Data analysis method and system based on logistics transportation

Abstract

The embodiment of the invention provides a data analysis method and system based on logistics transportation, and the method comprises the steps: carrying out the dialogue text processing through combining a basic sample logistics transportation dialogue text with a target text disturbance enhancement feature, and generating a first semantic feature space and a target reconstruction logistics transportation dialogue text. Furthermore, by fusing the difference semantic features and the first semantic feature space, a richer second semantic feature space is constructed, and the accuracy and comprehensiveness of dialogue intention mining are effectively improved. And finally, dialogue intention mining is performed according to the second semantic feature space, and model parameters are optimized in combination with the dialogue intention annotation data, so that the performance and precision of the logistic transportation dialogue intention mining model are remarkably improved, and powerful support is provided for intelligent dialogue processing in the field of logistic transportation.

Description

Technical Field

[0001] The present application relates to the field of smart logistics technology, and more specifically, to a data analysis method and system based on logistics transportation. Background Art

[0002] In the field of logistics and transportation, with the continuous expansion and complexity of business, a large amount of logistics and transportation conversation data is generated and accumulated. These data contain rich information and value, which is of great significance for improving logistics and transportation efficiency, optimizing service quality and reducing costs. However, how to effectively analyze and mine the intentions and information in these conversation data has always been a problem that needs to be solved.

[0003] Traditional data analysis methods often rely on manual labeling and rule-making, which is not only time-consuming and labor-intensive, but also difficult to adapt to large-scale and diverse logistics and transportation conversation data. Summary of the invention

[0004] In view of this, the purpose of this application is to provide a data analysis method and system based on logistics transportation.

[0005] In conjunction with the first aspect of the present application, a data analysis method based on logistics and transportation is provided, which is applied to a data analysis system based on logistics and transportation, and the method comprises:

[0006] Based on the basic sample logistics and transportation dialogue text, the logistics and transportation dialogue intention mining model is parameter-learned to generate a logistics and transportation dialogue intention mining model that has completed parameter learning; wherein the parameter learning includes:

[0007] Processing the dialogue text according to the basic sample logistics and transportation dialogue text and the target text disturbance enhancement features to generate a first semantic feature space and a target reconstructed logistics and transportation dialogue text;

[0008] A second semantic feature space is generated by fusing the difference semantic features corresponding to the basic sample logistics and transportation dialogue text and the first semantic feature space; the difference semantic features are determined by comparing the difference between the basic sample logistics and transportation dialogue text and the target reconstructed logistics and transportation dialogue text;

[0009] Conducting conversation intention mining based on the second semantic feature space to generate conversation intention mining results for the basic sample logistics transportation conversation text;

[0010] Based on the conversation intention mining results and the conversation intention annotation data of the basic sample logistics and transportation conversation text, the model parameter information of the logistics and transportation conversation intention mining model is optimized.

[0011] In a possible implementation of the first aspect, the step of fusing the differential semantic features corresponding to the basic sample logistics transportation dialogue text with the first semantic feature space to generate a second semantic feature space includes:

[0012] Performing attention analysis based on the differential semantic features corresponding to the basic sample logistics transportation dialogue text to determine first attention coefficient information; the first attention coefficient information includes first semantic importance factors under multiple dimensions;

[0013] The second semantic feature space is generated by fusing the first semantic feature space and the second attention coefficient information; the second attention coefficient information includes the second semantic importance factors under the multiple dimensions, and the sum of the first semantic importance factor and the second semantic importance factor under the same dimension is equal to 1.

[0014] In a possible implementation of the first aspect, performing attention analysis based on the difference semantic features corresponding to the basic sample logistics transportation dialogue text to determine the first attention coefficient information includes:

[0015] Performing spatial encoding representation on the difference semantic features to generate a third semantic feature space;

[0016] Performing attention allocation on the third semantic feature space to generate a fourth semantic feature space;

[0017] Linear mapping is performed according to the fourth semantic feature space to generate first attention coefficient information.

[0018] In a possible implementation of the first aspect, the fusing the first semantic feature space and the second attention coefficient information to generate the second semantic feature space includes:

[0019] Performing spatial deconstruction based on the first semantic feature space to generate a fifth semantic feature space;

[0020] The fifth semantic feature space and the second attention coefficient information are fused to generate a second semantic feature space.

