Method and device for identifying abnormal tumble

Abstract

The invention discloses a method and a device for identifying abnormal tumble, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of obtaining a to-be-identified logistics order, and querying customer feedback information and corresponding delivery track information of the logistics order; under the condition that the client feedback information and the delivery track information are inquired, abnormal delivery identification is carried out on the logistics order according to the client feedback information and the delivery track information; and when at least one of the client feedback information and the delivery track information is not queried, obtaining communication content and feature data of the logistics order, inputting the communication content and the feature data into a pre-trained identification model, and further outputting an identification result about whether the logistics order is abnormally delivered or not. According to the embodiment of the invention, accurate and comprehensive abnormal delivery identification is realized, misjudgment can be avoided, abnormal delivery identification can be carried out on logistics orders without customer feedback information, and comprehensive delivery supervision is facilitated.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for identifying abnormal delivery. Background Art

[0002] In the logistics scenario, delivery quality is an important indicator for evaluating express logistics services. If an abnormal delivery occurs, it will not only bring a bad user experience to customers, reduce their willingness to use the express service again, but also may cause economic losses to customers. At present, the identification of abnormal delivery mainly comes from customer feedback information. As long as the customer feedback information involves abnormal delivery of logistics orders, and the time of customer feedback information is later than the delivery time, it is determined that the logistics order has been abnormally delivered.

[0003] In the process of implementing the present invention, the inventors found that the prior art has the following problems:

[0004] The existing strategy for judging and identifying abnormal delivery is relatively simple, and it only judges whether a logistics order has been abnormally delivered based on customer feedback information, which results in a large number of misjudgments. In addition, the existing method can only identify abnormal deliveries for logistics orders with customer feedback information, and cannot identify logistics orders that have actually been abnormally delivered but for which no customer feedback information has been received. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method and device for identifying abnormal deliveries, which realizes accurate and comprehensive identification of abnormal deliveries. It can not only accurately identify abnormal deliveries of logistics orders with customer feedback information to avoid misjudgment, but also identify abnormal deliveries of logistics orders without customer feedback information, which is more conducive to comprehensive delivery supervision.

[0006] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for identifying abnormal delivery is provided, comprising:

[0007] Obtain the logistics order to be identified, and query the customer feedback information and corresponding delivery track information of the logistics order;

[0008] When the customer feedback information and the delivery track information are queried, abnormal delivery identification is performed on the logistics order according to the customer feedback information and the delivery track information;

[0009] In the case that at least one of the customer feedback information and the delivery track information is not found, the communication content of the logistics order and the characteristic data of the logistics order are obtained, and the communication content and the characteristic data are input into a pre-trained recognition model, and then the recognition result of whether the logistics order has been abnormally delivered is output.

[0010] Optionally, the training samples of the recognition model are constructed in the following manner: sample orders marked as abnormally delivered are obtained from a historical logistics order set, and it is determined whether the number of the sample orders reaches a sample number threshold; when the number of the sample orders does not reach the sample number threshold, the text similarity between the alternative orders in the historical logistics order set other than the sample order and the sample order is calculated, and based on the calculation result of the text similarity and the sample number threshold, a target order is screened out from the alternative orders, and it is marked whether the target order has been abnormally delivered; based on the sample order and the target order, the training samples of the recognition model are constructed.

[0011] Optionally, the communication content and the feature data are input into a pre-trained recognition model, and then an identification result as to whether the logistics order has been abnormally delivered is output, including: inputting the communication content into the word vector layer of the recognition model, performing text segmentation on the communication content, querying the vectors of each word segmented by the text segmentation, and then obtaining a word vector matrix of the communication content; inputting the word vector matrix into the convolution layer of the recognition model, and performing local feature extraction on the word vector matrix according to a set convolution kernel to obtain a feature vector; inputting the feature vector into the pooling layer of the recognition model, performing pooling calculation on the feature vector, and obtaining text features of the communication text; splicing the feature data and the text features, and inputting the splicing result into the fully connected layer of the recognition model, and outputting the probability of abnormal delivery of the logistics order through activation function mapping.

[0012] Optionally, according to the set convolution kernel, local feature extraction is performed on the word vector matrix to obtain a feature vector, including: determining the sizes of multiple convolution kernels according to the number of rows and columns of the word vector matrix; performing local feature extraction by sliding each convolution kernel up and down on the word vector matrix to obtain a local feature vector extracted by each convolution kernel from the word vector matrix, and taking a set of each local feature vector as a feature vector.

[0013] Optionally, when the recognition result of abnormal delivery identification of the logistics order based on the customer feedback information and the delivery track information is to be verified, the method further includes: using the recognition model to identify abnormal delivery of the logistics order based on the communication content of the logistics order and the characteristic data of the logistics order.

[0014] Optionally, based on the customer feedback information and the delivery track information, abnormal delivery of the logistics order is identified, including: determining the track verification time based on the delivery time of the logistics order and a preset time threshold; querying the track point of the logistics order at the track verification time, calculating the distance between the track point and the delivery location of the logistics order, and judging whether there is any abnormality in the delivery of the logistics order based on the distance and the customer feedback information.

