Vehicle public praise determination and public praise analysis model training method and electronic equipment

Abstract

The invention provides a vehicle word-of-mouth determination method, a word-of-mouth analysis model training method and an electronic device, the vehicle word-of-mouth determination method comprises the following steps: obtaining first information, the first information being evaluation information of a user on a target vehicle; the first information is input to a public praise analysis model, the public praise analysis model determines a first keyword vector corresponding to the first information and a first weight corresponding to the first keyword vector according to the first information, and determines a first score according to the first keyword vector and the first weight, and the first score represents the public praise of the target vehicle; wherein the public praise analysis model is a model obtained by training based on vehicle public praise data provided by a vehicle user and a vehicle evaluation report provided by a vehicle engineer. Therefore, the accuracy of determining the public praise of the vehicle is effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of big data analysis, and particularly relates to a method for determining vehicle word-of-mouth, training a word-of-mouth analysis model, and an electronic device. Background Art

[0002] Currently, vehicles have become an essential means of transportation for people to travel. During the process of using vehicles, people often make evaluations on vehicle performance, comfort, etc. The evaluation methods can be through communication with friends, or leaving messages and scoring on some vehicle-related websites. By analyzing these evaluation information, the attitudes, opinions, and emotional tendencies of users towards vehicles can be obtained.

[0003] In the prior art, when analyzing the evaluation information of users to obtain the word-of-mouth information corresponding to vehicles, most of them obtain the evaluation information related to vehicle word-of-mouth of users, and then extract evaluation classification and / or evaluation score data from these evaluation information, and determine the word-of-mouth of the vehicle according to the evaluation score and / or according to the evaluation classification. For example, the word-of-mouth of vehicles with high evaluation scores and good evaluation classifications is uniformly determined to be good, and the word-of-mouth of vehicles with low evaluation scores and low evaluation classifications is uniformly determined to be poor, etc. However, the evaluation classification and / or evaluation score data provided by users generally have strong personal emotions. This way of determining vehicle word-of-mouth based on the evaluation classification and / or scoring data provided by users has a strong individual subjectivity and there is a problem of being prone to overgeneralization. Therefore, there is a problem in the prior art that the vehicle word-of-mouth obtained through evaluation classification and / or evaluation score data is inaccurate. Summary of the Invention

[0004] The present application provides a method for determining vehicle word-of-mouth, training a word-of-mouth analysis model, and an electronic device, which can solve the problem in the prior art that the vehicle word-of-mouth obtained through evaluation classification and / or evaluation score data is inaccurate.

[0005] To solve the above technical problem, in a first aspect, an embodiment of the present application provides a method for determining vehicle word-of-mouth, the method includes: obtaining first information, where the first information is the evaluation information of a user on a target vehicle; inputting the first information into a word-of-mouth analysis model, and the word-of-mouth analysis model determines a first keyword vector corresponding to the first information and a first weight corresponding to the first keyword vector according to the first information, and determines a first score according to the first keyword vector and the first weight, where the first score represents the word-of-mouth of the target vehicle; wherein, the word-of-mouth analysis model is a model trained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers.

[0006] In the implementation manner of this application, the first information related to the user's evaluation information of the target vehicle is input into the word-of-mouth analysis model. The word-of-mouth analysis model determines the first keyword vector corresponding to the first information. In this process, the word-of-mouth analysis model can effectively extract all the keywords related to the vehicle word-of-mouth in the first information. Then, according to the first keyword vector and the first weight corresponding to the first keyword vector, the first score representing the word-of-mouth of the target vehicle is obtained. Compared with the method in the prior art of obtaining the vehicle word-of-mouth only through the evaluation classification and / or evaluation score data in the user evaluation information, the word-of-mouth analysis model in this application is a model based on big data. It can perform model analysis by extracting all the keywords in the first information and according to the weights related to the keywords, and then comprehensively analyze the user evaluation information, increasing the accuracy of determining the vehicle word-of-mouth.

[0007] Furthermore, the word-of-mouth analysis model is trained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers. The model training data not only includes the data provided by vehicle users, but also the professional data provided by vehicle engineers, making the output results of the trained word-of-mouth analysis model more comprehensive and objective, and also increasing the accuracy of determining the vehicle word-of-mouth.