[0021] In a possible implementation of the first aspect, the fusing the fifth semantic feature space and the second attention coefficient information to generate the second semantic feature space includes:

[0022] Generating a sixth semantic feature space by fusing the fifth semantic feature space and the second attention coefficient information;

[0023] The sixth semantic feature space is feature enhanced according to the first semantic feature space to generate a second semantic feature space.

[0024] In a possible implementation of the first aspect, the processing of the dialogue text according to the basic sample logistics and transportation dialogue text and the target text disturbance enhancement feature to generate the first semantic feature space and the target reconstructed logistics and transportation dialogue text includes:

[0025] Performing spatial encoding representation on the basic sample logistics transportation dialogue text to generate a first temporary semantic feature space;

[0026] Performing spatial encoding representation according to the first temporary semantic feature space to generate a first semantic feature space;

[0027] Performing spatial encoding representation according to the first semantic feature space to generate a second temporary semantic feature space;

[0028] Performing spatial encoding representation according to the second temporary semantic feature space and the target text disturbance enhancement feature to generate a seventh semantic feature space;

[0029] The logistics and transportation dialogue text is reconstructed according to the seventh semantic feature space to generate a target reconstructed logistics and transportation dialogue text.

[0030] In a possible implementation of the first aspect, optimizing the model parameter information of the logistics and transportation dialogue intention mining model according to the dialogue intention mining result and the dialogue intention annotation data of the basic sample logistics and transportation dialogue text includes:

[0031] Perform error calculation based on the conversation intention mining result and the conversation intention annotation data of the basic sample logistics transportation conversation text to generate a first training error parameter;

[0032] Perform error calculation based on the target reconstructed logistics and transportation dialogue text and the basic sample logistics and transportation dialogue text to generate a second training error parameter;

[0033] Based on the first training error parameter and the second training error parameter, the model parameter information of the logistics and transportation dialogue intention mining model is optimized.

[0034] In a possible implementation of the first aspect, the conversation intention mining result includes a first confidence level that the basic sample logistics transportation conversation text is classified as conversation intention annotation data and a second confidence level that the basic sample logistics transportation conversation text is classified as a set conversation intention, and the set conversation intention is not the conversation intention annotation data;

[0035] The step of performing error calculation based on the dialogue intention mining result and the dialogue intention annotation data of the basic sample logistics transportation dialogue text to generate a first training error parameter includes:

[0036] For the X first basic sample logistics transportation dialogue texts, determine the feature distance according to the first confidence level corresponding to the first basic sample logistics transportation dialogue text and the set value, and generate a first training error sub-parameter;

[0037] For Y second basic sample logistics transportation dialogue texts, determine the feature distance according to the second confidence level corresponding to the second basic sample logistics transportation dialogue text and the set value, and generate a second training error sub-parameter;

[0038] A first training error parameter is determined according to the first training error sub-parameter and the second training error sub-parameter.

[0039] In a possible implementation of the first aspect, the method further includes:

[0040] Based on the logistics and transportation dialogue intention mining model, dialogue text processing is performed on the candidate logistics and transportation dialogue text to generate a dialogue intention mining result of the candidate logistics and transportation dialogue text.

[0041] In combination with the second aspect of the present application, a data analysis system based on logistics and transportation is provided, wherein the data analysis system based on logistics and transportation comprises a machine-readable storage medium and a processor, wherein the machine-readable storage medium stores machine-executable instructions, and when the processor executes the machine-executable instructions, the data analysis system based on logistics and transportation implements the aforementioned data analysis method based on logistics and transportation.

[0042] In conjunction with the third aspect of the present application, a computer-readable storage medium is provided, in which computer-executable instructions are stored. When the computer-executable instructions are executed, the aforementioned data analysis method based on logistics and transportation is implemented.

[0043] In combination with any of the above aspects, the embodiment of the present application processes the dialogue text by combining the basic sample logistics and transportation dialogue text with the target text perturbation enhancement features to generate a first semantic feature space and a target reconstructed logistics and transportation dialogue text. Furthermore, by fusing the differential semantic features with the first semantic feature space, a richer second semantic feature space is constructed, which effectively improves the accuracy and comprehensiveness of dialogue intent mining. Finally, dialogue intent mining is performed based on the second semantic feature space, and the model parameters are optimized in combination with the dialogue intent annotation data, which significantly improves the performance and accuracy of the logistics and transportation dialogue intention mining model, providing strong support for intelligent dialogue processing in the field of logistics and transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained by combining these drawings without paying creative work.