[0015] Optionally, when the communication content is in voice form, before inputting the communication content and the feature data into a pre-trained recognition model, the method also includes: converting the communication content in voice form into communication text, and performing data cleaning and refining on the communication text; inputting the communication content and the feature data into a pre-trained recognition model, including: inputting the communication text and the feature data after data cleaning and refinement into the pre-trained recognition model.

[0016] Optionally, when the identification result of the abnormal delivery of the logistics order is abnormal delivery, the method further includes: connecting the identification result of the abnormal delivery of the logistics order to a courier scoring system, so that the courier scoring system scores the courier of the logistics order according to the delivery identification result.

[0017] According to a second aspect of an embodiment of the present invention, there is provided an abnormal delivery identification device, comprising:

[0018] An information query module is used to obtain the logistics order to be identified, and query the customer feedback information and corresponding delivery track information of the logistics order;

[0019] A first identification module is used to identify abnormal delivery of the logistics order according to the customer feedback information and the delivery track information when the customer feedback information and the delivery track information are queried;

[0020] The second recognition module is used to obtain the communication content of the logistics order and the characteristic data of the logistics order when at least one of the customer feedback information and the delivery track information is not queried, and input the communication content and the characteristic data into a pre-trained recognition model, and then output a recognition result on whether the logistics order has been abnormally delivered.

[0021] According to a third aspect of an embodiment of the present invention, there is provided an abnormally delivered electronic identification device, comprising:

[0022] one or more processors;

[0023] a storage device for storing one or more programs,

[0024] When the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in the first aspect of the embodiment of the present invention.

[0025] According to a fourth aspect of an embodiment of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method provided by the first aspect of the embodiment of the present invention is implemented.

[0026] An embodiment of the invention has the following advantages or beneficial effects: by obtaining a logistics order to be identified, and querying customer feedback information and corresponding delivery track information of the logistics order; when the customer feedback information and the delivery track information are queried, the logistics order is identified as abnormally delivered according to the customer feedback information and the delivery track information; when at least one of the customer feedback information and the delivery track information is not queried, the communication content of the logistics order and the characteristic data of the logistics order are obtained, and the communication content and the characteristic data are input into a pre-trained recognition model, and then a technical solution is outputted to identify whether the logistics order has been abnormally delivered. An accurate and comprehensive abnormal delivery identification method is implemented, which can not only accurately identify the abnormal delivery of logistics orders with customer feedback information to avoid misjudgment, but also identify the abnormal delivery of logistics orders without customer feedback information, which is more conducive to comprehensive delivery supervision. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0028] Figure 1 is a schematic diagram of the main process of the abnormal delivery identification method according to an embodiment of the present invention;

[0029] Figure 2 is a schematic diagram of a word vector matrix of communication content according to an embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram of the characteristic data composition of a logistics order according to an embodiment of the present invention;

[0031] Figure 4 Schematic diagram of an abnormal successful delivery recognition implemented by a recognition model according to an embodiment of the present invention;

[0032] Figure 5 is a schematic diagram of a specific implementation of the abnormal delivery identification method according to an embodiment of the present invention;

[0033] Figure 6 2 is a schematic diagram of main modules of an abnormal delivery identification device according to an embodiment of the present invention;

[0034] Figure 7is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0035] Figure 8 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that in the technical solution of the present invention, the collection / collection, update, analysis, use, transmission, storage and other aspects of user personal information involved are in compliance with the provisions of relevant laws and regulations, are used for legal and reasonable purposes, are not shared, leaked or sold outside of these legal uses, and are subject to supervision and management by national regulatory authorities. Necessary measures should be taken for user personal information to selectively block the use or access to personal information data to prevent illegal access to such personal information data, ensure that persons who have access to personal information data comply with the provisions of relevant laws and regulations, and ensure the security of user personal information. In addition, once these user personal information data are no longer needed, the risks should be minimized by limiting or even prohibiting data collection and / or deleting data.

[0037] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0038] The existing judgment and identification strategy for abnormal delivery is relatively simple, and it determines whether a logistics order has been abnormally delivered based solely on customer feedback information, which results in a large number of misjudgments. In addition, the existing method can only identify abnormal deliveries for logistics orders with customer feedback information. It is impossible to identify logistics orders that have actually been abnormally delivered but for which no customer feedback information has been received, and cannot meet practical application requirements.

[0039] In order to solve the above-mentioned problems existing in the prior art, the present invention proposes a method for identifying abnormal delivery. Based on the logistics order to be identified, the corresponding customer feedback information and delivery track information are queried. If both the customer feedback information and the delivery track information are queried, the logistics order is identified for abnormal delivery at the business strategy level. Otherwise, according to the communication content and feature data of the logistics order, a model recognition strategy is used to identify abnormal delivery, thereby realizing an accurate and comprehensive method for identifying abnormal delivery. Not only can the abnormal delivery of logistics orders with customer feedback information be accurately identified to avoid misjudgment, but also the abnormal delivery of logistics orders without customer feedback information can be identified, which is more conducive to comprehensive delivery supervision.

[0040] Figure 1 FIG. 1 is a schematic diagram of the main process of the abnormal delivery identification method according to an embodiment of the present invention. Figure 1 As shown, the abnormal delivery identification method of the embodiment of the present invention includes the following steps S101 to S103.