[0008] In a possible implementation of the above first aspect, determining the first keyword vector corresponding to the first information according to the first information includes: preprocessing the first information to obtain the preprocessed first information; determining all the keywords, the corresponding intensifiers of each keyword, and the corresponding negators of each keyword in the preprocessed first information; and determining the first keyword vector according to each keyword, the corresponding intensifiers of each keyword, and the corresponding negators of each keyword.

[0009] In the implementation manner of this application, first, the first information is preprocessed (such as word segmentation, stop word removal, etc.) to obtain the preprocessed first information. Then, all the keywords, the corresponding intensifiers of each keyword, and the corresponding negators of each keyword in the preprocessed first information are determined; and the first keyword vector is determined according to each keyword, the corresponding intensifiers of each keyword, and the corresponding negators of each keyword. Thus, the obtained first keyword vector can more accurately reflect the user's emotion towards the target vehicle, thereby ensuring that the vehicle word-of-mouth determined according to the first keyword vector is also more accurate.

[0010] In a possible implementation of the above first aspect, determining the first score according to the first keyword vector and the first weight includes: determining the number of negators corresponding to each keyword included in the first keyword vector; and determining the first score according to the number of negators and the first weight.

[0011] In the implementation manner of the present application, the first score determined according to the number of negative words and the first weight effectively increases the accuracy of vehicle word-of-mouth determination.

[0012] In a possible implementation of the above first aspect, determining the first weight corresponding to the first keyword vector includes: determining the degree adverbs corresponding to each keyword included in the first keyword vector; determining the first weight according to each degree adverb.

[0013] In the implementation manner of the present application, the first weight corresponding to the first keyword vector can be accurately obtained according to each degree adverb, thereby increasing the accuracy of the first score determined according to the first keyword vector and the first weight, that is, increasing the accuracy of vehicle word-of-mouth determination.

[0014] In a possible implementation of the above first aspect, training the word-of-mouth analysis model based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers includes: obtaining a vehicle sentiment dictionary based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers; training the word-of-mouth analysis model based on the vehicle sentiment dictionary.

[0015] In the implementation manner of the present application, a vehicle sentiment dictionary is obtained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers, which records keywords specifically for the vehicle industry. The word-of-mouth analysis model trained through the vehicle sentiment dictionary has more comprehensive and objective output results, increasing the accuracy of vehicle word-of-mouth determination.

[0016] In a possible implementation of the above first aspect, the vehicle sentiment dictionary includes a vehicle degree adverb dictionary; and, determining the first weight according to each degree adverb includes: determining the adverb gear of each degree adverb in the vehicle degree adverb dictionary; determining the first weight according to the adverb gear corresponding to each degree adverb.

[0017] In the implementation manner of the present application, the first weight determined according to the adverb gear of the degree adverb in the vehicle degree adverb dictionary is more accurate, thereby increasing the accuracy of the first score determined according to the first keyword vector and the first weight, that is, increasing the accuracy of vehicle word-of-mouth determination.

[0018] In a possible implementation of the above first aspect, the adverb gear includes five gears. Determining the first score according to the number of negative words and the first weight includes determining the first score through the following formula:

[0019]

[0020] where Sentiment Score is the first score, x P is the positive degree adverb among the degree adverbs, xN is a negative degree adverb among degree adverbs, is a positive degree adverb x P The corresponding first weight, is a negative degree adverb x N The corresponding first weight, i is the adverb gear corresponding to the degree adverb, j is the word segmentation position of the degree adverb in the corresponding adverb gear, c j is a parameter related to the number of negative words. When the number of negative words corresponding to the degree adverb is even, c j takes the value of 1. When the number of negative words corresponding to the degree adverb is odd, c j takes the value of -1.

[0021] In a possible implementation of the above first aspect, the method further includes: obtaining vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers through the BeautifulSoup tool.

[0022] In the implementation manner of this application, it is easier to obtain vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers through the BeautifulSoup tool.

[0023] In a possible implementation of the above first aspect, the method further includes: when the word-of-mouth of the target vehicle does not match the actual word-of-mouth of the target vehicle, inputting the first information into the weight determination model to determine the second weight, and updating the first weight according to the second weight to update the word-of-mouth analysis model.