[0045] Figure 1 A flow chart of a data analysis method based on logistics and transportation provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. According to the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0048] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0049] Figure 1 The flowchart of the data analysis method based on logistics and transportation provided by the embodiment of the present application is shown. It should be understood that in other embodiments, the order of some steps in the data analysis method based on logistics and transportation of this embodiment can be shared with each other according to actual needs, or some steps can be omitted or maintained. The data analysis method based on logistics and transportation includes:

[0050] Based on the basic sample logistics and transportation dialogue text, the logistics and transportation dialogue intention mining model is parameter-learned to generate a logistics and transportation dialogue intention mining model that has completed parameter learning. The parameter learning includes:

[0051] Step S110, processing the dialogue text according to the basic sample logistics and transportation dialogue text and the target text disturbance enhancement features to generate a first semantic feature space and a target reconstructed logistics and transportation dialogue text.

[0052] In this embodiment, the server is responsible for optimizing the parameters of the logistics and transportation dialogue intention mining model using the basic sample logistics and transportation dialogue text.

[0053] Assume that the server stores a series of basic sample logistics transportation dialogue texts, such as:

[0054] Conversation 1: "I want to find out the cost of shipping goods from city A to city B."

[0055] Conversation 2: "My goods are already in the warehouse. When can they be shipped?"

[0056] Conversation 3: "The goods I received are damaged, what should I do?"

[0057] First, the server performs spatial encoding representation based on these basic sample logistics and transportation dialogue texts. Taking dialogue 1 as an example, the server will map each word to a vector space. For example, the word "city A" may be mapped to a specific vector [x_1, y_1, z_1], "city B" is mapped to [x_2, y_2, z_2], and so on. The entire sentence will be represented as a combination of a series of vectors to generate the first temporary semantic feature space. This is like putting each element in the dialogue in a specific coordinate system for subsequent processing.

[0058] Then, the server again performs spatial encoding representation on this first temporary semantic feature space, further refines and adjusts the representation of these vectors, and generates the first semantic feature space. This step is similar to making more precise adjustments to the coordinates after preliminary positioning.

[0059] Next, the server performs spatial encoding again based on the first semantic feature space to generate a second temporary semantic feature space. This is a process of continuously deepening the extraction of text semantic features.

[0060] Assume that the target text perturbation enhancement feature is to simulate some possible disturbances or changes in actual conversations, such as adding some ambiguous words or changing the sentence structure. For example, the target text perturbation enhancement feature for conversation 1 may be "I want to roughly check the cost of transporting goods from city A to city B."

[0061] The server performs spatial encoding representation based on the second temporary semantic feature space and the target text perturbation enhanced features to generate the seventh semantic feature space. This process is like constructing a new semantic representation space after considering the original semantic features and possible interference factors.

[0062] Finally, the server reconstructs the logistics transportation dialogue text based on the seventh semantic feature space to generate the target reconstructed logistics transportation dialogue text. For example, according to the previous operations, for Dialogue 1, it may be reconstructed as "I want to roughly check the goods shipped from City A to City B. How much is the transportation fee?" This reconstructed dialogue text not only retains the main intention of the original dialogue but also includes the changes brought by the target text perturbation enhanced features.

[0063] 2. Taking Dialogue 2 as an example, for the original dialogue "My goods are already in the warehouse. When can they be shipped?" During spatial encoding, the word "goods" is encoded as the vector [a_1, b_1, c_1], "warehouse" is encoded as [a_2, b_2, c_2], etc. After a series of spatial encoding representation operations, considering the target text perturbation enhanced features (such as "My goods have been in the warehouse for some time. When will they be shipped?"), the target reconstructed logistics transportation dialogue text "My goods have been in the warehouse for a while. When can they be shipped out?" is reconstructed.

[0064] 3. For Dialogue 3 "The goods I received are damaged. What should I do?" The same process is carried out according to the above process. During the spatial encoding process, the word "damaged" is given a specific vector representation [m_1, n_1, o_1]. After considering the target text perturbation enhanced features (such as "The goods I received are broken. How to solve it?"), the corresponding target reconstructed logistics transportation dialogue text is reconstructed.

[0065] In step S120, the difference semantic features corresponding to the basic sample logistics transportation dialogue text and the first semantic feature space are fused to generate the second semantic feature space. The difference semantic features are determined by comparing the differences between the basic sample logistics transportation dialogue text and the target reconstructed logistics transportation dialogue text.