[0041] Step S101: obtain a logistics order to be identified, and query customer feedback information and corresponding delivery track information of the logistics order.

[0042] Specifically, in the logistics order scenario, the delivery supervision system periodically pulls the logistics orders that are displayed as delivered from the background according to the preset supervision tasks to identify whether abnormal delivery has occurred, wherein the delivered logistics orders are the logistics orders to be identified. Abnormal delivery in the embodiment of the present invention mainly refers to irregular delivery (for example, the courier privately delivers to other addresses without the consent of the customer) and false delivery (for example, the courier privately signs for the delivery without completing the delivery). According to the order number of the logistics order to be identified, the customer feedback information and the corresponding delivery track information of the logistics order are queried, wherein the customer feedback information can be the customer's active feedback to the logistics company on the abnormal delivery of the logistics order, or the delivery supervision system responds to the receipt of the logistics order delivery completion information, and sends a feedback information sheet to the customer, and the customer selects the feedback information triggered by the failure to receive the goods in the feedback information sheet. The embodiment of the present invention does not make specific restrictions. It should be noted that the customer of the embodiment of the present invention can be the recipient or the sender.

[0043] In addition, it can be understood that the acquisition of the logistics order to be identified can be triggered by a scheduled task, or the delivery supervision system receives an abnormal delivery identification request initiated by a supervisor, and obtains the logistics order to be identified by parsing the abnormal delivery identification request.

[0044] Step S102: when the customer feedback information and the delivery track information are queried, abnormal delivery of the logistics order is identified according to the customer feedback information and the delivery track information.

[0045] Specifically, for the above-mentioned action of querying the customer feedback information and delivery track information of the logistics order, the corresponding query result may be that the customer feedback information and delivery track information are not queried, or one of the customer feedback information and delivery track information is queried, or both the customer feedback information and delivery track information are queried. If both the customer feedback information and the delivery track information can be queried, from the starting point of the customer's subjective behavior, the customer will not report abnormal delivery to the logistics company for no reason. Only when encountering abnormal delivery that affects the experience will the customer report to the logistics company. Therefore, logistics orders with customer feedback information are likely to have abnormal delivery.

[0046] Furthermore, in order to ensure the rigor of identification, there may be other influencing factors that may cause the customer to fail to receive the parcel that was normally delivered, such as when no one is at home, the courier leaves the parcel at the door with permission, but it is taken away by someone else; or the delivery address is the mailroom, and the customer fails to find the parcel that is placed in a corner or blocked due to the parcel stacking. In view of the fact that the delivery track information can objectively represent the actual route of the courier in delivering the logistics order, the embodiment of the present invention uses customer feedback information and delivery track information to jointly identify and determine whether the logistics order has been abnormally delivered.

[0047] According to one embodiment of the present invention, based on the customer feedback information and the delivery track information, the abnormal delivery of the logistics order is identified, including: determining the track verification time according to the delivery time of the logistics order and a preset time threshold; querying the track point of the logistics order at the track verification time, calculating the distance between the track point and the delivery place of the logistics order, and judging whether there is an abnormality in the delivery of the logistics order based on the distance and the customer feedback information.

[0048] Specifically, it can be seen from the delivery process of logistics orders that the courier needs to deliver the parcel to the delivery location before completing the delivery. Then, before and after the corresponding time of the courier's delivery, the delivery track must be near the delivery address. In addition, considering the problem of weak signals in special places, the courier will first deliver the goods to the delivery location, and then complete the delivery when the signal is good. In this way, there may be a geographical deviation between the courier's track and the actual delivery location at the time of the delivery, but the distance will not be very far. According to business experience and actual investigations, the embodiment of the present invention sets the 10 minutes before the delivery time as the time threshold, and uses the first 10 minutes of the delivery time as the track verification time. For example, if the delivery time is 9:40, then the track verification time is 9:30. Of course, you can also set 5 minutes or 10 minutes after the delivery time. Accordingly, the track verification time is the last 5 minutes or 10 minutes after the delivery time.

[0049] Furthermore, the track point of the logistics order at the time of track verification is queried from the delivery track information, that is, the specific geographical location at the time of track verification, and the geographical distance between the track point and the logistics order delivery location is calculated. When the time threshold is set to 10 minutes, the corresponding reasonable distance for normal delivery is set to 2km. That is, if the geographical distance exceeds 2km at this time, and the logistics order has customer feedback information about abnormal delivery, then it is determined that the delivery of the logistics order is abnormal. If the distance between the track point and the delivery location at the time of track verification does not exceed 2km, although there is customer feedback information, the delivery supervision system will mark this as pending verification, that is, the identification result is pending verification, and relevant personnel will visit the customer by phone and verify the logistics order with the courier to determine whether abnormal delivery has occurred to ensure that no misjudgment occurs.

[0050] Step S103: When at least one of the customer feedback information and the delivery track information is not found, the communication content of the logistics order and the feature data of the logistics order are obtained, and the communication content and the feature data are input into a pre-trained recognition model, and then a recognition result of whether the logistics order has been abnormally delivered is output.

