Model generation method, information generation method, electronic device, and storage medium

Abstract

The application is suitable for the computer technical field, and provides a model generation method, an information generation method, an electronic device and a storage medium, wherein the method comprises the following steps: obtaining vehicle comment data, and extracting defeat relationship information from the vehicle comment data, wherein the defeat relationship information comprises a main competitive vehicle model, other competitive vehicle models and an attention factor concerned by the vehicle comment data; based on the defeat relationship information, performing labeling processing on the vehicle comment data to obtain labeled result data, and storing the vehicle comment data and the labeled result data as training samples in a training sample set; taking the vehicle comment data of the training sample in the training sample set as input, taking the labeled result data corresponding to the input vehicle comment data as expected output, and training to obtain a defeat relationship analysis model. The application can generate a defeat relationship analysis model for analyzing defeat relationship information in vehicle comment data.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and in particular relates to a model generation method, an information generation method, an electronic device, and a storage medium. Background Technology

[0002] In recent years, with the advent of the intelligent technology wave, automobile consumption has steadily increased, and the automotive industry has experienced rapid development. Now, more and more companies are entering this market. Major companies are vying for dominance, pulling out all the stops to improve their intelligent and service levels, offering customers an increasingly wider range of choices.

[0003] With the continuous emergence of new generations of vehicles, people are increasingly seeking information and sharing their experiences with automobiles. In practice, many car owners typically post reviews of various car models on different automotive media platforms, such as short video platforms, car forums, and automotive news outlets. Users, such as potential car owners, can learn about the competitive advantages and disadvantages of different models by viewing these reviews on automotive media platforms.

[0004] In related technologies, users typically need to read the entire vehicle review to understand the competitive advantages and disadvantages of the various types of vehicles described in the review, which requires a lot of time and effort and results in a poor user experience. Summary of the Invention

[0005] This application provides a model generation method, an information generation method, an electronic device, and a storage medium, aiming to solve the problem in related technologies that users usually need to read the entire vehicle review to understand the competitive advantages and disadvantages of various types of vehicles described in the review, which requires a lot of time and effort and results in a low user experience.

[0006] In a first aspect, embodiments of this application provide a model generation method, the method comprising:

[0007] Acquire vehicle review data and extract defeat relationship information from it. This defeat relationship information includes the main competitor's model, other competitor models, and the factors of interest highlighted in the vehicle review data.

[0008] Based on the information about the defeat relationship, the vehicle comment data is labeled to obtain the labeled result data, and the vehicle comment data and the labeled result data are stored as training samples in the training sample set;

[0009] The vehicle review data of the training samples in the training sample set is used as input, and the labeled result data corresponding to the input vehicle review data is used as the expected output to train the defeat relationship analysis model. The defeat relationship analysis model is used to analyze the defeat relationship information in the vehicle review data.

[0010] Secondly, embodiments of this application provide an information generation method, the method comprising:

[0011] Upon receiving the target vehicle review data, the target vehicle review data is input into the defeat relationship analysis model to obtain defeat relationship information for the target vehicle review data. The defeat relationship analysis model is generated using any of the above model generation methods.

[0012] Output the defeat relationship information for the target vehicle's review data.

[0013] Thirdly, embodiments of this application provide a model generation apparatus, including:

[0014] The data acquisition unit is used to acquire vehicle review data and extract defeat relationship information from the vehicle review data. The defeat relationship information includes the main competitor model, other competitor models, and the factors of interest that the vehicle review data focuses on.

[0015] The sample generation unit is used to annotate vehicle review data based on defeat relationship information to obtain annotated result data, and to store vehicle review data and annotated result data as training samples into the training sample set;

[0016] The model training unit is used to take the vehicle review data of the training samples in the training sample set as input and the labeled result data corresponding to the input vehicle review data as the expected output to train the defeat relationship analysis model. The defeat relationship analysis model is used to analyze the defeat relationship information in the vehicle review data.

[0017] Fourthly, embodiments of this application provide an information generation apparatus, including:

[0018] The data receiving unit is used to input the target vehicle review data into the defeat relationship analysis model when it receives the target vehicle review data, so as to obtain the defeat relationship information for the target vehicle review data. The defeat relationship analysis model is generated by any of the above-mentioned model generation methods.

[0019] The information output unit is used to output the defeat relationship information for the target vehicle review data.

[0020] Fifthly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any of the above-described model generation methods or the steps of any of the above-described information generation methods.

[0021] Sixthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described model generation methods or the steps of any of the above-described information generation methods.

[0022] In a seventh aspect, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to execute any of the above-described model generation methods or any of the above-described information generation methods.

[0023] The beneficial effects of this application embodiment compared with related technologies are: it can generate a defeat relationship analysis model for analyzing defeat relationship information in vehicle review data, which helps to obtain defeat relationship information in vehicle review data based on the defeat relationship analysis model. That is, it helps to quickly extract the advantages and disadvantages of various types of cars described in vehicle review data. Compared with related technologies, where users need to read the entire vehicle review data to understand the competitive advantages and disadvantages of various types of cars described in the vehicle review data, it can greatly improve the user experience.

[0024] It is understood that the beneficial effects of the second to seventh aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic flowchart of a model generation method provided in an embodiment of this application;

[0027] Figure 2 This is a flowchart illustrating an information generation method provided in an embodiment of this application;

[0028] Figure 3 This is a flowchart illustrating the implementation process from model generation to model application, provided in an embodiment of this application.