[0066] First, calculate the difference semantic features. Taking Dialogue 1 as an example, the original basic sample logistics transportation dialogue text "I want to query the transportation fee of the goods from City A to City B." is compared with the target reconstructed logistics transportation dialogue text "I want to roughly check the goods shipped from City A to City B. How much is the transportation fee?"

[0067] The server will analyze the addition and subtraction of words, subtle semantic changes, etc. For example, although "query" and "roughly check" have similar meanings, there are certain semantic differences, and these differences will be quantified.

[0068] Then, attention analysis is performed based on this difference semantic feature to determine the first attention coefficient information.

[0069] Specifically, the difference semantic features are spatially encoded to generate a third semantic feature space. For example, for the difference between "query" and "check", a new vector representation space may be obtained after spatial encoding, and the vectors in this space reflect the degree of their semantic difference.

[0070] Then, the third semantic feature space is allocated attention, just like allocating different attention levels to different semantically different parts. For example, the difference between "query" and "check it out" may be assigned a higher attention weight because they are changes in the key action words in the sentence. Then, a linear mapping is performed based on the fourth semantic feature space after the attention allocation to generate the first attention coefficient information. This first attention coefficient information includes the first semantic importance factors in multiple dimensions. Assume that in a certain dimension, the semantic importance factor of the word "query" is 0.6, indicating that it is relatively important in this dimension.

[0071] Then determine the second attention coefficient information, the sum of the first semantic importance factor and the second semantic importance factor in the same dimension is equal to 1. So for the previous example, the second semantic importance factor in this dimension is 0.4.

[0072] The fifth semantic feature space is generated by performing spatial deconstruction based on the first semantic feature space. For example, the vectors in the first semantic feature space are decomposed into different sub-parts according to certain rules to facilitate subsequent fusion operations.

[0073] The sixth semantic feature space is generated by fusing the fifth semantic feature space and the second attention coefficient information. Assuming a subvector [p_1, p_2, p_3] in the fifth semantic feature space, it is adjusted according to the weight in the second attention coefficient information to obtain a new vector [q_1, q_2, q_3], which constitutes the sixth semantic feature space.

[0074] Finally, the sixth semantic feature space is enhanced based on the first semantic feature space to generate the second semantic feature space. For example, some feature information in the first semantic feature space is added to the sixth semantic feature space, so that the second semantic feature space contains the information after the difference semantic feature adjustment and retains the important information in the original first semantic feature space.

[0075] For dialogue 2, in the calculation of the difference semantic features between the original dialogue and the target reconstructed dialogue, the difference between "already" and "for a while" is analyzed. After spatial encoding representation, attention allocation and other operations, the first attention coefficient information and the second attention coefficient information are generated, and then a similar fusion operation is performed to generate the second semantic feature space.

[0076] For dialogue 3, such as the difference analysis between “was damaged” and “was broken”, the same process was followed to finally obtain the second semantic feature space corresponding to dialogue 3.

[0077] Step S130, performing conversation intention mining based on the second semantic feature space to generate a conversation intention mining result of the basic sample logistics transportation conversation text.

[0078] Taking dialogue 1 as an example, in the second semantic feature space, the server will mine dialogue intent based on the vector representation in this space and the pre-set classification rules. Assume that there are three possible dialogue intent classifications: querying transportation costs, querying delivery time, and querying cargo damage handling.

[0079] The server analyzes the vectors in the second semantic feature space and calculates the first confidence level that conversation 1 is classified as a query about transportation costs (this is the correct intention annotation data of conversation 1), for example, 0.8, indicating that there is an 80% probability that the intention of this conversation is to query about transportation costs.

[0080] At the same time, the second confidence level classified as querying the delivery time is calculated, for example, 0.1, indicating that there is a 10% probability that the intention of this conversation is to query the delivery time. Querying the delivery time here is to set the conversation intention (no data is labeled for the conversation intention).

[0081] For dialogue 2, in its corresponding second semantic feature space, the server calculates the first confidence level of the query classified as shipping time query (correct intent labeling data), for example, 0.75, and the second confidence level of the query classified as cargo damage handling query is 0.05, etc.

