Machine reading understanding method applied to traffic infrastructure detection text
By collecting and converting voice and text data in traffic infrastructure detection and using bidirectional neural network module for deep learning, the problem of different units and detection workers' semantic interpretations of word semantics is solved, and more accurate and comprehensive traffic infrastructure detection is achieved Text understanding and analysis.
Patent Information
- Application Number
- CN202411182259.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing machine reading comprehension model of the detection text of traffic infrastructure is not comprehensive enough when understanding and processing the detection text, and it is difficult to achieve accurate information extraction and analysis due to different semantic interpretations of the same word by different units and detection workers.
By collecting voice data and traffic sign text information in traffic infrastructure detection scenarios, converting it into text data, and using the bidirectional neural network module for deep learning, identifying and marking standard semantics, a more comprehensive understanding of traffic infrastructure detection text is achieved.
It improves the accuracy and reliability of text data information, can more comprehensively capture the context information and deep semantics in traffic infrastructure detection text, and enhances the performance and efficiency of machine reading comprehension.
Smart Images

Figure CN118916467A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of artificial intelligence technology, and specifically is a machine reading comprehension method applied to traffic infrastructure detection text. Background Art
[0002] With the acceleration of urbanization and the continuous improvement of transportation networks, the safety and reliability of transportation infrastructure are crucial to ensuring people's travel safety. Traditional transportation infrastructure detection methods often rely on manual inspections and regular maintenance, which is not only inefficient but also prone to misjudgment and missed detection. Therefore, artificial intelligence-based transportation infrastructure detection text reading comprehension technology has emerged, aiming to improve detection efficiency and accuracy through intelligent means.
[0003] The rapid development of artificial intelligence technology has provided new ideas and methods for traffic infrastructure detection. By introducing technologies such as deep learning and natural language processing, it is possible to realize the automatic understanding and processing of traffic infrastructure detection texts, thereby improving detection efficiency and accuracy.
[0004] Traffic infrastructure inspection requires processing a large amount of text information, including inspection reports, maintenance records, etc. These text information contain a lot of key information, such as fault type, location, time, etc. Traditional manual processing methods are inefficient and cannot meet actual needs. Therefore, an efficient and accurate automated processing method is needed to meet this challenge.
[0005] Machine reading comprehension technology is a technology that can automatically extract key information from text. By training the model, it can understand the meaning of natural language text and accurately answer questions related to the text. This technology has broad application prospects in the field of transportation infrastructure inspection and can realize the automatic understanding and processing of inspection text.
[0006] The text reading comprehension technology for traffic infrastructure detection based on artificial intelligence has broad application prospects. It can be applied to different types of traffic infrastructure detection such as bridges, tunnels, and roads to improve the efficiency and accuracy of detection. At the same time, this technology can also be combined with other traffic management systems to achieve more intelligent traffic management and services. With the continuous development of technology and the expansion of application scenarios, the potential of this technology will be more fully utilized.
[0007] The existing machine reading comprehension of transportation infrastructure inspection texts comprehensively uses deep learning and NLP technologies to identify and understand key information in the inspection report text by training and optimizing the model, thereby realizing automatic detection and analysis of the status of transportation infrastructure. It usually requires learning and retrieving a large amount of inspection report texts, but at the same time, since the semantics of the inspection report texts are relatively standard and single, different inspection workers from different units have different semantic interpretations of the same word in the inspection results, making it difficult to understand different statements of the same word, making the reading comprehension model not comprehensive enough. Therefore, it is necessary to propose a machine reading comprehension method for transportation infrastructure inspection texts. Summary of the invention
[0008] In order to solve the problem that the machine reading comprehension of the above-mentioned common traffic infrastructure detection text is difficult to understand different ways of saying the same word due to different inspection workers in different units in the inspection results, making the reading comprehension model not comprehensive enough, the purpose of the present invention is to provide a machine reading comprehension method for traffic infrastructure detection text, which collects voice data information of conversations between staff members who inspect traffic infrastructure in different traffic infrastructure inspection scenarios through a collection module, extracts keywords and key information therein, and converts traffic sign text data information in the collection scenario into text data information to train the bidirectional neural network module to learn and train, and continuously trains to identify different semantics in real scenarios and finds the corresponding standard semantics through self-learning, so as to achieve a more comprehensive understanding of the traffic infrastructure detection text reading by the bidirectional neural network module in combination with real scenarios.
[0009] In order to achieve the above object, the technical solution of the present invention is as follows: A machine reading comprehension method for traffic infrastructure detection text comprises the following steps:
[0010] Step 1: Take a photo of the traffic infrastructure detection scene through the collection module to collect image data information, and collect voice data information of the conversation of the staff who detect the traffic infrastructure in the traffic infrastructure detection scene;
[0011] Step 2: converting the voice data information into text data information through a non-text preprocessing module, and converting the text information on the image data information into text data information;
[0012] Step 3: Using a semantic analysis module, the text data information obtained by the non-text preprocessing module is compared and identified with the standard traffic infrastructure detection text, and key information is extracted to obtain key text data information;
[0013] Step 4: preprocessing the key text data extracted by the semantic analysis module and the text data collected by the collection module through the preprocessing module;
[0014] Step 5, converting the result processed by the preprocessing module into a numerical vector through the text vectorization module;
[0015] Step six: Convert the numerical vector obtained by the text vectorization module into key text data information through the bidirectional neural network module, and compare the key text data information with the standard traffic infrastructure detection text to conduct deep learning training. The learning results are used for machine reading comprehension of the traffic infrastructure detection text.