[0029] Figure 4 This is a schematic diagram of the structure of the model generation apparatus provided in the embodiments of this application;

[0030] Figure 5 This is a schematic diagram of the structure of the information generation device provided in the embodiments of this application;

[0031] Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0032] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0033] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0034] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0035] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0036] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0037] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0038] To illustrate the technical solution of this application, the following embodiments will be used for explanation.

[0039] Example 1

[0040] Please see Figure 1 This application provides a model generation method, such as... Figure 1 As shown, the model generation method may include the following steps 101-103.

[0041] Step 101: Obtain vehicle review data and extract defeat relationship information from the vehicle review data.

[0042] Among them, the information on the relationship between the defeated and the defeated includes the main competitor's model, other competitor's models, and the factors of interest reflected in vehicle review data.

[0043] The aforementioned main competitor models are typically the most competitive models reviewed in the vehicle review data. The aforementioned other competitor models are usually models reviewed in the vehicle review data other than the main competitor models. In some application scenarios, the main competitor model is often an alternative model that the reviewer has not yet purchased, while the other competitor models are usually models used for comparison that the reviewer has not yet purchased. The aforementioned factors of interest are typically those considered in the vehicle review data. In practice, these factors of interest can indicate why a particular model is favored by users. In practical applications, these factors of interest may include interior design, exterior appearance, powertrain, and handling.

[0044] The vehicle review data can be in the form of text, audio, video, etc. When analyzing vehicle review data, it is usually advisable to first convert non-textual vehicle review data into text format. This facilitates analysis and improves efficiency.

[0045] In this embodiment, the execution entity of the above-described model generation method is typically an electronic device. It should be noted that the electronic device can be hardware or software. When the electronic device is hardware, it can be implemented as a cluster of multiple devices or as a single device. When the electronic device is software, it can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are imposed here.

[0046] In practice, web crawler technology can be used to scrape vehicle review data from various automotive media platforms. This scraped data can be stored locally or on other electronic devices connected to the executing entity. When the vehicle review data is stored locally, the executing entity can directly retrieve and process it. When the vehicle review data is stored on other electronic devices connected to the executing entity, the executing entity can obtain and process the data via wired or wireless connections.

[0047] After obtaining vehicle review data, the aforementioned entity can extract defeat / loss relationship information from the vehicle review data. For example, the vehicle review data may directly contain variables for the main competitor model, other competitor models, and factors of interest. The entity can directly extract the variable values ​​corresponding to these three variables to extract defeat / loss relationship information from the vehicle review data.

[0048] It should be noted that the vehicle review data obtained in step 101 is usually multiple. For each vehicle review data, the aforementioned executing entity can extract the defeat relationship information from the vehicle review data.

[0049] In some alternative implementations, after obtaining the vehicle review data in step 101 above, the following steps are usually included: preprocessing the vehicle review data and switching the vehicle review data to the preprocessed vehicle review data.

[0050] The aforementioned data preprocessing may include, but is not limited to, data cleaning, data processing, and data structuring.

[0051] In this embodiment, after acquiring the vehicle review data and before applying it, the vehicle review data is preprocessed to improve data quality and ensure that subsequent data processing and analysis are more stable and reliable.

[0052] Step 102: Based on the defeat relationship information, the vehicle comment data is labeled to obtain the labeled result data, and the vehicle comment data and the labeled result data are stored as training samples in the training sample set.

[0053] Here, the aforementioned implementing entity can use the defeat / victory relationship information extracted from the vehicle review data to annotate the vehicle review data, thereby obtaining the annotated result data. When directly using the defeat / victory relationship information to annotate the vehicle review data, the resulting annotated result data includes the defeat / victory relationship information.

[0054] Subsequently, the aforementioned implementing entity can use the vehicle review data and its corresponding annotation results as a training sample, and store this training sample in a training sample set. In practice, one training sample can be obtained for each piece of vehicle review data. When there are many pieces of vehicle review data, multiple training samples can be obtained. Each training sample includes the corresponding vehicle review data and annotation results.

[0055] It should be noted that, by default, the above training sample set can be empty.

[0056] Step 103: Take the vehicle review data of the training samples in the training sample set as input, and take the labeled result data corresponding to the input vehicle review data as the expected output to train the defeat relationship analysis model.

[0057] Among them, the defeat relationship analysis model is used to analyze defeat relationship information in vehicle review data.

[0058] Here, the aforementioned implementing entity can use a training sample set to train the initial deep learning model, thereby obtaining the aforementioned defeat relationship analysis model.

[0059] The initial deep learning model mentioned above can be an untrained deep learning model or a deeply learning model that has not been fully trained. Each layer of the initialized deep learning model can have initial parameters, which can be continuously adjusted during the training process. The initial deep learning model can be various types of untrained or under-trained artificial neural networks, or a model obtained by combining multiple untrained or under-trained artificial neural networks. For example, the initial deep learning model can be an untrained convolutional neural network, an untrained recurrent neural network, or a model obtained by combining untrained convolutional neural networks, untrained recurrent neural networks, and untrained fully connected layers. In practice, the initial deep learning model mentioned above is usually a Bidirectional Encoder Representations from Transformer (BERT) model. When applying the BERT model to this application, the labels of the CRF layers of the BERT model are redefined. For example, the BA label can indicate the main competitor model, and the BM label can indicate other competitor models. For example, for the BA-Tank 300, it can be known that the Tank 300 model is the main competitor model. And regarding the BM-Tank 500, it can be inferred that the Tank 500 model is a competitor of other models.