[0051] Specifically, if no customer feedback information and delivery track information are found, or only one of the customer feedback information and delivery track information is found, the above-mentioned business strategy is no longer used to identify the logistics order. The embodiment of the present invention pre-trains a recognition model that can be used to identify abnormal delivery based on a deep learning algorithm. The recognition model introduces the communication content between the courier and the customer, the customer service and the customer of the logistics order, and the characteristic data of the logistics order, such as the courier's delivery time (if the delivery time is very short, the courier is likely to not actually deliver the goods properly), the time between the courier's receipt time and the planned delivery time (if the time is very short, the courier is worried about timeout and often chooses to deliver the goods first and then deliver the goods), the courier's abnormal delivery rate, etc. The communication content and characteristic data are input into the recognition model, and the recognition model is used to perform recognition calculations for abnormal delivery, and the recognition result of whether the logistics order has abnormal delivery output by the recognition model is obtained.

[0052] According to one embodiment of the present invention, the training samples of the recognition model are constructed in the following manner: sample orders marked as abnormally delivered are obtained from a historical logistics order set, and it is determined whether the number of the sample orders reaches a sample number threshold; when the number of the sample orders does not reach the sample number threshold, the text similarity between the alternative orders in the historical logistics order set other than the sample order and the sample order is calculated, and based on the calculation result of the text similarity and the sample number threshold, a target order is screened out from the alternative orders, and it is marked whether the target order has been abnormally delivered; based on the sample order and the target order, the training samples of the recognition model are constructed.

[0053] Specifically, the performance of the recognition model determines the accuracy of recognition, and the performance of the recognition model mainly depends on the training samples. High-quality training samples ensure the training effect of the recognition model. In order to improve the recognition ability of the recognition model, the embodiment of the present invention uses manually labeled standard samples to construct training samples. First, sample orders marked as abnormal delivery are obtained from the historical logistics order set, the number of sample orders is counted, and it is determined whether the number reaches the preset sample quantity threshold. If the sample quantity threshold is reached, these sample orders are used to construct training samples. If the sample quantity threshold is not reached, the orders other than the sample orders in the historical logistics order set are used as alternative orders. Based on the text features of each sample order, the text similarity algorithm is used to calculate the text similarity between the alternative orders and the sample orders. The purpose is to screen out some target orders with high text similarity with the sample orders from the alternative orders.

[0054] Then, the specific screening criteria are mainly based on the gap between the sample orders and the sample quantity threshold. For example, if the gap is 500, then considering the loss of manual annotation and the possibility that orders similar to the sample orders are delivered normally, the required number of target orders can be expanded to 3,000. According to the required number of target orders determined by the sample quantity threshold, the calculation results of the above text similarity are sorted from large to small according to the similarity, and the orders with the required number are selected from front to back as target orders. Then, with the help of manual annotation capabilities, these target orders are marked as abnormally delivered, and the target orders and sample orders marked as abnormally delivered are used to construct training samples for the recognition model.

[0055] According to another embodiment of the present invention, the communication content and the feature data are input into a pre-trained recognition model, and then a recognition result as to whether the logistics order has been abnormally delivered is output, including: inputting the communication content into the word vector layer of the recognition model, performing text segmentation on the communication content, querying the vectors of each word segmented by the text segmentation, and then obtaining a word vector matrix of the communication content; inputting the word vector matrix into the convolution layer of the recognition model, and performing local feature extraction on the word vector matrix according to a set convolution kernel to obtain a feature vector; inputting the feature vector into the pooling layer of the recognition model, performing pooling calculation on the feature vector, and obtaining text features of the communication text; splicing the feature data and the text features, and inputting the splicing result into the fully connected layer of the recognition model, and outputting the probability of abnormal delivery of the logistics order through activation function mapping.

[0056] Specifically, considering that the recognition model mainly relies on text to identify abnormal deliveries, the recognition model specifically includes a word vector layer, a convolution layer, a pooling layer, and a fully connected layer. First, the communication content is input into the word vector layer, and the text of the communication content is divided into several words through the word segmentation tool. For example, "Logistics shows that I have signed for it, but I have not received the goods." After word segmentation, it becomes "logistics", "show", "I", "have", "signed", "but", "I", "and", "no", "received", "goods". After completing the text segmentation, query the index corresponding to each word from the word vector file, and obtain the vector corresponding to the word from the pre-trained word vector matrix based on the index. Finally, the vectors of each word are aggregated together to obtain the word vector matrix of the communication content.

[0057] Figure 2 It is a schematic diagram of the word vector matrix of the communication content of an embodiment of the present invention. Each row represents the vector of the corresponding word segment. All word segments are aggregated together in order to form a word vector matrix. It can be understood that the vector dimension of each word segment is the same, and the same dimension facilitates subsequent calculation and processing.

[0058] Furthermore, the purpose of the convolution layer in the recognition model is to extract local features. The word vector matrix of the communication content is input into the convolution layer, and the word vector matrix is ​​extracted locally using multiple set convolution kernels to obtain a feature vector. The feature vector is then input into the pooling layer. The purpose of pooling is to further extract higher-order features based on the features extracted by the convolution layer. Some detailed information is ignored through the pooling operation, so that the model can learn more needed high-order information. There are currently two main calculation methods for specific pooling operations: average pooling and maximum pooling. Average pooling is to find the average of all values ​​in each channel, and maximum pooling is to find the maximum value of all values ​​in each channel. The embodiment of the present invention chooses to use the maximum pooling calculation method to obtain the most critical text features in the communication text.