[0016] Furthermore, in step one, the collection module includes a camera and a microphone, the camera is used to collect image data information and text data information in the scene, and the microphone is used to collect voice data information.
[0017] Furthermore, in step one, the transportation infrastructure scene includes one or more of a subway station, a bridge, a highway toll station, a railway station and a tunnel.
[0018] Further, in step 2, the text information includes text data information on the traffic sign.
[0019] Furthermore, in step two, the non-text preprocessing module includes an RNN convolutional neural network model unit, and the RNN convolutional neural network model unit is used to convert the non-text data information collected by the collection unit into text data information through its own learning model.
[0020] Furthermore, in step 4, the preprocessing steps include word segmentation, stop word removal, and part-of-speech restoration.
[0021] Furthermore, in step five, the text vectorization module takes the preprocessed numerical vector as input and sequentially sends it to the bidirectional neural network module according to the time step.
[0022] Furthermore, in step six, the semantic analysis module and the bidirectional neural network module are linked to each other, and the semantic analysis module uses the learning results of the bidirectional neural network module to identify and extract key text data information.
[0023] Furthermore, in step six, the bidirectional neural network module initializes the parameters of the model when the module is used for the first time, and the parameters include input weights, hidden state weights and bias terms.
[0024] Furthermore, in step six, the bidirectional neural network module runs using two parallel RNN layers in chronological order, the first RNN layer processes data from front to back, and the second RNN layer processes data from back to front, capturing the dependencies in the text data information.
[0025] Beneficial effects: 1. The principle of learning the bidirectional neural network module is to collect the conversational voice information of the workers in the scene during the inspection, maintenance and repair of traffic facilities, as well as the text data information on the traffic signs in the scene, and use this information to find the corresponding standard semantics in the traffic infrastructure inspection text and mark them. By continuously training to identify different semantics in real scenes and finding the corresponding standard semantics through self-learning, the bidirectional neural network module can achieve a more comprehensive understanding of the traffic infrastructure inspection text reading in combination with real scenes.
[0026] 2. The semantic analysis module improves the accuracy and reliability of text data information by utilizing the learning results of the bidirectional neural network module. At the same time, this module is linked to the bidirectional neural network module to form a cyclic iterative process, which continuously optimizes the performance of machine reading comprehension.
[0027] 3. This method can be applied to various transportation infrastructure inspection scenarios, including subway stations, bridges, highway toll stations, railway stations and tunnels. Through machine reading comprehension technology, key information in the inspection text can be automatically extracted and analyzed, providing effective technical support for the inspection and maintenance of transportation infrastructure.
[0028] 4. The bidirectional neural network module uses two parallel RNN layers to run in chronological order. The first RNN layer processes data from front to back, and the second RNN layer processes data from back to front, capturing the dependencies in the text data information. This enables the bidirectional neural network model to capture higher-level contextual semantic information. For understanding complex traffic infrastructure detection texts, the bidirectional neural network model can comprehensively and accurately capture the contextual information and deep semantic information in the traffic infrastructure detection text, thereby more accurately understanding the meaning of the text. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 The present invention is a flowchart of a machine reading comprehension method for traffic infrastructure detection text in an embodiment of the present invention. DETAILED DESCRIPTION
[0030] The following is further described in detail through specific implementation methods:
[0031] The embodiment is basically as shown in the attached Figure 1 As shown: A machine reading comprehension method for traffic infrastructure detection text includes the following steps:
[0032] Step one, take pictures of the traffic infrastructure detection scene through the camera on the collection module to collect image data information, and use the microphone on the collection module to collect voice data information of the conversation between the staff who detect the traffic infrastructure in the traffic infrastructure detection scene, such as the voice information of the workers' conversations during the detection, maintenance and repair of traffic facilities, etc., as well as text data information on the traffic signs in the collection scene. The collection module sends the text data information, voice data information and image data information to the text preprocessing module.
[0033] Among them, the transportation infrastructure scenarios include one or more of subway stations, bridges, highway toll stations, railway stations and tunnels.
[0034] Step 2: Convert the voice data information into text data information through a non-text preprocessing module, and convert the text information on the image data information into text data information.
[0035] Among them, the text information includes text data information on traffic signs, and the non-text preprocessing module includes an RNN convolutional neural network model unit, which is used to convert the voice data information collected by the collection unit into text data information through its own learning model.
[0036] Step three, the text data information obtained by the non-text preprocessing module is compared and identified with the standard traffic infrastructure detection text through the semantic analysis module, and the key information is extracted to obtain the key text data information.