[0060] In practice, the aforementioned execution entity can input vehicle review data from the training samples into the input side of the deep learning model, process the parameters of each layer in the deep learning model in sequence, and output the data from the output side of the deep learning model. The information output from the output side is the labeled result data.

[0061] The model generation method provided in this embodiment can generate a defeat relationship analysis model for analyzing defeat relationship information in vehicle review data. This helps to analyze the defeat relationship information in vehicle review data based on the defeat relationship analysis model. In other words, it helps to quickly extract the advantages and disadvantages of various types of cars described in the vehicle review data. Compared with related technologies where users need to read the entire vehicle review data to understand the competitive advantages and disadvantages of various types of cars described in the vehicle review data, this can greatly improve the user experience.

[0062] In some optional implementations of this embodiment, step 101 above, in which the information on the defeat relationship is extracted from the vehicle review data, may include: extracting the vehicle feature datasets corresponding to each vehicle model from the vehicle review data, and determining the main competitor model and other competitor models in the vehicle review data based on the vehicle feature datasets corresponding to each vehicle model.

[0063] The vehicle feature data in the aforementioned vehicle feature dataset typically consists of parameters corresponding to the vehicle model. In practice, vehicle feature data may include body parameters and other feature parameters. For example, vehicle feature data in the dataset may include, but is not limited to, parameters such as length, width, height, wheelbase, appearance, and power.

[0064] In some application scenarios, when the vehicle review data includes parameter variables for each vehicle model, the aforementioned execution entity can directly extract the parameter variables and their values ​​for the corresponding vehicle model, thereby obtaining the vehicle feature dataset for that vehicle model.

[0065] In some application scenarios, since the vehicle feature dataset for a specific car model in vehicle review data is usually located immediately after that model, the aforementioned execution entity can extract the vehicle feature dataset for each car model in the following way: First, identify the information indicating the car model from the corresponding text in the vehicle review data, such as "Tank 300" indicating the Tank 300 model; then, continue to identify keywords indicating parameter names after this information, such as "wheelbase"; subsequently, the data following this keyword can be used as the value of the corresponding parameter. For example, the number identified after "wheelbase" can be used as the value of the wheelbase parameter. Similarly, for other parameters, such as appearance, the adjective "very stylish" identified after "appearance" can be used as the value of the appearance parameter.

[0066] Here, after extracting the vehicle feature datasets corresponding to each model from the vehicle review data, the aforementioned execution entity can use the respective vehicle feature datasets to determine the main competitor model and other competitor models in the vehicle review data. In practice, since the features described for a preferred model are usually numerous, the aforementioned execution entity can determine the model corresponding to the vehicle feature dataset with the most corresponding model feature data as the main competitor model, and determine the models corresponding to other model feature datasets as other competitor models. For example, if model A's corresponding model feature dataset 1 has 10 model feature data, model B's corresponding model feature dataset 2 has 5 model feature data, and model C's corresponding model feature dataset 3 has 1 model feature data, then model A corresponding to model feature dataset 1 can be determined as the main competitor model, and models B and C can be determined as other competitor models.

[0067] This embodiment can accurately extract the main competitor model and other competing models from vehicle review data.

[0068] In some optional implementations, the process of determining the main competitor model and other competitor models in the vehicle review data based on the vehicle feature datasets corresponding to each vehicle model can include: when the vehicle review data includes a target model, determining the relevance between the corresponding model and the target model based on the vehicle feature datasets corresponding to each model; identifying models whose relevance meets preset relevance conditions as main competitor models; and identifying models other than the main competitor models as other competitor models. Other models refer to models in the vehicle review data other than the main competitor model and the target model.

[0069] The target vehicle model mentioned above can be the model purchased by the person sharing the vehicle review data, or the model that the person sharing the data indicates is most likely to purchase. In practice, the vehicle review data may contain a target vehicle model variable, and the executing entity can directly extract the variable value corresponding to the target vehicle model variable to extract the target vehicle model from the vehicle review data.

[0070] In some application scenarios, vehicle review data can be generated based on a pre-defined template, which may include an input field for users to specify their desired vehicle model. Thus, when retrieving the target vehicle model option, the executing entity can directly locate the recorded target vehicle model based on the template's format and extract it.

[0071] Here, when the vehicle review data includes the target vehicle model, the aforementioned execution entity can calculate the correlation between each vehicle model feature dataset and the target vehicle model. For example, for each vehicle model, the correlation between the vehicle model feature dataset corresponding to that model and the vehicle model feature dataset corresponding to the target vehicle model can be calculated to obtain the correlation for that vehicle model. In practice, the aforementioned execution entity can use the Minimum Redundancy Maximum Relevance (MRMR) algorithm to calculate the correlation for each vehicle model.

[0072] Subsequently, the aforementioned implementing entity can identify models that meet the preset relevant conditions as the main competitor models, and identify models other than the main competitor models as other competitor models.

[0073] The aforementioned preset relevant conditions are usually pre-defined filtering criteria. For example, preset relevant conditions could be used to filter for the car models with the highest relevance, or to filter for the car models with the second highest relevance.

