CN114328886A - Intelligent logistics address entity recognition system based on deep learning - Google Patents
Intelligent logistics address entity recognition system based on deep learningDownload PDFInfo
- Publication number
- CN114328886A CN114328886ACN202111523965.5ACN202111523965ACN114328886ACN 114328886 ACN114328886 ACN 114328886ACN 202111523965 ACN202111523965 ACN 202111523965ACN 114328886 ACN114328886 ACN 114328886A
- Authority
- CN
- China
- Prior art keywords
- address
- data
- entity
- model
- deep learning
- 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
Landscapes
- Sorting Of Articles (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Character Discrimination (AREA)
Abstract
Translated fromChinese
Description
Translated fromChinese技术领域technical field
本发明涉及物流技术领域，特别是一种基于深度学习的智能物流地址实体提取识别系统及物流地址实体的提取识别方法。The invention relates to the technical field of logistics, in particular to a deep learning-based intelligent logistics address entity extraction and identification system and a logistics address entity extraction and identification method.
背景技术Background technique
当前智能化越来越成为时代的关键词互联网、物联网、大数据、人工智能等技术不断发展给多个行业带来创新的推动力，所谓智能物流是指通过先进的物流网技术实现物资运输过程的自动化运作和高效化管理.物流行业的智能化对于中国物流行业提高利润、降低物流成本具有积极的推动作用。At present, intelligence is increasingly becoming the keyword of the times. The continuous development of technologies such as the Internet, Internet of Things, big data, and artificial intelligence has brought innovation to many industries. The so-called intelligent logistics refers to the realization of material transportation through advanced logistics network technology. Process automation and efficient management. The intelligence of the logistics industry has a positive effect on improving profits and reducing logistics costs in China's logistics industry.
而作为智能物流的核心组成部分，“三段码”由三段编码构成：一段码(转运中心)+二段码(独立网点)+三段码(派件员)+四段码(末端实体)。一二三段码是通过对转运中心、独立网点和派件员进行编码。上述四段码则是根据地址数据对每个网点下派送的末端实体进行识别提取，从而提升物流的分拣效率，节约人员成本。在现有技术中，尚没有对四段码进行应用和提取的案例，即使有也是人工实现，存在着识别困难和识别准确率过低的问题。本发明拟就末端实体的应用和提取进行探索，以促进物流业的快速发展。As the core component of intelligent logistics, "three-segment code" consists of three-segment codes: one-segment code (transshipment center) + two-segment code (independent network point) + three-segment code (dispatcher) + four-segment code (end entity ). One, two and three-segment codes are used to encode transit centers, independent outlets and dispatchers. The above four-segment codes are used to identify and extract the end entities dispatched under each outlet based on the address data, thereby improving the sorting efficiency of logistics and saving personnel costs. In the prior art, there is no case of applying and extracting the four-segment code, and even if there is, it is implemented manually, and there are problems of difficulty in recognition and low recognition accuracy. The present invention intends to explore the application and extraction of end entities to promote the rapid development of the logistics industry.
发明内容SUMMARY OF THE INVENTION
本申请将利用深度学习技术解决目前智能物流领域物流地址实体识别困难、识别准确率过低的问题。提供一种基于深度学习的智能物流地址实体提取识别系统及物流地址实体的提取识别方法。This application will use the deep learning technology to solve the current problems of difficult identification of logistics address entities and low identification accuracy in the field of intelligent logistics. Provided are a deep learning-based intelligent logistics address entity extraction and identification system and a logistics address entity extraction and identification method.