[0059] Although text features include communication information during the entire delivery process of logistics orders, it is also necessary to consider the characteristic data of logistics orders other than communication information. The characteristic data of logistics orders is also of certain significance for the identification of abnormal delivery. For example, if the courier has a high abnormal delivery rate, the risk of abnormal delivery of the logistics order will also increase. Therefore, it is necessary to introduce the characteristic data of logistics orders.

[0060] Figure 3 : is a schematic diagram of the characteristic data composition of the logistics order of the embodiment of the present invention. It can be seen that the characteristic data of the logistics order mainly includes the characteristics of customer feedback information, the delivery characteristics provided by the map data management department, the external interaction characteristics, the portrait characteristics, and the timeliness characteristics. Among them, the characteristics of the customer feedback information mainly come from the customer feedback information of the logistics order. The customer may send customer feedback information to the logistics company multiple times, each time as an event, so the number of changes in the content of the customer feedback information, the first event summary, the final event summary, as well as the duration of each communication and the total duration of communication will be recorded.

[0061] The delivery features are provided by the cooperating map data management department. Based on the map data features during the delivery process, it is preliminarily determined whether there are any abnormal deliveries in the logistics order, and the credibility of the abnormal deliveries.

[0062] The external interaction features are mainly used to characterize the outbound call records between the courier and the customer during the delivery process, including whether an outbound call was made, the duration of the outbound call, and the duration of the call that was not connected but only rang.

[0063] The portrait features mainly count the relevant logistics order volumes from the sender, recipient, courier and site respectively, and finely divide the number of different types of customer feedback information, including the number of all types of feedback information (all-type feedback information unit volume) and the number of customer feedback information only involving abnormal delivery mentioned in the embodiment of the present invention (customer feedback information unit volume), as well as the corresponding proportion of all-type feedback information unit volume (all-type feedback rate) and the proportion of customer feedback information unit volume (feedback rate).

[0064] The timeliness feature is mainly characterized by the inspection time, the reserved time for promised delivery, the planned time from inspection to delivery, and the total planned time.

[0065] Finally, the feature data of the logistics order and the text features processed by the pooling layer are concatenated, and the concatenated result is input into the fully connected layer, and the probability of abnormal delivery of the logistics order is output through the activation function softmax mapping.

[0066] According to another embodiment of the present invention, local features are extracted from the word vector matrix according to the set convolution kernel to obtain a feature vector, including: determining the sizes of multiple convolution kernels according to the number of rows and columns of the word vector matrix; performing local feature extraction by sliding each convolution kernel up and down on the word vector matrix to obtain a local feature vector extracted from the word vector matrix by each convolution kernel, and taking a set of each local feature vector as a feature vector.

[0067] For example, Figure 4 Schematic diagram of the recognition model of the embodiment of the present invention to realize abnormal delivery recognition. Taking the communication content of "Logistics shows that I have signed for the goods but I have not received them" as an example, based on the word vector matrix after the word vector layer has vectorized the communication content, three convolution kernels are set, and the number of rows of the convolution kernel is 2, 3, and 4 respectively, and the number of columns is kept the same as the number of columns of the word vector matrix. See Figure 4 Thick solid line boxes with three different numbers of rows in the word vector matrix.

[0068] The three convolution kernels are slid up and down in the word vector matrix respectively. Each sliding position is a local feature extraction. When the three convolution kernels complete the local feature extraction in the word vector matrix, the corresponding three local feature vectors are obtained. See Figure 4 The three local feature vectors A, B, and C correspond to the convolution kernels in rows 2, 3, and 4, respectively.

[0069] The feature vector composed of these three local feature vectors is then input into the pooling layer, which performs maximum pooling on them respectively, selects the maximum value in each local feature vector, and performs feature splicing together with the feature data of the logistics order. The splicing result is input into the fully connected layer, and then the probability of abnormal delivery of the logistics order is output through the activation function.

[0070] According to another embodiment of the present invention, when the recognition result of abnormal delivery identification of the logistics order based on the customer feedback information and the delivery track information is to be verified, the method further includes: using the recognition model to identify abnormal delivery of the logistics order based on the communication content of the logistics order and the characteristic data of the logistics order.

[0071] Specifically, considering the above-mentioned use of business strategies, in the process of identifying abnormal delivery of logistics orders based on customer feedback information and delivery track information, special situations will inevitably be encountered, and the distance between the track point and the destination does not exceed the set reasonable distance of 2km. At this time, the identification result is pending verification. For this special situation, manual verification can be used to solve it, or the recognition model of this case can be used to obtain the communication content and characteristic data of the logistics order according to the above-mentioned model for the identification method of abnormal delivery, and the communication content and characteristic data are input into the recognition model, and then the recognition result of whether the logistics order has been abnormally delivered is output.

[0072] According to another embodiment of the present invention, when the communication content is in voice form, before inputting the communication content and the feature data into a pre-trained recognition model, the method further includes: converting the communication content in voice form into communication text, and performing data cleaning and refining on the communication text; inputting the communication content and the feature data into a pre-trained recognition model, including: inputting the communication text after data cleaning and refining and the feature data into the pre-trained recognition model.