[0074] In practice, the model with the highest correlation is usually taken as the main competitor model, and other models are taken as other competitors models.

[0075] This embodiment can accurately extract the main competitor model and other competitor models when the vehicle review data includes the target model.

[0076] In practice, for each vehicle model, if the corresponding vehicle feature dataset S is {x1, x2, ... x...} n}, where n is the total number of feature data in the vehicle model feature dataset S. If the vehicle model feature dataset corresponding to the target vehicle model is c, then when using the MRMR algorithm to calculate the relevance of that vehicle model, the relevance of that vehicle model can be obtained as:

[0077]

[0078] Where I(x) i c) represents the mutual information value between the i-th feature in the vehicle feature dataset S and the vehicle feature dataset c. After obtaining the correlation degree for each vehicle model, the vehicle model with the highest correlation degree can be identified as the main competitor model, and the other models can be identified as other competitor models.

[0079] In practical applications, the aforementioned execution entity can also use MRMR to remove redundancy from the feature data in the vehicle feature datasets corresponding to each vehicle model. Specifically, m feature data can be selected from the vehicle feature dataset S in descending order of mutual information value to obtain a vehicle feature dataset S1 containing m feature data. Then, several highly correlated feature data are removed from S1 to obtain a new vehicle feature dataset with the lowest total correlation after redundancy removal. The specific filtering process is shown in the following formula:

[0080]

[0081] Where I(x) i ,x j Let be the mutual information value between the i-th feature data and the j-th feature data, where i ≠ j, i ≤ m, and j ≤ m.

[0082] It should be noted that by using MRMR to remove redundancy from the feature data in the vehicle feature datasets corresponding to each vehicle model, a simplified vehicle feature dataset can be obtained. Using this simplified dataset to annotate the corresponding vehicle review data reduces the amount of information to be labeled, thereby improving annotation efficiency.

[0083] In some optional implementations of this embodiment, step 101 above, in which the information on the defeat relationship is extracted from the vehicle review data, may include: extracting multiple target keywords from the vehicle review data, selecting a preset number of target keywords from the multiple target keywords in descending order of frequency of occurrence according to the frequency of occurrence of each target keyword in the vehicle review data, and determining the selected target keywords as factors of interest.

[0084] The target keywords mentioned above are usually pre-defined keywords, such as interior design, powertrain, operability, and exterior design. The preset number mentioned above is usually a pre-defined value, such as 5.

[0085] In practice, the aforementioned implementing entity extracts keywords from the text corresponding to vehicle review data, thereby obtaining multiple target keywords. For example, three target keywords can be extracted: interior, powertrain, and exterior. Then, the implementing entity can select a preset number of target keywords from the extracted keywords, in descending order of frequency, and use these selected target keywords as focus factors. For instance, if the preset number is two, and three target keywords are extracted (interior, powertrain, and exterior), with interior appearing at 10%, powertrain at 5%, and exterior at 2%, then the focus factors could be interior and powertrain.

[0086] This embodiment can accurately analyze the factors of interest that sharers of vehicle review data focus on. These factors are likely the main reasons that prompt the sharer to purchase the target model or a main competitor's model. It should be noted that accurately analyzing the factors of interest corresponding to vehicle review data can assist vehicle manufacturers in making targeted improvements to certain vehicle models.

[0087] In some optional implementations of this embodiment, the extraction of defeat relationship information from vehicle review data may further include: performing topic clustering analysis on the vehicle review data to obtain multiple topics and corresponding vocabulary sets for each topic; selecting a preset number of vocabulary sets in descending order of vocabulary quantity based on the vocabulary sets corresponding to each topic; and determining the topics corresponding to the selected vocabulary sets as factors of interest.

[0088] Here, the aforementioned implementing entity can employ a topic clustering model to perform topic clustering analysis on vehicle review data, thereby obtaining multiple topics and a vocabulary set corresponding to each topic. The vocabulary set corresponding to a topic is typically a collection of words belonging to that topic. In practice, the aforementioned topic clustering model can be a document topic generation model (Latent Dirichlet Allocation, LDA). In practice, when performing topic clustering analysis on vehicle review data, the topics are usually pre-defined; for example, five topics can be pre-defined: exterior, powertrain, interior, operability, and comfort.

[0089] After obtaining multiple topics and their corresponding vocabulary sets, the aforementioned execution entity can select topics corresponding to vocabulary sets with a large number of words as factors of interest. This embodiment can extract factors of interest from vehicle review data through topic clustering.

[0090] In some optional implementations of this embodiment, when the vehicle review data includes the target model, the annotation result data includes the target model, the vehicle series to which the target model belongs, and the defeat / defeat relationship information.

[0091] Here, when the vehicle review data includes the target model, the vehicle model purchased, the target model's series, and the aforementioned defeat / loss relationship information can be used to label the vehicle review data together. This results in labeled data that includes the target model, its series, and the defeat / loss relationship information. It's important to note that including the target model, its series, and defeat / loss relationship information in the labeled data allows the trained defeat / loss relationship analysis model to simultaneously analyze both the defeat / loss relationship information and the target model from the vehicle review data. In other words, this helps the model extract more comprehensive data from the vehicle review data, further improving the user experience.