为了达到上述发明目的，本发明专利提供的技术方案如下：In order to achieve the above-mentioned purpose of the invention, the technical scheme provided by the patent of the present invention is as follows:
一种基于深度学习的智能物流地址实体识别系统，该系统组成包括有地址实体数据标注模块、BERT编码器模块和结果解析模块，其中，An intelligent logistics address entity recognition system based on deep learning, the system comprises an address entity data labeling module, a BERT encoder module and a result parsing module, wherein,
所述地址实体数据标注模块接收地址结构化数据，并对接收的地址结构化数据进行数据清洗，获得高质量的已标注地址实体数据；The address entity data labeling module receives the address structured data, and performs data cleaning on the received address structured data to obtain high-quality labelled address entity data;
所述的BERT编码器模块基于已标注地址实体数据，进行深度学习算法学习物流业务实际应用场景的特定句子的编码表示，训练完成获得地址实体识别模型；The BERT encoder module, based on the marked address entity data, performs a deep learning algorithm to learn the coding representation of a specific sentence in the actual application scenario of the logistics business, and completes the training to obtain an address entity recognition model;
所述的结构解析模块基于训练完成的地址实体识别模块，在输入物流地址时，对输入地址信息数据进行解析，解析完成后输出由所输入地址提取的实体内容。The structure analysis module is based on the trained address entity recognition module, and when the logistics address is input, it analyzes the input address information data, and after the analysis is completed, the entity content extracted from the input address is output.
在本发明的一种基于深度学习的智能物流实体识别系统中，在所述的BERT编码器模块中设有BERT编码器、全连接层、softmax分类器和分类优化器，所述BERT编码器处理输入的地址实体数据得到数据的向量表示，全连接层对数据向量表示增加权重，softmax分类器得到预测结果，分类优化器逐层调整模型权重实现模型迭代优化。In a deep learning-based intelligent logistics entity recognition system of the present invention, the BERT encoder module is provided with a BERT encoder, a fully connected layer, a softmax classifier and a classification optimizer, and the BERT encoder processes The input address entity data obtains the vector representation of the data, the fully connected layer adds weights to the data vector representation, the softmax classifier obtains the prediction result, and the classification optimizer adjusts the model weight layer by layer to achieve model iterative optimization.
本发明还提供一种基于深度学习的智能物流地址实体的识别提取方法，该识别提取方法包括如下步骤：The present invention also provides a deep learning-based method for identifying and extracting intelligent logistics address entities, and the method for identifying and extracting includes the following steps:
第一步，数据预处理，将历史运单数据筛选，删除重复数据，获得干净的地址实体集；The first step, data preprocessing, filter the historical waybill data, delete duplicate data, and obtain a clean address entity set;
第二步，地址实体集清洗，并对清洗后数据进行标注提取，经筛选过滤后提取有意义的实体数据，获得模型训练所用的数据；The second step is to clean the address entity set, label and extract the cleaned data, extract meaningful entity data after screening and filtering, and obtain the data used for model training;
第三步，BERT编码及模型训练，由BERT编码得到地址数据的向量表示，经模型训练和优化得到物流地址实体识别模型；The third step, BERT coding and model training, the vector representation of address data is obtained from BERT coding, and the logistics address entity recognition model is obtained after model training and optimization;
第四步，输入地址及结果解析，将物流地址输入至训练得到的物流地址实体识别模型，经过解析输出获得物流地址中的实体内容。The fourth step: input address and result analysis, input the logistics address into the trained logistics address entity recognition model, and obtain the entity content in the logistics address through the analysis output.
在本发明一种基于深度学习的智能物流地址实体的识别提取方法中，第一步将历史运单数据进行筛选过滤，对省市区+详细地址拼接的数据进行省市区递归删除，删除省市区部分重复的数据，对进行了递归删除的地址进行MD5值计算，删除MD5值一致的数据，该MD5值一致表示为全部重复的数据。In the method for identifying and extracting intelligent logistics address entities based on deep learning of the present invention, the first step is to filter historical waybill data, perform recursive deletion of provinces and cities + detailed address splicing data, and delete provinces and cities. For the partially duplicated data in the area, the MD5 value is calculated for the address that has been recursively deleted, and the data with the same MD5 value is deleted, and the consistent MD5 value is expressed as all the duplicated data.