[0073] Specifically, usually, the communication content of logistics orders is in voice form, such as telephone recording. Understandably, before the telephone recording, the customer will be reminded that the call content will be recorded so that the customer is aware of the recording behavior. The voice communication content is converted into communication text through speech recognition technology. Considering that the converted text may contain some useless text data such as symbols, special characters, and modal particles, it is also necessary to clean the communication text, refine the communication text, remove redundant text information, and improve the quality of the text data, and then input the high-quality communication text and feature data after data cleaning and refinement into the recognition model.

[0074] According to another embodiment of the present invention, when the identification result of the abnormal delivery of the logistics order is abnormal delivery, the method further includes: connecting the identification result of the abnormal delivery of the logistics order to a courier scoring system, so that the courier scoring system scores the courier of the logistics order according to the delivery identification result.

[0075] Specifically, when an abnormal delivery of a logistics order is identified, the embodiment of the present invention connects the abnormal delivery identification result to the courier scoring system through a dedicated interface. The courier scoring system queries the corresponding courier based on the received abnormally delivered logistics order, updates the courier's score according to the scoring rules, and notifies the courier.

[0076] Figure 5 It is a schematic diagram of a specific implementation of the abnormal delivery identification method of an embodiment of the present invention. For each logistics order in the full amount of logistics orders to be identified, first query the customer feedback information and corresponding delivery track information of the logistics order. If both can be queried, then execute the identification method of the business strategy, and determine the track verification time according to the delivery time of the logistics order and the preset time threshold; query the track point of the logistics order at the track verification time, and calculate the distance between the track point and the logistics order receiving place. If the distance exceeds the preset reasonable distance, and there is customer feedback information about abnormal delivery, then the logistics order is determined to be abnormally delivered. Of course, if the distance does not exceed the preset reasonable distance, the identification result is to be verified at this time, and the corresponding logistics order can be transferred to the abnormal delivery identification method of the model for identification.

[0077] If at least one of the customer feedback information and the delivery trajectory information is not found, the model's abnormal delivery identification method is executed, and the recorded communication content is first converted and preprocessed such as data cleaning, and then the communication content is vectorized through the word vector layer of the recognition model to generate a word vector matrix of the communication content; then the local feature extraction and key feature extraction of the convolution layer and pooling layer of the recognition model are performed to obtain the text features of the communication content, and the text features and the feature data of the logistics order are combined and spliced, and the probability of abnormal delivery of each logistics order is predicted and calculated through the fully connected layer of the recognition model, and then the logistics orders are sorted according to the probability, and the logistics orders with abnormal delivery are determined according to the preset probability threshold.

[0078] Furthermore, when the identification result of abnormal delivery of a logistics order is abnormal delivery, the identification result of abnormal delivery of the logistics order can also be connected to the courier scoring system so that the courier scoring system can score the courier of the logistics order according to the delivery identification result.

[0079] The embodiment of the present invention queries the corresponding customer feedback information and delivery track information based on the logistics order to be identified. If both the customer feedback information and the delivery track information are queried, the logistics order is identified for abnormal delivery at the business strategy level. Otherwise, the abnormal delivery is identified using a model recognition strategy based on the communication content and feature data of the logistics order, thereby realizing an accurate and comprehensive method for identifying abnormal delivery. Not only can the abnormal delivery of logistics orders with customer feedback information be accurately identified to avoid misjudgment, but also the abnormal delivery of logistics orders without customer feedback information can be identified, which is more conducive to comprehensive delivery supervision.

[0080] Figure 6 FIG. 2 is a schematic diagram of the main modules of the abnormal delivery identification device according to an embodiment of the present invention. Figure 6 As shown, the abnormal delivery identification device 600 mainly includes an information query module 601 , a first identification module 602 and a second identification module 603 .

[0081] The information query module 601 is used to obtain the logistics order to be identified, and query the customer feedback information and corresponding delivery track information of the logistics order;

[0082] A first identification module 602 is used to identify abnormal delivery of the logistics order according to the customer feedback information and the delivery track information when the customer feedback information and the delivery track information are queried;

[0083] The second recognition module 603 is used to obtain the communication content of the logistics order and the characteristic data of the logistics order when at least one of the customer feedback information and the delivery track information is not found, and input the communication content and the characteristic data into a pre-trained recognition model, and then output a recognition result on whether the logistics order has been abnormally delivered.

[0084] According to one embodiment of the present invention, the training samples of the recognition model are constructed in the following manner: sample orders marked as abnormally delivered are obtained from a historical logistics order set, and it is determined whether the number of the sample orders reaches a sample number threshold; when the number of the sample orders does not reach the sample number threshold, the text similarity between the alternative orders in the historical logistics order set other than the sample order and the sample order is calculated, and based on the calculation result of the text similarity and the sample number threshold, a target order is screened out from the alternative orders, and it is marked whether the target order has been abnormally delivered; based on the sample order and the target order, the training samples of the recognition model are constructed.