[0092] Example 2

[0093] Figure 2 This is a flowchart illustrating an information generation method provided in an embodiment of this application. Figure 2 As shown, the information generation method may include the following steps 201-202.

[0094] Step 201: Upon receiving the target vehicle review data, input the target vehicle review data into the defeat relationship analysis model to obtain defeat relationship information for the target vehicle review data.

[0095] The defeat relationship analysis model can be generated using the model generation method described in Example 1. This model is used to analyze defeat relationship information in the target vehicle review data.

[0096] Among them, the information on the relationship between the defeated and the defeated includes the factors of interest reflected in the review data of the main competitor model, other competitor models, and the target vehicle.

[0097] The aforementioned primary competitor models are typically the most competitive models reviewed in the target vehicle review data. The aforementioned other competitor models are typically models reviewed in the target vehicle review data other than the primary competitor models. In some application scenarios, the primary competitor model is usually an alternative model that the reviewer has not yet purchased, while the other competitor models are usually models used for comparison that the reviewer has not yet purchased. The aforementioned factors of interest are typically those considered in the target vehicle review data. In practice, these factors of interest can indicate why a particular model is favored by users. In practical applications, these factors of interest may include interior design, exterior appearance, powertrain, and handling.

[0098] In this embodiment, the executing entity of the above-described information generation method can be an electronic device. It should be noted that the executing entity of the above-described information generation method can be the same as or different from the executing entity of the above-described model generation method.

[0099] Here, the executing entity can receive target vehicle review data. This target vehicle review data can be user-inputted vehicle review data, or vehicle review data sent from the user terminal or other terminals.

[0100] Here, upon receiving the target vehicle review data, the executing entity can input the target vehicle review data into the defeat relationship analysis model, thereby obtaining the defeat relationship information in the target vehicle review data output by the defeat relationship analysis model.

[0101] Step 202: Output the defeat relationship information for the target vehicle review data.

[0102] After obtaining the battle-response relationship information from the target vehicle review data, the executing entity can output this information. For example, it can be presented directly to the user or sent to the user's terminal. This allows the user to quickly learn about the battle-response relationship information of the target vehicle review data, greatly improving the user experience.

[0103] In some application scenarios, when a user is reading vehicle review data on an automotive media platform, if the system detects that the user has performed an action to trigger a battle-or-defeat relationship analysis of that vehicle review data, the executing entity can perform steps 201-202 as described above on the vehicle review data. This allows the system to obtain the battle-or-defeat relationship information of the vehicle review data and present this information near the vehicle review data, for example, in the position where the vehicle review data is being read. This helps users quickly understand the battle-or-defeat relationship information of the vehicle review data, greatly improving the user experience.

[0104] The information generation method provided in this embodiment can quickly analyze the defeat relationship information in vehicle review data using a pre-generated defeat relationship analysis model. In other words, it helps to quickly extract the advantages and disadvantages of various types of cars described in the vehicle review data. Compared with related technologies where users need to read the entire vehicle review data to understand the competitive advantages and disadvantages of various types of cars described in the vehicle review data, it can greatly improve the user experience.

[0105] In some optional implementations of this embodiment, the information generation method described above may further include the following steps: when the target vehicle review data includes the target model, the target model, the vehicle series to which the target model belongs, and the information on the relationship between the target and untargeted vehicles are stored as associated information. Based on the associated information corresponding to each target model, the main competing models, other competing models, and factors of interest for the corresponding target model are determined; and based on the main competing models, other competing models, and factors of interest for the corresponding target model, the competitiveness of the corresponding target model is determined.

[0106] The target models mentioned above are usually those purchased by the people who share vehicle review data.

[0107] Here, when the vehicle review data includes the target model, the aforementioned implementing entity can store the target model, the vehicle series to which the target model belongs, and the defeat / defeat relationship information as associated information. Each associated information includes the corresponding target model, the vehicle series to which the target model belongs, and the defeat / defeat relationship information.

[0108] In practice, the implementing entity can use the stored related information to conduct statistical analysis.

[0109] Here, for each target vehicle model, all associated information corresponding to that target vehicle model can be extracted from the stored associated information, resulting in multiple associated information sets. Since each associated information set contains information about competitive relationships (i.e., main competitor models, other competitor models, and factors of interest), the executing entity can sort the main competitor models in each associated information set corresponding to the target vehicle model according to their frequency of occurrence from highest to lowest, selecting several main competitor models, such as five, and then using these selected main competitor models as the main competitor models for that target vehicle model. To give a further example, if there are 30 related information entries for target model A, and if 6 of these entries are for competitor model B, 10 for competitor model C, 2 for competitor model D, 3 for competitor model E, 4 for competitor model F, and 5 for competitor model G, then B appears 6 times, C appears 6 times, D appears 2 times, E appears 3 times, F appears 4 times, and G appears 5 times. If we need to select 3 competitors for the target model, we can select B, C, and G in descending order of frequency. In other words, the competitors for target model A are B, C, and G. In some application scenarios, after obtaining the main competing models of the target vehicle, the main competing models can be scored based on the frequency of their appearance. Specifically, the more frequently a main competing model appears, the higher its score. This helps users to understand the situation of various models more intuitively.

[0110] Here, the methods for obtaining information on other competing models of the target vehicle and the factors of interest for the target vehicle are basically the same as those for obtaining information on the main competing models of the target vehicle, and will not be elaborated upon here.

