speed of 65, an amplitude of 3.5, a temperature of 42 and the like.
And respectively bringing a plurality of groups of segmented training data into a Bert model and a random forest algorithm model for training to achieve the optimal training effect, and storing and outputting the trained optimal model.
As shown in fig. 4, by selecting the optimal language model and machine learning model, the original text feature and non-text feature data are converted into multiple sets of failure output values derived by model inference. Wherein the failure output value of the text feature transition comprises: filling fault 0.67, water leakage fault 0.12, driving fault 0.07, and displaying fault 0.04; the failure output values of the non-text feature data transition include: the filling fault is 0.35, the water leakage fault is 0.42, the driving fault is 0.17, and the fault is 0.06.
And utilizing the obtained multiple groups of output values, then starting a model integration step, selecting an mlp model containing two layers of neural networks as an integrated model, randomly initializing the network parameters of the integrated model, selecting the output values of the two submodels, namely the probability values of fault categories output by the two submodels as input characteristic values of the integrated model, using a back propagation method for the training method, enabling the randomly initialized network parameter values to be close to the optimal weight parameter values of the two models through multiple back propagation, finally obtaining the weight parameter values of the two submodels and the two submodels on the two layers of neural networks, and enabling the precision of the integrated model on a verification set to be optimal.
Finally, verification is carried out, as shown in fig. 5, the fault output value converted by the text characteristic and the fault output value converted by the non-text characteristic data in the model a and the model B are used as input characteristics and input into the mlp model for reasoning, and the probability value result is obtained as follows: the filling fault is 0.54, the water leakage fault is 0.27, the driving fault is 0.09, and the fault is 0.1.
This shows that when the voltage value is 187, the rotation speed is 65, the amplitude is 3.5, the temperature is 42, and the fault is described as "coolant cannot be filled, and the equipment displays a low-pressure fault", the probability of a filling fault is 54%, the probability of a water leakage fault is 27%, the probability of a driving fault is 9%, the probability of a display fault is 10%, it is found that the vehicle fault is most likely to be a filling fault, and the fault judgment accuracy is high in accordance with the actual fault cause.
The diagnosis method integrates the two models, improves the accuracy of fault diagnosis, can carry out inspection according to the output probability when the vehicle fault is inspected, and can carry out inspection and maintenance on the vehicle more quickly.
The method realizes model optimization of specific data processing by respectively using the language model to process the text characteristics and using the machine learning model to process the non-text characteristic data, and determines the final result by training the multi-level integrated model processing language model and the machine learning model to output, so that the result is more accurate.
It should be understood that, although the steps in the flowcharts of the embodiments of the present invention are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in various embodiments may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a non-volatile computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile 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), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. The intelligent vehicle fault diagnosis method combining the text analysis and the machine learning method is characterized by specifically comprising the following steps of:
step 1, training on text data and non-text characteristic data respectively by using a language model and a machine learning model to obtain two fault diagnosis models;
and 2, training the multi-stage integrated model to integrate two fault diagnosis model results to achieve the best fault diagnosis result.
2. The intelligent vehicle fault diagnosis method combining text analysis and machine learning method according to claim 1, wherein the intelligent vehicle fault diagnosis method combining text analysis and machine learning method further comprises establishing multi-modal model training: and respectively training the models of the plurality of modes by using the labeled data.
3. The intelligent vehicle fault diagnosis method combining text analysis and machine learning method according to claim 2, wherein the specific step of establishing multi-modal model training comprises:
and (3) multi-modal data processing: dividing training data into text data and non-text characteristic data, and selecting a corresponding language model and a machine learning model;
training a model: and respectively training a language model and a machine learning model on the text data and the non-text data to obtain a model optimized on the verification set.
4. The intelligent vehicle fault diagnosis method combining text analysis and machine learning method according to claim 1, characterized in that the language model adopts a BERT model, the input fault text is used as source data, the corresponding multiple types of the fault text are used as verification sets, and the BERT model is trained by using two new unsupervised prediction tasks.
5. An intelligent vehicle fault diagnosis method in combination with text analysis and machine learning methods according to claim 4, wherein the fault text comprises standard fault phrases and similar fault phrases, which are masked with a probability of 15%.
6. An intelligent vehicle fault diagnosis method in combination with text analysis and machine learning method according to claim 2, wherein the fault phrase-like masking step is: acquiring input texts with similar fault meanings; generating a phrase using Autophrase; phrases were generated using Autophrase.
7. The intelligent vehicle fault diagnosis method combining the text analysis and the machine learning method according to any one of claims 1 to 4, wherein the intelligent vehicle fault diagnosis method combining the text analysis and the machine learning method further specifically comprises model fusion.