[0085] According to another embodiment of the present invention, the second recognition module 603 is also used to: input the communication content into the word vector layer of the recognition model, perform text segmentation on the communication content, query the vectors of each word segmentation obtained by text segmentation, and then obtain the word vector matrix of the communication content; input the word vector matrix into the convolution layer of the recognition model, and extract local features of the word vector matrix according to the set convolution kernel to obtain a feature vector; input the feature vector into the pooling layer of the recognition model, perform pooling calculation on the feature vector, and obtain text features of the communication text; splice the feature data and the text features, and input the splicing result into the fully connected layer of the recognition model, and output the probability of abnormal delivery of the logistics order through activation function mapping.

[0086] According to another embodiment of the present invention, the second recognition module 603 is also used to: determine the sizes of multiple convolution kernels according to the number of rows and columns of the word vector matrix; extract local features by sliding each convolution kernel up and down on the word vector matrix, obtain the local feature vector extracted by each convolution kernel from the word vector matrix, and use the set of each local feature vector as a feature vector.

[0087] According to another embodiment of the present invention, when the recognition result of abnormal delivery identification of the logistics order based on the customer feedback information and the delivery track information is to be verified, the abnormal delivery identification device 600 also includes an access recognition model module (not shown in the figure), which is used to: use the recognition model to identify abnormal delivery of the logistics order based on the communication content of the logistics order and the characteristic data of the logistics order.

[0088] According to another embodiment of the present invention, the first identification module 602 is further used to: determine the track verification time according to the successful delivery time of the logistics order and a preset time threshold; query the track point of the logistics order at the track verification time, calculate the distance between the track point and the delivery place of the logistics order, and determine whether there is any abnormality in the successful delivery of the logistics order based on the distance and the customer feedback information.

[0089] According to another embodiment of the present invention, when the communication content is in voice form, the abnormal successful delivery identification device 600 also includes a voice processing module (not shown in the figure), which is used to: convert the communication content in voice form into communication text before inputting the communication content and the feature data into a pre-trained recognition model, and perform data cleaning and refinement on the communication text; the second recognition module 603 is also used to: input the communication text and the feature data after data cleaning and refinement into the pre-trained recognition model.

[0090] According to another embodiment of the present invention, when the identification result of abnormal delivery of the logistics order is abnormal delivery, the abnormal delivery identification device 600 also includes a result access module (not shown in the figure), which is used to: connect the abnormal delivery identification result of the logistics order to the courier rating system, so that the courier rating system scores the courier of the logistics order according to the abnormal delivery identification result.

[0091] Figure 7 is an exemplary system architecture diagram to which embodiments of the present invention can be applied.

[0092] like Figure 7 As shown, system architecture 700 may include terminal devices 701, 702, 703, network 704 and server 705. Network 704 is used to provide a medium for communication links between terminal devices 701, 702, 703 and server 705. Network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0093] Users can use terminal devices 701, 702, 703 to interact with server 705 through network 704 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 701, 702, 703, such as abnormal delivery identification applications, etc. (only as an example).

[0094] The terminal devices 701 , 702 , and 703 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0095] Server 705 may be a server that provides various services, such as a background management server that provides support for abnormal delivery performed by users using terminal devices 701, 702, and 703 (for example only). The background management server may obtain the logistics order to be identified, and query the customer feedback information and corresponding delivery track information of the logistics order; in the case where the customer feedback information and the delivery track information are queried, the abnormal delivery of the logistics order is identified based on the customer feedback information and the delivery track information; in the case where at least one of the customer feedback information and the delivery track information is not queried, the communication content of the logistics order and the characteristic data of the logistics order are obtained, and the communication content and the characteristic data are input into a pre-trained recognition model, and then the recognition result of whether the logistics order has been abnormally delivered is output, and the processing result (such as the recognition result, etc. - for example only) is fed back to the terminal device.

[0096] It should be noted that the abnormal delivery identification method provided in the embodiment of the present invention is generally executed by the server 705 , and accordingly, the abnormal delivery identification device is generally set in the server 705 .

[0097] It should be understood that Figure 7 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0098] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. Figure 8 The terminal device or server shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0099] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0100] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0101] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0102] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination thereof.

[0103] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0104] The units involved in the embodiments of the present invention may be implemented by software or hardware. The units described may also be arranged in a processor, for example, it may be described as follows: a processor includes: an information query module, a first identification module and a second identification module.

[0105] Among them, the names of these modules do not constitute a limitation of the module itself in some cases. For example, the information query module can also be described as "a module for obtaining the logistics order to be identified, and querying the customer feedback information and corresponding delivery track information of the logistics order."

[0106] On the other hand, the present invention also provides a computer-readable medium, which may be included in the device described in the embodiment; or it may exist independently and not be assembled into the device. The computer-readable medium carries one or more programs, and when the one or more programs are executed by a device, the device includes: obtaining a logistics order to be identified, and querying the customer feedback information and corresponding delivery track information of the logistics order; in the case where the customer feedback information and the delivery track information are queried, the logistics order is abnormally delivered according to the customer feedback information and the delivery track information; in the case where at least one of the customer feedback information and the delivery track information is not queried, the communication content of the logistics order and the characteristic data of the logistics order are obtained, and the communication content and the characteristic data are input into a pre-trained recognition model, and then the recognition result of whether the logistics order has been abnormally delivered is output.