[0082] For dialogue 3, the first confidence level of dialogue classified as inquiry about cargo damage handling (correct intention labeling data) is calculated to be, for example, 0.9, and the second confidence level of dialogue classified as inquiry about transportation costs is calculated to be 0.02, and so on.

[0083] Step S140, optimizing the model parameter information of the logistics and transportation dialogue intention mining model based on the dialogue intention mining results and the dialogue intention annotation data of the basic sample logistics and transportation dialogue text.

[0084] For dialogue 1, the first confidence level of the dialogue intent mining result classified as querying transportation costs (correct intent labeled data) is 0.8.

[0085] Assuming that the set value is 0.9 (this set value can be set according to the needs of the model and experience), for the first basic sample logistics transportation dialogue text, the feature distance is determined according to the first confidence level and the set value. Here, a simple difference calculation can be used, for example, the feature distance is (0.9-0.8=0.1), to generate the first training error sub-parameter.

[0086] For the second confidence level classified as querying the delivery time (setting the conversation intention), which is 0.1, the feature distance is also determined with the set value (assuming that the set value here is 0.05, indicating a lower possibility that this intention should be judged), and the feature distance is (0.1-0.05=0.05), generating the second training error sub-parameter.

[0087] The first training error parameter is determined according to the first training error sub-parameter and the second training error sub-parameter. For example, the first training error parameter can be obtained by adding the two together or adding them according to a certain weight.

[0088] At the same time, the error calculation is performed based on the target reconstruction of the logistics and transportation dialogue text "I want to check the transportation cost of goods from city A to city B" and the basic sample logistics and transportation dialogue text "I want to check the transportation cost of goods from city A to city B" to generate the second training error parameter. The number of word differences, the degree of semantic differences, etc. can be calculated and quantified into a value as the second training error parameter.

[0089] Finally, the model parameter information of the logistics and transportation dialogue intention mining model is optimized based on the first training error parameter and the second training error parameter. For example, if the first training error parameter is large, it means that the model has a large deviation in the confidence judgment of intent classification, and the model may adjust the weight parameters related to intent classification; if the second training error parameter is large, it means that there may be problems in the text reconstruction process, and the model will adjust the parameters related to text processing and reconstruction.

[0090] Follow the same process for Conversation 2 and Conversation 3.

[0091] For dialogue 2, the first training error sub-parameter related to querying the delivery time (correct intent labeling data) and the second training error sub-parameter related to querying the cargo damage treatment (setting dialogue intent) are calculated, and then the first training error parameter is determined. At the same time, the second training error parameters of the target reconstructed dialogue and the original dialogue are calculated, and then the model parameters are optimized.

[0092] For dialogue 3, the error parameters related to querying the handling of damaged goods (correct intent labeling data) and the error parameters related to other set dialogue intentions are similarly calculated. The model parameters are optimized by combining the errors of the target reconstructed dialogue and the original dialogue.

[0093] By continuously using basic sample logistics and transportation dialogue texts to conduct such a parameter learning process, the server can gradually optimize the logistics and transportation dialogue intention mining model, making the model more accurate and efficient in processing logistics and transportation dialogue intention mining.

[0094] Based on the above steps, the embodiment of the present application processes the dialogue text by combining the basic sample logistics and transportation dialogue text with the target text perturbation enhancement features to generate a first semantic feature space and a target reconstructed logistics and transportation dialogue text. Furthermore, by fusing the differential semantic features with the first semantic feature space, a richer second semantic feature space is constructed, which effectively improves the accuracy and comprehensiveness of dialogue intent mining. Finally, dialogue intent mining is performed based on the second semantic feature space, and the model parameters are optimized in combination with the dialogue intent annotation data, which significantly improves the performance and accuracy of the logistics and transportation dialogue intention mining model, providing strong support for intelligent dialogue processing in the field of logistics and transportation.

[0095] In a possible implementation manner, the step of fusing the differential semantic features corresponding to the basic sample logistics transportation dialogue text with the first semantic feature space to generate a second semantic feature space includes:

[0096] Attention analysis is performed based on the difference semantic features corresponding to the basic sample logistics transportation dialogue text to determine first attention coefficient information. The first attention coefficient information includes first semantic importance factors under multiple dimensions.

[0097] The first semantic feature space and the second attention coefficient information are fused to generate a second semantic feature space. The second attention coefficient information includes the second semantic importance factors under the multiple dimensions, and the sum of the first semantic importance factor and the second semantic importance factor under the same dimension is equal to 1.