在本发明一种基于深度学习的智能物流地址实体的识别提取方法中，第二步使用词频/逆文本频率策略对地址提取出的实体进行筛选过滤，只提取具有具体意义的主要实体的数据，得到模型训练所用的标注地址数据。In a method for identifying and extracting intelligent logistics address entities based on deep learning of the present invention, in the second step, the word frequency/inverse text frequency strategy is used to screen and filter the entities extracted from the addresses, and only the data of the main entities with specific meanings are extracted, Get the labeled address data used for model training.
在本发明一种基于深度学习的智能物流地址实体的识别提取方法中，第三步使用BERT模型对处理好的地址进行命名实体识别任务，得到地址数据的向量表示，在命名实体识别任务中，对训练结果迭代优化，数据表示向量分别经过全连接层和Softmax分类层输出预测结果；预测结果与数据标签计算分类损失；模型优化器通过将损失逐层回传并通过优化算法来迭代优化BERT模型的权重。In a method for identifying and extracting address entities of intelligent logistics based on deep learning of the present invention, in the third step, the BERT model is used to perform a named entity identification task on the processed addresses, and a vector representation of the address data is obtained. In the named entity identification task, The training results are iteratively optimized, and the data representation vector outputs the prediction results through the fully connected layer and the Softmax classification layer respectively; the prediction results and the data labels calculate the classification loss; the model optimizer iteratively optimizes the BERT model by returning the loss layer by layer and using the optimization algorithm. the weight of.
在本发明一种基于深度学习的智能物流地址实体的识别提取方法中，在第三步的基础上，BERT编码和分类优化器的实现过程为：In a deep learning-based method for identifying and extracting intelligent logistics address entities of the present invention, on the basis of the third step, the implementation process of the BERT encoding and classification optimizer is as follows:
S1.输入地址实体数据通过BERT编码器得到数据的向量表示C：S1. Enter the address entity data to obtain the vector representation C of the data through the BERT encoder:
C＝EncoderTransformer(x1,x2,x3,...,xM) (1)C=EncoderTransformer(x1 ,x2 ,x3 ,...,xM ) (1)
S2.C经过全连接层加一层权重后，进到Softmax分类器得到预测结果：After S2.C goes through the fully connected layer and adds a layer of weight, it enters the Softmax classifier to get the prediction result:
pred＝softmax(CWT) (2)pred=softmax(CWT ) (2)
S3.根据预测值pred和真实标签label计算分类损失：S3. Calculate the classification loss based on the predicted value pred and the true label label:
S4.通过优化器来逐层调整模型权重实现模型迭代优化，最终得到收敛的地址实体识别模型。S4. The model weight is adjusted layer by layer through the optimizer to realize the iterative optimization of the model, and finally a converged address entity recognition model is obtained.
在本发明一种基于深度学习的智能物流地址实体的识别提取方法中，还包括有第五步，将训练完成的实体识别模型部署为Triton Inference服务，进行物流地址实体识别模型计算解析加速，承担大批量地址实体数据处理。In the method for identifying and extracting intelligent logistics address entities based on deep learning of the present invention, the fifth step is to deploy the trained entity identification model as a Triton Inference service, to accelerate the calculation and analysis of the logistics address entity identification model, and to undertake Bulk address entity data processing.
基于上述技术方案，本发明一种基于深度学习的智能物流地址实体提取识别系统及物流地址实体的提取识别方法与现有技术相比，取得了如下技术效果：Based on the above technical solutions, compared with the prior art, a deep learning-based intelligent logistics address entity extraction and identification system and a logistics address entity extraction and identification method of the present invention have achieved the following technical effects:
1.本发明基于深度学习的智能物流地址实体识别系统中，在物流地址数据获得时使用了第三方的地址结构化数据，无需采用人工再进行标注，既节省人工标注成本又能够获取一批比较高质量的地址实体标注数据。1. In the intelligent logistics address entity recognition system based on deep learning of the present invention, the third-party address structured data is used when the logistics address data is obtained, and manual labeling is not required, which not only saves the cost of manual labeling, but also enables a batch of comparisons to be obtained. High-quality address entity annotation data.