8. The intelligent vehicle fault diagnosis method combining the text analysis and the machine learning method according to claim 5, wherein fusing the models specifically comprises:
s1, selecting mlp models to integrate language models and machine learning model results;
s2, training by using output values of the text model and the machine learning model as input features, and optimizing the integration effect by adjusting the number of model layers, the model weight and the loss function option in the training process;
and S3, obtaining the integrated model with better effect on the verification set.
9. The intelligent vehicle fault diagnosis method combining text analysis and machine learning methods according to any one of claims 1-4, wherein the intelligent vehicle fault diagnosis method combining text analysis and machine learning methods further comprises a model implementation step.
10. An intelligent vehicle fault diagnosis method in combination with text analysis and machine learning methods according to claim 7, wherein said model implementation step comprises: selecting mlp a model integration language model and a machine learning model; when the integrated model is used for prediction, firstly, input data is split into text data and numerical data, the text data and the numerical data enter a text model and a machine learning model respectively for reasoning to obtain output values, and then output probability values of the two models are input into an mlp model as input characteristics for reasoning to obtain a final probability value result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210262718.2A CN114611485A (en) | 2022-03-17 | 2022-03-17 | Intelligent vehicle fault diagnosis method combining text analysis and machine learning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210262718.2A CN114611485A (en) | 2022-03-17 | 2022-03-17 | Intelligent vehicle fault diagnosis method combining text analysis and machine learning method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114611485A true CN114611485A (en) | 2022-06-10 |
Family
ID=81863105
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210262718.2A Pending CN114611485A (en) | 2022-03-17 | 2022-03-17 | Intelligent vehicle fault diagnosis method combining text analysis and machine learning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114611485A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117714193A (en) * | 2023-12-28 | 2024-03-15 | 中国电子技术标准化研究院 | Diagnostic method, diagnostic device, electronic equipment and storage medium |
-
2022
- 2022-03-17 CN CN202210262718.2A patent/CN114611485A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117714193A (en) * | 2023-12-28 | 2024-03-15 | 中国电子技术标准化研究院 | Diagnostic method, diagnostic device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210174023A1 (en) | Systems and Methods for Explicit Memory Tracker with Coarse-To-Fine Reasoning in Conversational Machine Reading | |
| CN111026847B (en) | Text emotion recognition method based on attention network and long-short term memory network | |
| CN113688244A (en) | Text classification method, system, device and storage medium based on neural network | |
| CN110516065A (en) | A method of the attention neural network based on multichannel dynamic mask | |
| CN115599901B (en) | Machine question-answering method, device, equipment and storage medium based on semantic prompt | |
| CN113987179A (en) | Knowledge enhancement and backtracking loss-based conversational emotion recognition network model, construction method, electronic device and storage medium | |
| CN113254604B (en) | Reference specification-based professional text generation method and device | |
| CN112016300B (en) | Pre-training model processing method, pre-training model processing device, downstream task processing device and storage medium | |
| CN114611485A (en) | Intelligent vehicle fault diagnosis method combining text analysis and machine learning method | |
| CN115630156A (en) | Mongolian emotion analysis method and system fusing Prompt and SRU | |
| CN114490950A (en) | Training method and storage medium of encoder model, and similarity prediction method and system | |
| CN115687609A (en) | Zero sample relation extraction method based on Prompt multi-template fusion | |
| CN115221332A (en) | Construction method and system of dangerous chemical accident event map | |
| CN116432663B (en) | Controllable diversity professional text generation method and system based on element diagram | |
| CN111444328B (en) | Natural language automatic prediction and inference method with interpretation generation | |
| CN113065324A (en) | Text generation method and device based on structured triples and anchor templates | |
| Li et al. | 3M: Multi-style image caption generation using Multi-modality features under Multi-UPDOWN model | |
| CN108573275B (en) | Construction method of online classification micro-service | |
| CN116010563A (en) | Multi-round dialogue data analysis method, electronic equipment and storage medium | |
| CN115238903B (en) | Model compression method, system, electronic device and storage medium | |
| US20230368085A1 (en) | Automated intelligence facilitation of routing operations | |
| CN114926272B (en) | Behavior overdue prediction method, system, equipment and medium based on end-to-end model | |
| CN115470786A (en) | Entity information extraction method for electric power defect text based on improved Transformer encoder | |
| Thaw et al. | A Question and Answer System for Program Comprehension based on Encoder-Decoder Recurrent Neural Networks | |
| US20200125924A1 (en) | Method and system for analyzing a neural network |
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 |