[0098] In a possible implementation manner, performing attention analysis based on the difference semantic features corresponding to the basic sample logistics transportation dialogue text to determine the first attention coefficient information includes:

[0099] The difference semantic features are spatially encoded to generate a third semantic feature space.

[0100] Attention is allocated to the third semantic feature space to generate a fourth semantic feature space.

[0101] Linear mapping is performed according to the fourth semantic feature space to generate first attention coefficient information.

[0102] In a possible implementation, the fusing the first semantic feature space and the second attention coefficient information to generate the second semantic feature space includes:

[0103] Perform spatial deconstruction based on the first semantic feature space to generate a fifth semantic feature space.

[0104] The fifth semantic feature space and the second attention coefficient information are fused to generate a second semantic feature space.

[0105] In a possible implementation, the fusing the fifth semantic feature space and the second attention coefficient information to generate the second semantic feature space includes:

[0106] The sixth semantic feature space is generated by fusing the fifth semantic feature space and the second attention coefficient information.

[0107] The sixth semantic feature space is feature enhanced according to the first semantic feature space to generate a second semantic feature space.

[0108] In a possible implementation, the processing of the dialogue text according to the basic sample logistics and transportation dialogue text and the target text disturbance enhancement feature to generate the first semantic feature space and the target reconstructed logistics and transportation dialogue text includes:

[0109] The basic sample logistics and transportation dialogue text is spatially encoded and represented to generate a first temporary semantic feature space.

[0110] Perform spatial encoding representation according to the first temporary semantic feature space to generate a first semantic feature space.

[0111] Perform spatial encoding representation according to the first semantic feature space to generate a second temporary semantic feature space.

[0112] A seventh semantic feature space is generated by performing spatial encoding representation based on the second temporary semantic feature space and the target text perturbation enhancement feature.

[0113] The logistics and transportation dialogue text is reconstructed according to the seventh semantic feature space to generate a target reconstructed logistics and transportation dialogue text.

[0114] In a possible implementation, optimizing the model parameter information of the logistics and transportation dialogue intention mining model based on the dialogue intention mining result and the dialogue intention annotation data of the basic sample logistics and transportation dialogue text includes:

[0115] An error calculation is performed based on the dialogue intention mining results and the dialogue intention annotation data of the basic sample logistics and transportation dialogue text to generate a first training error parameter.

[0116] The error is calculated based on the target reconstructed logistics and transportation dialogue text and the basic sample logistics and transportation dialogue text to generate a second training error parameter.

[0117] Based on the first training error parameter and the second training error parameter, the model parameter information of the logistics and transportation dialogue intention mining model is optimized.

[0118] In one possible implementation, the conversation intent mining result includes a first confidence level that the basic sample logistics transportation conversation text is classified as conversation intent annotation data and a second confidence level that the basic sample logistics transportation conversation text is classified as a set conversation intent, and the set conversation intent is not the conversation intent annotation data.

[0119] The step of performing error calculation based on the dialogue intention mining result and the dialogue intention annotation data of the basic sample logistics transportation dialogue text to generate a first training error parameter includes:

[0120] For the X first basic sample logistics transportation dialogue texts, a feature distance is determined according to a first confidence level corresponding to the first basic sample logistics transportation dialogue text and a set value, and a first training error sub-parameter is generated.

[0121] For Y second basic sample logistics and transportation dialogue texts, the feature distance is determined according to the second confidence level corresponding to the second basic sample logistics and transportation dialogue text and the set value, and a second training error sub-parameter is generated.

[0122] A first training error parameter is determined according to the first training error sub-parameter and the second training error sub-parameter.

[0123] In a possible implementation, the method further includes:

[0124] Based on the logistics and transportation dialogue intention mining model, dialogue text processing is performed on the candidate logistics and transportation dialogue text to generate a dialogue intention mining result of the candidate logistics and transportation dialogue text.

[0125] In the above embodiments, the logistics and transportation-based data analysis system for executing the above method embodiments has at least one processor, a control module (chip set) coupled to at least one of the (at least one) processors, a memory coupled to the control module, a non-volatile memory (NVM) / storage device coupled to the control module, at least one input / output device coupled to the control module, and a network interface coupled to the control module.