2.本发明基于深度学习的智能物流地址实体识别系统中，用TF-IDF(词频/逆文本频率)的策略对地址实体进行筛选过滤，筛选出高质量的地址实体，提取具有具体意义的主要实体数据，为了模型训练提供了良好的数据源基础。2. In the intelligent logistics address entity recognition system based on deep learning of the present invention, the strategy of TF-IDF (word frequency/inverse text frequency) is used to screen and filter the address entities, screen out high-quality address entities, and extract the main objects with specific meanings. Entity data provides a good data source basis for model training.
3.本发明基于深度学习的智能物流地址实体识别系统中，利用提取的出地址实体信息可以在三段码的基础上生成四段码，即一段码(转运中心)+二段码(独立网点)+三段码(派件员)+四段码(末端实体)，有利于节省物流的分拣人工成本，从而提高物流企业的核心竞争力。3. In the intelligent logistics address entity recognition system based on deep learning of the present invention, the extracted address entity information can be used to generate four-segment codes on the basis of three-segment codes, that is, one-segment code (transit center) + two-segment code (independent network point). ) + three-segment code (dispatcher) + four-segment code (end entity), which is conducive to saving the labor cost of logistics sorting, thereby improving the core competitiveness of logistics enterprises.
附图说明Description of drawings
图1是本发明基于深度学习的智能物流地址实体识别系统框图。Fig. 1 is a block diagram of the intelligent logistics address entity recognition system based on deep learning of the present invention.
具体实施方式Detailed ways
下面我们结合附图和具体的实施例来对本发明一种基于深度学习的智能物流地址实体提取识别系统及物流地址实体的提取识别方法做进一步的详细说明，以求更为清楚明了地理解本发明的组成和工作过程，但不以此来限制本发明的保护范围。In the following, we will further describe in detail a deep learning-based intelligent logistics address entity extraction and identification system and a logistics address entity extraction and identification method of the present invention in conjunction with the accompanying drawings and specific embodiments, in order to understand the present invention more clearly and clearly. The composition and working process of the invention are not intended to limit the protection scope of the present invention.
如图1所示，本发明涉及到一种基于深度学习的智能物流地址实体识别系统，该系统组成包括有地址实体数据标注模块、BERT编码器模块和结果解析模块，其中，As shown in Figure 1, the present invention relates to a deep learning-based intelligent logistics address entity recognition system, which comprises an address entity data labeling module, a BERT encoder module and a result parsing module, wherein,
所述地址实体数据标注模块接收地址结构化数据，并对接收的地址结构化数据进行数据清洗，获得高质量的已标注地址实体数据；The address entity data labeling module receives the address structured data, and performs data cleaning on the received address structured data to obtain high-quality labelled address entity data;
所述的BERT编码器模块基于已标注地址实体数据，进行深度学习算法学习物流业务实际应用场景的特定句子的编码表示，训练完成获得地址实体识别模型；The BERT encoder module, based on the marked address entity data, performs a deep learning algorithm to learn the coding representation of a specific sentence in the actual application scenario of the logistics business, and completes the training to obtain an address entity recognition model;
所述的结构解析模块基于训练完成的地址实体识别模块，在输入物流地址时，对输入地址信息数据进行解析，解析完成后输出由所输入地址提取的实体内容。The structure analysis module is based on the trained address entity recognition module, and when the logistics address is input, it analyzes the input address information data, and after the analysis is completed, the entity content extracted from the input address is output.
上述基于深度学习的智能物流地址实体识别系统地址标注处理模块在第三方公司的地址结构化解析服务的基础上根据TF-IDF以及一些统计学上的方式进行数据清洗。既节省人工标注成本又能够获取一批比较高质量的地址实体标注数据；BERT编码器模块是基于已完成标注的地址实体数据进行深度学习算法学习物流业务实际应用场景下的特定句子的编码表示。The above-mentioned address labeling processing module of the intelligent logistics address entity recognition system based on deep learning performs data cleaning according to TF-IDF and some statistical methods on the basis of the address structure analysis service of a third-party company. It not only saves the cost of manual labeling, but also can obtain a batch of high-quality address entity labeling data; the BERT encoder module is based on the labelled address entity data to perform deep learning algorithms to learn the coding representation of specific sentences in practical application scenarios of logistics business.