[0037] Step 4: Preprocess the key text data information extracted by the semantic analysis module and the text data information collected by the collection module through a preprocessing module.
[0038] Among them, the preprocessing steps include word segmentation, stop word removal and part-of-speech restoration.
[0039] Step 5: Convert the results processed by the preprocessing module into a numerical vector through the text vectorization module.
[0040] Among them, the text vectorization module takes the preprocessed numerical vector as input and sends it to the bidirectional neural network module in sequence according to the time step.
[0041] Step six: Convert the numerical vector obtained by the text vectorization module into key text data information through the bidirectional neural network module, and compare the key text data information with the standard traffic infrastructure detection text to conduct deep learning training. The learning results are used for machine reading comprehension of the traffic infrastructure detection text.
[0042] Among them, the semantic analysis module and the bidirectional neural network module are linked to each other. The semantic analysis module uses the learning results of the bidirectional neural network module to identify and extract key text data information. The bidirectional neural network module initializes the parameters of the model when the module is used for the first time. The parameters include input weights, hidden state weights and bias items. The bidirectional neural network module runs in chronological order using two parallel RNN layers. The first RNN layer processes data from front to back, and the second RNN layer processes data from back to front to capture the dependencies in the text data information.
[0043] The principle of learning the bidirectional neural network module is to collect the conversational voice information of the workers in the scene when they perform inspection, maintenance and repair of traffic facilities, as well as the text data information on the traffic signs in the scene, and use this information to find the corresponding standard semantics in the traffic infrastructure inspection text and mark them. By continuously training to identify different semantics in real scenes and finding the corresponding standard semantics through self-learning, deep recognition and learning of semantics in different scenes can be achieved, so that the bidirectional neural network module can combine real scenes to more comprehensively understand the reading of traffic infrastructure inspection text.
[0044] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0045] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.
Claims
1. A machine reading comprehension method for traffic infrastructure inspection text, characterized by: The following steps are involved: Step 1: Take a photo of the traffic infrastructure detection scene through the collection module to collect image data information, and collect voice data information of the conversation of the staff who detect the traffic infrastructure in the traffic infrastructure detection scene; Step 2: converting the voice data information into text data information through a non-text preprocessing module, and converting the text information on the image data information into text data information; Step 3: Using a semantic analysis module, the text data information obtained by the non-text preprocessing module is compared and identified with the standard traffic infrastructure detection text, and key information is extracted to obtain key text data information; Step 4: preprocessing the key text data information extracted by the semantic analysis module and the text data information collected by the collection module through a preprocessing module; Step 5, converting the result processed by the preprocessing module into a numerical vector through the text vectorization module; Step six: Convert the numerical vector obtained by the text vectorization module into key text data information through the bidirectional neural network module, and compare the key text data information with the standard traffic infrastructure detection text to conduct deep learning training. The learning results are used for machine reading comprehension of the traffic infrastructure detection text.
2. The machine reading comprehension method for traffic infrastructure detection text according to claim 1, characterized in that: In step one, the collection module includes a camera and a microphone. The camera is used to collect image data information and text data information in the scene, and the microphone is used to collect voice data information.
3. The machine reading comprehension method for traffic infrastructure detection text according to claim 2, characterized in that: In step 1, the transportation infrastructure scene includes one or more of a subway station, a bridge, a highway toll station, a railway station, and a tunnel.
4. The machine reading comprehension method for traffic infrastructure detection text according to claim 3 is characterized by: In step 2, the text information includes text data information on the traffic sign.
5. The machine reading comprehension method for traffic infrastructure detection text according to claim 4, characterized in that: In step 2, the non-text preprocessing module includes an RNN convolutional neural network model unit, which is used to convert the language data information collected by the collection unit into text data information through its own learning model and extract key text data information.
6. The machine reading comprehension method for traffic infrastructure detection text according to claim 5, characterized in that: In step 4, the preprocessing steps include word segmentation, stop word removal, and part-of-speech restoration.
7. The machine reading comprehension method for traffic infrastructure detection text according to claim 6, characterized in that: In step 5, the text vectorization module takes the preprocessed numerical vector as input and sends it to the bidirectional neural network module in sequence according to the time step.
8. The machine reading comprehension method for traffic infrastructure detection text according to claim 7, characterized in that: In step six, the semantic analysis module and the bidirectional neural network module are linked to each other, and the semantic analysis module uses the learning results of the bidirectional neural network module to identify and extract key text data information.
9. The machine reading comprehension method for traffic infrastructure detection text according to claim 8, characterized in that: In step six, when the bidirectional neural network module is used for the first time, the parameters of the model are initialized, including input weights, hidden state weights, and bias terms.
10. The machine reading comprehension method for traffic infrastructure detection text according to claim 9, characterized in that: In step six, the bidirectional neural network module runs in chronological order using two parallel RNN layers. The first RNN layer processes data from front to back, and the second RNN layer processes data from back to front, capturing the dependencies in the text data information.