[0126] The processor may include at least one single-core or multi-core processor, and the processor may include any combination of general-purpose processors or special-purpose processors (such as graphics processors, application processors, baseband processors, etc.). For some alternative implementations, the data analysis system based on logistics transportation can be used as the data analysis system device based on logistics transportation such as the gateway described in the embodiments of this application.

[0127] For some alternative embodiments, the logistics and transportation-based data analysis system may include at least one computer-readable medium (e.g., a memory or NVM / storage device) having instructions and at least one processor integrated with the at least one computer-readable medium and configured to execute the instructions to implement a module to perform the actions described in the present disclosure.

[0128] For one embodiment, the control module may include any suitable interface controller to provide any suitable interface to at least one of the processor(s) and / or any suitable device or component in communication with the control module.

[0129] The control module may include a memory controller module to provide an interface to the memory. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0130] The memory may be used, for example, to load and store data and / or instructions for a data analysis system based on logistics transportation. For one embodiment, the memory may include any suitable volatile memory, for example, a suitable DRAM.

[0131] For one embodiment, the control module may include at least one input / output controller to provide an interface to the NVM / storage device and (at least one) input / output device.

[0132] For example, the NVM / storage device may be used to store data and / or instructions. The NVM / storage device may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (at least one) non-volatile storage device (e.g., at least one hard disk drive (HDD), at least one compact disk (CD) drive, and / or at least one digital versatile disk (DVD) drive).

[0133] The NVM / storage device may include storage resources that are physically part of the device on which the logistics-based data analysis system is installed, or it may be accessible to the device without being part of the device. For example, the NVM / storage device may be accessed via (at least one) input / output device based on a network.

[0134] (At least one) input / output device may provide an interface for the data analysis system based on logistics and transportation to communicate with any other appropriate device, and the input / output device may include a communication component, a phonetic component, a sensor component, etc. The network interface may provide an interface for the data analysis system based on logistics and transportation to predict network communication based on at least one user preference, and the data analysis system based on logistics and transportation may wirelessly communicate with at least one component of a wireless network based on any of at least one wireless network priors and / or protocols, such as accessing a wireless network based on a communication prior.

[0135] For one embodiment, at least one of the (at least one) processor may be loaded together with the logic of at least one controller of the control module (e.g., a memory controller module). For one embodiment, at least one of the (at least one) processor may be loaded together with the logic of at least one controller of the control module to form a system level load. For one embodiment, at least one of the (at least one) processor may be fused on the same die with the logic of at least one controller of the control module. For one embodiment, at least one of the (at least one) processor may be fused on the same die with the logic of at least one controller of the control module to form a system on chip (SoC).

[0136] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0137] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the data analysis method based on logistics and transportation described in the aforementioned embodiment.

[0138] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps in the data analysis method based on logistics and transportation described in the aforementioned embodiment.

[0139] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple user preference prediction network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0140] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, a magnetic disk storage, a magnetic tape storage, or any other medium that can be used to be computer-readable with or store data.

[0141] Finally, it should be noted that what is disclosed above is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A data analysis method based on logistics transportation, characterized in that: The method comprises: Based on the basic sample logistics and transportation dialogue text, the logistics and transportation dialogue intention mining model is parameter-learned to generate a logistics and transportation dialogue intention mining model that has completed parameter learning; wherein the parameter learning includes: Processing the dialogue text according to the basic sample logistics and transportation dialogue text and the target text disturbance enhancement features to generate a first semantic feature space and a target reconstructed logistics and transportation dialogue text; A second semantic feature space is generated by fusing the difference semantic features corresponding to the basic sample logistics and transportation dialogue text and the first semantic feature space; the difference semantic features are determined by comparing the difference between the basic sample logistics and transportation dialogue text and the target reconstructed logistics and transportation dialogue text; Conducting conversation intention mining based on the second semantic feature space to generate conversation intention mining results for the basic sample logistics transportation conversation text; Based on the conversation intention mining results and the conversation intention annotation data of the basic sample logistics and transportation conversation text, the model parameter information of the logistics and transportation conversation intention mining model is optimized.

2. The data analysis method based on logistics transportation according to claim 1 is characterized in that: The step of fusing the difference semantic features corresponding to the basic sample logistics transportation dialogue text with the first semantic feature space to generate a second semantic feature space includes: Performing attention analysis based on the differential semantic features corresponding to the basic sample logistics transportation dialogue text to determine first attention coefficient information; the first attention coefficient information includes first semantic importance factors under multiple dimensions; The second semantic feature space is generated by fusing the first semantic feature space and the second attention coefficient information; the second attention coefficient information includes the second semantic importance factors under the multiple dimensions, and the sum of the first semantic importance factor and the second semantic importance factor under the same dimension is equal to 1.