在本发明的一种基于深度学习的智能物流实体识别系统中，在所述的BERT编码器模块中设有BERT编码器、全连接层、softmax分类器和分类优化器，所述BERT编码器处理输入的地址实体数据得到数据的向量表示，全连接层对数据向量表示增加权重，softmax分类器得到预测结果，分类优化器逐层调整模型权重实现模型迭代优化。In a deep learning-based intelligent logistics entity recognition system of the present invention, the BERT encoder module is provided with a BERT encoder, a fully connected layer, a softmax classifier and a classification optimizer, and the BERT encoder processes The input address entity data obtains the vector representation of the data, the fully connected layer adds weights to the data vector representation, the softmax classifier obtains the prediction result, and the classification optimizer adjusts the model weight layer by layer to achieve model iterative optimization.
本发明还提供一种基于深度学习的智能物流地址实体的识别提取方法，该识别提取方法包括如下步骤：The present invention also provides a deep learning-based method for identifying and extracting intelligent logistics address entities, and the method for identifying and extracting includes the following steps:
第一步，数据预处理，将历史运单数据筛选，删除重复数据，获得干净的地址实体集；The first step, data preprocessing, filter the historical waybill data, delete duplicate data, and obtain a clean address entity set;
第二步，地址实体集清洗，并对清洗后数据进行标注提取，经筛选过滤后提取有意义的实体数据，获得模型训练所用的数据；The second step is to clean the address entity set, label and extract the cleaned data, extract meaningful entity data after screening and filtering, and obtain the data used for model training;
第三步，BERT编码及模型训练，由BERT编码得到地址数据的向量表示，经模型训练和优化得到物流地址实体识别模型；The third step, BERT coding and model training, the vector representation of address data is obtained from BERT coding, and the logistics address entity recognition model is obtained after model training and optimization;
第四步，输入地址及结果解析，将物流地址输入至训练得到的物流地址实体识别模型，经过解析输出获得物流地址中的实体内容。The fourth step: input address and result analysis, input the logistics address into the trained logistics address entity recognition model, and obtain the entity content in the logistics address through the analysis output.
上述第一步中，将历史运单数据进行筛选过滤，对省市区+详细地址拼接的数据进行省市区递归删除，删除省市区部分重复的数据，对进行了递归删除的地址进行MD5值计算，删除MD5值一致的数据，该MD5值一致表示为全部重复的数据。In the first step above, the historical waybill data is filtered, recursively deletes the data of the provinces and cities + detailed addresses, deletes the repeated data of the provinces and cities, and performs the MD5 value of the recursively deleted addresses. Calculate and delete the data with the same MD5 value, and the consistent MD5 value is expressed as all duplicate data.
上述第二步中，使用词频/逆文本频率策略对地址提取出的实体进行筛选过滤，只提取具有具体意义的主要实体的数据，得到模型训练所用的标注地址数据。In the second step above, the word frequency/inverse text frequency strategy is used to filter the entities extracted from the addresses, and only the data of the main entities with specific meanings are extracted to obtain the marked address data used for model training.
上述第三步中，使用BERT模型对处理好的地址进行命名实体识别任务，得到地址数据的向量表示，在命名实体识别任务中，对训练结果迭代优化，数据表示向量分别经过全连接层和Softmax分类层输出预测结果；预测结果与数据标签计算分类损失；模型优化器通过将损失逐层回传并通过优化算法来迭代优化BERT模型的权重。In the third step above, the BERT model is used to perform the named entity recognition task on the processed address, and the vector representation of the address data is obtained. In the named entity recognition task, the training result is iteratively optimized, and the data representation vector is passed through the fully connected layer and Softmax respectively. The classification layer outputs the prediction result; the prediction result and the data label calculate the classification loss; the model optimizer iteratively optimizes the weight of the BERT model by passing the loss back layer by layer and through the optimization algorithm.