[0111] In this embodiment, for each target vehicle model, after obtaining the main competing models, other competing models, and factors of interest, the competitiveness of the target vehicle model can be calculated by using at least one of the main competing models, other competing models, and factors of interest.

[0112] In some implementations, the competitiveness of a target model can be determined based on one of the following: the main competing models, other competing models, or a factor of interest. For example, the competitiveness of a target model can be determined using factors of interest. As an example, the competitiveness of a target model can be calculated by considering each factor of interest, the frequency of occurrence of each factor, and the weight of each factor. For instance, if target model A has factors 1, 2, and 3, and factor 1 appears 20 times with a weight of 0.2, factor 2 appears 50 times with a weight of 0.1, and factor 3 appears 30 times with a weight of 0.5, then the competitiveness of target model A can be calculated as 24, where 24 = 20 × 0.2 + 50 × 0.1 + 30 × 0.5. Here, the weights of each factor can be pre-set.

[0113] In some implementations, the competitiveness of a target vehicle can be determined based on three factors: the main competing models, other competing models, and the factors of interest. For example, the competitiveness of the target vehicle can be calculated by summing the frequency of occurrence of each main competing model, each other competing model, and each factor of interest. To give a further example, if target model A has two main competing models, model B and model C, with model B appearing 20 times and model C appearing 50 times, and target model A has three other competing models, model D, model E, and model F, with model D appearing 10 times, model E appearing 5 times, and model F appearing once, and target model A has three factors of interest, factor 1, factor 2, and factor 3, with factor 1 appearing 20 times, factor 2 appearing 50 times, and factor 3 appearing 30 times, then the competitiveness of target model A can be calculated to be 176, where 176 = (20 + 50) + (10 + 5 + 1) + (20 + 50 + 30).

[0114] In some application scenarios, the main competing models, other competing models, and factors of interest for each car series can be identified based on their respective related information. Furthermore, the competitiveness of a given car series can be determined based on these main competing models, other competing models, and factors of interest. The method for determining the competitiveness of a car series is essentially the same as the method for determining the competitiveness of a target car model, and will not be elaborated upon here.

[0115] In some application scenarios, the main competing models, other competing models, and factors of interest for each vehicle type can be displayed visually. The competitiveness of each vehicle type can also be displayed visually. This helps users to understand the situation of various vehicle types more intuitively and further improves the user experience.

[0116] Example 3

[0117] Figure 3 This is a flowchart illustrating the implementation process from model generation to model application, provided as an embodiment of this application. Figure 3 As shown, the process may include steps 301 to 308. The entity performing steps 301 to 308 is typically an electronic device.

[0118] Step 301: Obtain vehicle review data from various automotive media platforms.

[0119] Step 302: Clean and structure the acquired vehicle review data.

[0120] Step 303: Train the defeat relationship analysis model.

[0121] Here, the defeat relationship analysis model can be obtained by training through the following steps 3031 to 3033.

[0122] Step 3031: Label the comment data for each vehicle to obtain the labeled data.

[0123] The labeled data may include the user's target car model, the car series to which the target model belongs, the main competing models, other competing models, and the reason for purchase. Here, the reason for purchase refers to the factors of interest mentioned above.

[0124] Step 3032: Train the initial deep learning model using the labeled vehicle comment data.

[0125] Step 3033: Use the labeled vehicle review data to evaluate and optimize the intermediate model obtained from training.

[0126] In practice, setting the batch size to 2k and the number of training steps to 15k can achieve faster model convergence and better predictive performance. This, in turn, helps to obtain a more stable and reliable model for analyzing the relationship between war and defeat.

[0127] Step 304: Extract the comment area portion from the vehicle comment data.

[0128] Here, because the vehicle review data contains a lot of information, some review data that is not related to the vehicle review can be removed, and only the review area related to the review can be extracted.

[0129] Step 305: Using the defeat relationship analysis model, predict the vehicle review data to obtain the target model, the series to which the target model belongs, the main competitor model, other competitor models, and the reason for purchase.

[0130] Step 306: Store the predicted data corresponding to the vehicle review data.

[0131] Here, the article ID corresponding to the vehicle review data, the target model in the vehicle review data, the series to which the target model belongs, the main competitor model, other competitor models, and the reason for purchase can be stored in a structured manner.

[0132] Step 307: Calculate the competitiveness of each vehicle model entity.

[0133] Here, we can use the main competing models, other competing models, and reasons for purchase for each model to calculate its competitiveness.

[0134] Step 308: Display the data in charts and graphs for each dimension.

[0135] Here, for each target model, all defeat relationship information corresponding to that target model can be extracted, and can be displayed graphically from the dimensions of the main competitor model, other competitor models, and reasons for purchase.

[0136] Example 4

[0137] Corresponding to the model generation method in the above embodiments, Figure 4 A structural block diagram of the model generation apparatus 400 provided in an embodiment of this application is shown. For ease of explanation, only the parts relevant to the embodiment of this application are shown. (Refer to...) Figure 4 The device includes a data acquisition unit 401, a sample generation unit 402, and a model training unit 403.

[0138] The data acquisition unit 401 is used to acquire vehicle review data and extract defeat relationship information from the vehicle review data. The defeat relationship information includes the main competitor model, other competitor models, and the factors of interest that the vehicle review data focuses on.