3. The data analysis method based on logistics transportation according to claim 2 is characterized in that: The performing attention analysis based on the difference semantic features corresponding to the basic sample logistics transportation dialogue text to determine the first attention coefficient information includes: Performing spatial encoding representation on the difference semantic features to generate a third semantic feature space; Performing attention allocation on the third semantic feature space to generate a fourth semantic feature space; Linear mapping is performed according to the fourth semantic feature space to generate first attention coefficient information.

4. The data analysis method based on logistics transportation according to claim 2 is characterized in that: The step of fusing the first semantic feature space and the second attention coefficient information to generate a second semantic feature space includes: Performing spatial deconstruction based on the first semantic feature space to generate a fifth semantic feature space; The fifth semantic feature space and the second attention coefficient information are fused to generate a second semantic feature space.

5. The data analysis method based on logistics transportation according to claim 4 is characterized in that: The step of fusing the fifth semantic feature space and the second attention coefficient information to generate a second semantic feature space includes: Generating a sixth semantic feature space by fusing the fifth semantic feature space and the second attention coefficient information; The sixth semantic feature space is feature enhanced according to the first semantic feature space to generate a second semantic feature space.

6. The data analysis method based on logistics transportation according to claim 1 is characterized in that: The processing of the dialogue text according to the basic sample logistics and transportation dialogue text and the target text disturbance enhancement feature to generate a first semantic feature space and a target reconstructed logistics and transportation dialogue text includes: Performing spatial encoding representation on the basic sample logistics transportation dialogue text to generate a first temporary semantic feature space; Performing spatial encoding representation according to the first temporary semantic feature space to generate a first semantic feature space; Performing spatial encoding representation according to the first semantic feature space to generate a second temporary semantic feature space; Performing spatial encoding representation according to the second temporary semantic feature space and the target text disturbance enhancement feature to generate a seventh semantic feature space; The logistics and transportation dialogue text is reconstructed according to the seventh semantic feature space to generate a target reconstructed logistics and transportation dialogue text.

7. The data analysis method based on logistics transportation according to claim 1 is characterized in that: The optimizing the model parameter information of the logistics and transportation dialogue intention mining model based on the dialogue intention mining result and the dialogue intention annotation data of the basic sample logistics and transportation dialogue text includes: Perform error calculation based on the conversation intention mining result and the conversation intention annotation data of the basic sample logistics transportation conversation text to generate a first training error parameter; Perform error calculation based on the target reconstructed logistics and transportation dialogue text and the basic sample logistics and transportation dialogue text to generate a second training error parameter; Based on the first training error parameter and the second training error parameter, the model parameter information of the logistics and transportation dialogue intention mining model is optimized.

8. The data analysis method based on logistics transportation according to claim 7 is characterized in that: The conversation intention mining result includes a first confidence level that the basic sample logistics transportation conversation text is classified as conversation intention annotation data and a second confidence level that the basic sample logistics transportation conversation text is classified as a set conversation intention, and the set conversation intention is not the conversation intention annotation data; The step of performing error calculation based on the dialogue intention mining result and the dialogue intention annotation data of the basic sample logistics transportation dialogue text to generate a first training error parameter includes: For the X first basic sample logistics transportation dialogue texts, determine the feature distance according to the first confidence level corresponding to the first basic sample logistics transportation dialogue text and the set value, and generate a first training error sub-parameter; For Y second basic sample logistics transportation dialogue texts, determine the feature distance according to the second confidence level corresponding to the second basic sample logistics transportation dialogue text and the set value, and generate a second training error sub-parameter; A first training error parameter is determined according to the first training error sub-parameter and the second training error sub-parameter.

9. The data analysis method based on logistics transportation according to claim 1 is characterized in that: The method further comprises: Based on the logistics and transportation dialogue intention mining model, dialogue text processing is performed on the candidate logistics and transportation dialogue text to generate a dialogue intention mining result of the candidate logistics and transportation dialogue text.

10. A data analysis system based on logistics transportation, characterized in that: It includes a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by a computer, the data analysis method based on logistics transportation described in any one of claims 1 to 9 is implemented.

Images