在上述第三步的基础上，BERT编码和分类优化器的实现过程为：On the basis of the third step above, the implementation process of the BERT encoding and classification optimizer is as follows:
S1.输入地址实体数据通过BERT编码器得到数据的向量表示C：S1. Enter the address entity data to obtain the vector representation C of the data through the BERT encoder:
C＝EncoderTransformer(x1,x2,x3,...,xM) (1)C=EncoderTransformer(x1 ,x2 ,x3 ,...,xM ) (1)
S2.C经过全连接层加一层权重后，进到Softmax分类器得到预测结果：After S2.C goes through the fully connected layer and adds a layer of weight, it enters the Softmax classifier to get the prediction result:
pred＝softmax(CWT) (2)pred=softmax(CWT ) (2)
S3.根据预测值pred和真实标签label计算分类损失：S3. Calculate the classification loss based on the predicted value pred and the true label label:
S4.通过优化器来逐层调整模型权重实现模型迭代优化，最终得到收敛的地址实体识别模型。S4. The model weight is adjusted layer by layer through the optimizer to realize the iterative optimization of the model, and finally a converged address entity recognition model is obtained.
在本发明一种基于深度学习的智能物流地址实体的识别提取方法中，还包括有第五步，将训练完成的实体识别模型部署为Triton Inference服务，进行物流地址实体识别模型计算解析加速，承担大批量地址实体数据处理。In the method for identifying and extracting intelligent logistics address entities based on deep learning of the present invention, the fifth step is to deploy the trained entity identification model as a Triton Inference service, to accelerate the calculation and analysis of the logistics address entity identification model, and to undertake Bulk address entity data processing.
本发明的系统和方法中，采用深度学习与地址实体提取相结合的技术实现了地址实体提取的自动学习，在第三方公司例如阿里、美团等提供的地址结构化解析服务的基础上获得标注的地址实体信息数据，根据TF-IDF以及一些统计学上的方式进行数据清洗，这种操作既节省人工标注成本又能够获取一批比较高质量的地址实体标注数据。再根据地址数据对每个网点下派送的末端实体进行识别提取，从而提升物流的分拣效率，节约人员成本。In the system and method of the present invention, the technology of combining deep learning and address entity extraction is used to realize the automatic learning of address entity extraction, and tags are obtained on the basis of address structured analysis services provided by third-party companies such as Ali, Meituan, etc. According to TF-IDF and some statistical methods to clean the address entity information data, this operation not only saves the cost of manual labeling, but also can obtain a batch of relatively high-quality address entity labeling data. Then, according to the address data, the terminal entities dispatched under each outlet are identified and extracted, thereby improving the sorting efficiency of logistics and saving personnel costs.