[0139] The sample generation unit 402 is used to annotate vehicle review data based on defeat relationship information to obtain annotated result data, and to store the vehicle review data and annotated result data as training samples into the training sample set.

[0140] The model training unit 403 is used to take the vehicle review data of the training samples in the training sample set as input and the labeled result data corresponding to the input vehicle review data as the expected output to train the defeat relationship analysis model. The defeat relationship analysis model is used to analyze the defeat relationship information in the vehicle review data.

[0141] In some embodiments, the data acquisition unit 401 is specifically used to: extract vehicle feature datasets corresponding to each vehicle model from the vehicle review data, and determine the main competitor model and other competitor models in the vehicle review data based on the vehicle feature datasets corresponding to each vehicle model.

[0142] In some embodiments, in the data acquisition unit 401, the process of determining the main competitor model and other competitor models in the vehicle review data based on the vehicle feature datasets corresponding to each vehicle model includes: when the vehicle review data includes a target model, determining the correlation between the corresponding model and the target model based on the vehicle feature datasets corresponding to each model; identifying models whose correlation meets preset correlation conditions as main competitor models; and identifying models other than the main competitor models as other competitor models. Other models are those in the vehicle review data other than the main competitor model and the target model.

[0143] In some embodiments, the data acquisition unit 401 is further configured to: extract multiple target keywords from vehicle review data, select a preset number of target keywords from the multiple target keywords in descending order of frequency of occurrence according to the frequency of occurrence of each target keyword in the vehicle review data, and determine the selected target keywords as factors of interest.

[0144] In some embodiments, the data acquisition unit 401 is further configured to: perform topic clustering analysis on vehicle review data to obtain multiple topics and vocabulary sets corresponding to each topic; select a preset number of vocabulary sets according to the vocabulary sets corresponding to each topic in descending order of vocabulary quantity; and determine the topics corresponding to the selected vocabulary sets as factors of interest.

[0145] In some embodiments, when the vehicle review data includes the target model, the annotation result data includes the target model, the vehicle series to which the target model belongs, and the defeat / defeat relationship information.

[0146] The apparatus provided in this embodiment can generate a defeat relationship analysis model for analyzing defeat relationship information in vehicle review data. This helps to obtain defeat relationship information in vehicle review data based on the defeat relationship analysis model. In other words, it helps to quickly extract the advantages and disadvantages of various types of cars described in the vehicle review data. Compared with related technologies where users need to read the entire vehicle review data to understand the competitive advantages and disadvantages of various types of cars described in the vehicle review data, this can greatly improve the user experience.

[0147] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the model generation method embodiment of this application. For details on their specific functions and technical effects, please refer to the model generation method embodiment section, which will not be repeated here.

[0148] Example 5

[0149] Corresponding to the information generation method in the above embodiments, Figure 5 A structural block diagram of the information generation apparatus 500 provided in an embodiment of this application is shown. For ease of explanation, only the parts related to the embodiment of this application are shown. (Refer to...) Figure 5 The device includes a data receiving unit 501 and an information output unit 502.

[0150] The data receiving unit 501 is used to input the vehicle review data into the defeat relationship analysis model when it receives vehicle review data, so as to obtain the defeat relationship information for the vehicle review data. The defeat relationship analysis model is generated by a model generation method.

[0151] The information output unit 502 is used to output defeat relationship information for vehicle review data.

[0152] In some embodiments, the apparatus further includes an information storage unit and a data analysis unit (not shown in the figure).

[0153] The information storage unit is used to store the target model, the vehicle series to which the target model belongs, and the defeat relationship information as associated information when the vehicle review data includes the target model;

[0154] The data analysis unit is used to determine the main competing models, other competing models, and factors of interest for each target model based on the relevant information corresponding to each target model, and to determine the competitiveness of the target model based on the main competing models, other competing models, and factors of interest for each target model.

[0155] The device provided in this embodiment can quickly analyze the defeat relationship information in vehicle review data using a pre-generated defeat relationship analysis model. In other words, it helps to quickly extract the advantages and disadvantages of various types of cars described in the vehicle review data. Compared with related technologies where users need to read the entire vehicle review data to understand the competitive advantages and disadvantages of various types of cars described in the vehicle review data, it can greatly improve the user experience.

[0156] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the information generation method embodiments of this application. For details on their specific functions and technical effects, please refer to the information generation method embodiments section, which will not be repeated here.

[0157] Example 6

[0158] Figure 6 This is a schematic diagram of the structure of an electronic device 600 provided in an embodiment of this application. Figure 6As shown, the electronic device 600 of this embodiment includes: at least one processor 601 ( Figure 6 The diagram shows only one processor, memory 602, and a computer program 603 stored in memory 602 and executable on at least one processor 601, such as a model generation program or an information generation program. When processor 601 executes computer program 603, it implements the steps in any of the above-described method embodiments. When processor 601 executes computer program 603, it implements the steps in the embodiments of the above-described model generation methods, or the steps in the embodiments of the above-described information generation methods. When processor 601 executes computer program 603, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 4 The functions of the data acquisition unit 401, sample generation unit 402, and model training unit 403 shown are as follows: Figure 5 The functions of the data receiving unit 501 and the information output unit 502 shown are illustrated.