[0107] The technical solution according to the embodiment of the present invention has the following advantages or beneficial effects: by obtaining the logistics order to be identified, and querying the customer feedback information and the corresponding delivery track information of the logistics order; when the customer feedback information and the delivery track information are queried, the logistics order is identified as abnormally delivered according to the customer feedback information and the delivery track information; when at least one of the customer feedback information and the delivery track information is not queried, the communication content of the logistics order and the characteristic data of the logistics order are obtained, and the communication content and the characteristic data are input into a pre-trained recognition model, and then the technical solution of outputting the recognition result of whether the logistics order has been abnormally delivered is realized, an accurate and comprehensive abnormal delivery identification method is realized, which can not only accurately identify the abnormal delivery of logistics orders with customer feedback information to avoid misjudgment, but also identify the abnormal delivery of logistics orders without customer feedback information, which is more conducive to comprehensive delivery supervision.

[0108] The specific implementation methods described herein do not constitute limitations on the scope of protection of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A method for identifying abnormal delivery. It is characterized in that include: Obtain the logistics order to be identified, and query the customer feedback information and corresponding delivery track information of the logistics order; When the customer feedback information and the delivery track information are queried, abnormal delivery identification is performed on the logistics order according to the customer feedback information and the delivery track information; In the case that at least one of the customer feedback information and the delivery track information is not found, the communication content of the logistics order and the characteristic data of the logistics order are obtained, and the communication content and the characteristic data are input into a pre-trained recognition model, and then the recognition result of whether the logistics order has been abnormally delivered is output.

2. The method according to claim 1, It is characterized in that The training samples of the recognition model are constructed in the following way: Obtain sample orders marked as abnormally delivered from a historical logistics order set, and determine whether the quantity of the sample orders reaches a sample quantity threshold; When the number of the sample orders does not reach the sample number threshold, the text similarity between the candidate orders in the historical logistics order set other than the sample order and the sample order is calculated, and according to the calculation result of the text similarity and the sample number threshold, the target order is screened out from the candidate orders, and whether the target order is abnormally delivered is marked; A training sample for the recognition model is constructed according to the sample order and the target order.

3. The method according to claim 1, It is characterized in that The communication content and the feature data are input into a pre-trained recognition model, and then a recognition result on whether the logistics order has been abnormally delivered is output, including: Input the communication content into the word vector layer of the recognition model, perform text segmentation on the communication content, query the vectors of each word segmentation obtained by the text segmentation, and then obtain the word vector matrix of the communication content; Input the word vector matrix into the convolution layer of the recognition model, and extract local features of the word vector matrix according to the set convolution kernel to obtain a feature vector; Inputting the feature vector into the pooling layer of the recognition model, performing pooling calculation on the feature vector, and obtaining text features of the communication text; The feature data and the text feature are concatenated, and the concatenation result is input into the fully connected layer of the recognition model, and the probability of abnormal delivery of the logistics order is output through activation function mapping.

4. The method according to claim 3, It is characterized in that According to the set convolution kernel, local features are extracted from the word vector matrix to obtain feature vectors, including: Determine the sizes of multiple convolution kernels according to the number of rows and columns of the word vector matrix; Local feature extraction is performed by sliding each convolution kernel up and down on the word vector matrix to obtain a local feature vector extracted by each convolution kernel from the word vector matrix, and a set of each local feature vector is used as a feature vector.

5. The method according to claim 1, It is characterized in that When the result of abnormal successful delivery identification for the logistics order is to be verified according to the customer feedback information and the delivery track information, the method further includes: According to the communication content of the logistics order and the characteristic data of the logistics order, the recognition model is used to identify abnormal deliveries of the logistics order.

6. The method according to claim 1, It is characterized in that According to the customer feedback information and the delivery track information, abnormal delivery of the logistics order is identified, including: Determine the track verification time according to the delivery time of the logistics order and the preset time threshold; Query the track point of the logistics order at the track verification time, calculate the distance between the track point and the delivery location of the logistics order, and determine whether there is any abnormality in the delivery of the logistics order based on the distance and the customer feedback information.

7. The method according to claim 1, It is characterized in that In the case where the communication content is in the form of speech, before inputting the communication content and the feature data into a pre-trained recognition model, the method further includes: Convert the communication content in the form of voice into communication text, and perform data cleaning and refining on the communication text; Inputting the communication content and the feature data into a pre-trained recognition model includes: The cleaned and refined communication text and the feature data are input into the pre-trained recognition model.

8. The method according to claim 1, It is characterized in that When the identification result of the abnormal delivery of the logistics order is that abnormal delivery has occurred, the method further includes: The abnormal successful delivery identification result of the logistics order is connected to the courier scoring system, so that the courier scoring system scores the courier of the logistics order according to the successful delivery identification result.

9. An abnormal delivery identification device, It is characterized in that include: An information query module is used to obtain the logistics order to be identified, and query the customer feedback information and corresponding delivery track information of the logistics order; A first identification module is used to identify abnormal delivery of the logistics order according to the customer feedback information and the delivery track information when the customer feedback information and the delivery track information are queried; The second recognition module is used to obtain the communication content of the logistics order and the characteristic data of the logistics order when at least one of the customer feedback information and the delivery track information is not queried, and input the communication content and the characteristic data into a pre-trained recognition model, and then output a recognition result on whether the logistics order has been abnormally delivered.

10. A mobile electronic device terminal, It is characterized in that include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

11. A computer readable medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images