Claims (8)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111523965.5ACN114328886A (en) | 2021-12-14 | 2021-12-14 | Intelligent logistics address entity recognition system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111523965.5ACN114328886A (en) | 2021-12-14 | 2021-12-14 | Intelligent logistics address entity recognition system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114328886Atrue CN114328886A (en) | 2022-04-12 |
Family
ID=81051614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111523965.5APendingCN114328886A (en) | 2021-12-14 | 2021-12-14 | Intelligent logistics address entity recognition system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114328886A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115859986A (en)* | 2022-12-27 | 2023-03-28 | 上海捷晓信息技术有限公司 | Method, system and computer medium for realizing address named entity recognition based on tensorrT accelerated reasoning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569322A (en)* | 2019-07-26 | 2019-12-13 | 苏宁云计算有限公司 | Address information analysis method, device and system and data acquisition method |
| CN110688449A (en)* | 2019-09-20 | 2020-01-14 | 京东数字科技控股有限公司 | Address text processing method, device, equipment and medium based on deep learning |
| CN111209362A (en)* | 2020-01-07 | 2020-05-29 | 苏州城方信息技术有限公司 | Address data analysis method based on deep learning |
| CN111522901A (en)* | 2020-03-18 | 2020-08-11 | 大箴(杭州)科技有限公司 | Method and device for processing address information in text |
| CN112560484A (en)* | 2020-11-09 | 2021-03-26 | 武汉数博科技有限责任公司 | Improved BERT training model and named entity recognition method and system |
| CN112613312A (en)* | 2020-12-18 | 2021-04-06 | 平安科技（深圳）有限公司 | Method, device and equipment for training entity naming recognition model and storage medium |
- 2021
- 2021-12-14CNCN202111523965.5Apatent/CN114328886A/enactivePending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569322A (en)* | 2019-07-26 | 2019-12-13 | 苏宁云计算有限公司 | Address information analysis method, device and system and data acquisition method |
| CN110688449A (en)* | 2019-09-20 | 2020-01-14 | 京东数字科技控股有限公司 | Address text processing method, device, equipment and medium based on deep learning |
| CN111209362A (en)* | 2020-01-07 | 2020-05-29 | 苏州城方信息技术有限公司 | Address data analysis method based on deep learning |
| CN111522901A (en)* | 2020-03-18 | 2020-08-11 | 大箴(杭州)科技有限公司 | Method and device for processing address information in text |
| CN112560484A (en)* | 2020-11-09 | 2021-03-26 | 武汉数博科技有限责任公司 | Improved BERT training model and named entity recognition method and system |
| CN112613312A (en)* | 2020-12-18 | 2021-04-06 | 平安科技（深圳）有限公司 | Method, device and equipment for training entity naming recognition model and storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115859986A (en)* | 2022-12-27 | 2023-03-28 | 上海捷晓信息技术有限公司 | Method, system and computer medium for realizing address named entity recognition based on tensorrT accelerated reasoning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111340334B (en) | Intelligent work order assignment method, system and medium | |
| CN109783637A (en) | Electric power overhaul text mining method based on deep neural network | |
| CN113362083B (en) | Repair reporting method and device, electronic equipment and storage medium | |
| CN109117715B (en) | A non-component identification method of architectural drawings based on SSD model | |
| CN111462752A (en) | Client intention identification method based on attention mechanism, feature embedding and BI-L STM | |
| CN114462556A (en) | Enterprise association industry chain classification method, training method, device, equipment and medium | |
| CN112257425A (en) | Power data analysis method and system based on data classification model | |
| CN115757124A (en) | A Method of Test Case Generation Based on Neural Network | |
| CN111651597A (en) | Multi-source heterogeneous commodity information classification method based on Doc2Vec and convolutional neural network | |
| CN118626771B (en) | Digital management system and method based on multi-source data information | |
| CN111310762A (en) | Intelligent medical bill identification method based on Internet of things | |
| CN115953788B (en) | Intelligent green melt attribute identification method and system based on OCR and NLP technologies | |
| CN114328886A (en) | Intelligent logistics address entity recognition system based on deep learning | |
| CN115438190B (en) | A distribution network fault auxiliary decision-making knowledge extraction method and system | |
| CN116468945A (en) | Detection method, device and equipment of transformer substation and storage medium | |
| Runhu et al. | Semantic segmentation using deeplabv3+ model for fabric defect detection | |
| CN110688312A (en) | A Transformer-based Software Defect Prediction Method | |
| CN115457304A (en) | A baggage damage analysis method and system based on target detection | |
| CN112488551B (en) | Hot line intelligent order dispatching method based on XGboost algorithm | |
| CN114491010B (en) | Training method and device for information extraction model | |
| CN112530582A (en) | Intelligent system for assisting cause of death classified coding | |
| CN109657691B (en) | Image semantic annotation method based on energy model | |
| CN111460160B (en) | An Event Clustering Method Based on Reinforcement Learning for Streaming Text Data | |
| CN116861905A (en) | Deep learning-based power standard entity relation extraction method and device | |
| CN116341520A (en) | Label recognition model training method, recognition method and related equipment |
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 |