[0159] For example, computer program 603 can be divided into one or more modules / units, one or more of which are stored in memory 602 and executed by processor 601 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of computer program 603 in electronic device 600. For example, computer program 603 can be divided into a data acquisition unit, a sample generation unit, a model training unit, a data receiving unit, and an information output unit. The specific functions of each unit have been described in the above embodiments and will not be repeated here.

[0160] Electronic device 600 may include, but is not limited to, processor 601 and memory 602. Those skilled in the art will understand that... Figure 6 This is merely an example of electronic device 600 and does not constitute a limitation on electronic device 600. It may include more or fewer components than shown, or combine certain components, or different components. For example, a vehicle may also include input / output devices, network access devices, buses, etc.

[0161] The processor 601 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0162] The memory 602 can be an internal storage unit of the electronic device 600, such as a hard disk or RAM of the electronic device 600. The memory 602 can also be an external storage device of the electronic device 600, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or FlashCard equipped on the electronic device 600. Furthermore, the memory 602 can include both internal and external storage units of the electronic device 600. The memory 602 is used to store computer programs and other programs and data required by the vehicle. The memory 602 can also be used to temporarily store data that has been output or will be output.

[0163] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0164] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0165] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0166] In the embodiments provided in this application, it should be understood that the disclosed devices / vehicles and methods can be implemented in other ways. For example, the device / vehicle embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0167] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0168] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0169] If an integrated module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer-readable storage medium can be non-volatile or volatile. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the contents of a computer-readable storage medium may be appropriately added to or subtracted from the contents as required by the legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, a computer-readable storage medium may not include electrical carrier signals and telecommunication signals.

[0170] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A model generation method, characterized in that, The method includes: Vehicle review data is acquired, and defeat relationship information is extracted from the vehicle review data. The defeat relationship information includes the main competitor model, other competitor models, and the factors of interest that the vehicle review data focuses on. The main competitor model is an alternative model that the sharer of the vehicle review data has not yet purchased, and the other competitor models are models used for comparison that the sharer has not yet purchased. Based on the defeat relationship information, the vehicle review data is labeled to obtain labeled result data, and the vehicle review data and the labeled result data are stored as training samples in the training sample set; The vehicle review data of the training samples in the training sample set is used as input, and the labeled result data corresponding to the input vehicle review data is used as the expected output. The initial deep learning model is trained to obtain the defeat relationship analysis model, wherein the defeat relationship analysis model is used to analyze the defeat relationship information in the vehicle review data; the initial deep learning model is a bidirectional encoder language representation model, wherein the BA label of the CRF layer of the bidirectional encoder language representation model indicates the main competitor model and the BM label indicates other competitor models; The extraction of defeat relationship information from the vehicle review data includes: Extract the vehicle feature datasets corresponding to each vehicle model from the vehicle review data; When the vehicle review data includes the target model, the minimum redundancy maximum relevance algorithm is used to determine the relevance between the corresponding model and the target model based on the model feature dataset corresponding to each model. Models whose relevance meets the preset relevance conditions are identified as the main competitor models, and other models are identified as the other competitor models. The other models are models in the vehicle review data other than the main competitor models and the target model. For each vehicle model, if the corresponding vehicle feature dataset S is... , where n is the total number of feature data in the vehicle model feature dataset S; if the vehicle model feature dataset corresponding to the target vehicle model is c, then when using the minimum redundancy maximum relevance algorithm to calculate the relevance of the vehicle model, the relevance of the vehicle model is: in, Let be the mutual information value between the i-th feature in the vehicle feature dataset S and the vehicle feature dataset c.

2. The model generation method according to claim 1, characterized in that, The extraction of defeat relationship information from the vehicle review data includes: Multiple target keywords are extracted from the vehicle review data. A preset number of target keywords are selected from the multiple target keywords according to their frequency of occurrence in the vehicle review data, in descending order of frequency. The selected target keywords are then identified as the factors of interest.

3. The model generation method according to claim 1, characterized in that, The extraction of defeat relationship information from the vehicle review data includes: The vehicle review data was subjected to topic clustering analysis to obtain multiple topics and the vocabulary sets corresponding to each topic; Based on the vocabulary sets corresponding to each topic, a preset number of vocabulary sets are selected in descending order of vocabulary quantity, and the topics corresponding to the selected vocabulary sets are determined as the factors of interest.

4. The model generation method according to any one of claims 1-3, characterized in that, When the vehicle review data includes a target model, the annotation result data includes the target model, the vehicle series to which the target model belongs, and the defeat / defeat relationship information.

5. An information generation method, characterized in that, The method includes: Upon receiving target vehicle review data, the target vehicle review data is input into a defeat relationship analysis model to obtain defeat relationship information for the target vehicle review data, wherein the defeat relationship analysis model is generated by any one of the model generation methods as described in claims 1-4; Output the defeat relationship information for the target vehicle review data.

6. The information generation method according to claim 5, characterized in that, The method further includes: When the target vehicle review data includes the target model, the target model, the vehicle series to which the target model belongs, and the defeat relationship information are stored as associated information; Based on the relevant information for each target model, identify the main competing models, other competing models, and factors of interest for each target model. Based on the main competing models, other competing models, and factors of interest for each target model, determine the competitiveness of each target model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the model generation method as described in any one of claims 1 to 4, or the information generation method as described in claim 5 or 6.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the model generation method as described in any one of claims 1 to 4, or the information generation method as described in claim 5 or 6.

Images