[0024] In the implementation manner of this application, when the word-of-mouth of the target vehicle does not match the actual word-of-mouth of the target vehicle, the second weight corresponding to the word-of-mouth analysis model is updated in time through the weight determination model to update the word-of-mouth analysis model, making the output result of the word-of-mouth analysis model more accurate, and then increasing the accuracy of vehicle word-of-mouth determination.

[0025] In a possible implementation of the above first aspect, the method further includes: updating the vehicle sentiment dictionary when the word-of-mouth of the target vehicle does not match the actual word-of-mouth of the target vehicle.

[0026] In the implementation manner of this application, when the word-of-mouth of the target vehicle does not match the actual word-of-mouth of the target vehicle, the vehicle sentiment dictionary is updated in time, increasing the accuracy and comprehensiveness of the word-of-mouth analysis model trained by the vehicle sentiment dictionary, making the output result of the word-of-mouth analysis model more comprehensive and objective, and increasing the accuracy of vehicle word-of-mouth determination.

[0027] Second aspect, an implementation manner of the present application provides a method for training a word-of-mouth analysis model, the method including: determining first training data and first test data, where the first training data and the first test data are obtained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers; inputting the first training data into an initial word-of-mouth analysis model for model training to obtain a first word-of-mouth analysis model, where the first word-of-mouth analysis model includes a third weight; inputting the first test data into the first word-of-mouth analysis model for testing to determine a test result; in the case where the test result does not meet the requirements, inputting the first training data into a weight determination model to determine a fourth weight, and updating the third weight according to the fourth weight to update the first word-of-mouth analysis model to obtain a target word-of-mouth analysis model; in the case where the test result meets the requirements, determining the first word-of-mouth analysis model as the target word-of-mouth analysis model.

[0028] In the implementation manner of the present application, by inputting the first training data obtained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers into an initial word-of-mouth analysis model for model training to obtain a first word-of-mouth analysis model, the output result of the trained word-of-mouth analysis model is more comprehensive and objective, can accurately output a first score representing the word-of-mouth of the target vehicle, and can increase the accuracy of vehicle word-of-mouth determination. Moreover, inputting the first test data into the first word-of-mouth analysis model for testing to determine a test result; in the case where the test result does not meet the requirements, timely updating the third weight corresponding to the first word-of-mouth analysis model through a weight determination model to update the first word-of-mouth analysis model to obtain a target word-of-mouth analysis model, so that the output result of the target word-of-mouth analysis model is more accurate, thereby increasing the accuracy of vehicle word-of-mouth determination.

[0029] In a possible implementation of the above second aspect, obtaining the first training data and the first test data based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers includes: obtaining a vehicle sentiment dictionary based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers; obtaining the first training data and the first test data based on the vehicle sentiment dictionary.

[0030] In the implementation manner of the present application, obtaining a vehicle sentiment dictionary based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers, which records keywords specifically for the vehicle industry, the target word-of-mouth analysis model trained with the first training data and the first test data obtained through the vehicle sentiment dictionary has a more comprehensive and objective output result, increasing the accuracy of vehicle word-of-mouth determination.

[0031] In a possible implementation of the above second aspect, the method further includes: updating the vehicle sentiment dictionary in the case where the test result does not meet the requirements.

[0032] In a third aspect, an embodiment of the present application provides a vehicle word-of-mouth determination device, including: a first processing module, configured to obtain first information, where the first information is evaluation information of a user on a target vehicle; a second processing module, configured to input the first information into a word-of-mouth analysis model, and the word-of-mouth analysis model determines a first keyword vector corresponding to the first information and a first weight corresponding to the first keyword vector according to the first information, and determines a first score according to the first keyword vector and the first weight, where the first score represents the word-of-mouth of the target vehicle; wherein, the word-of-mouth analysis model is a model trained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers.

[0033] In a fourth aspect, an embodiment of the present application provides a word-of-mouth analysis model training device, including: a third processing module, configured to determine first training data and first test data, where the first training data and the first test data are obtained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers; a fourth processing module, configured to input the first training data into an initial word-of-mouth analysis model for model training to obtain a first word-of-mouth analysis model, where the first word-of-mouth analysis model includes a third weight; a fifth processing module, configured to input the first test data into the first word-of-mouth analysis model for testing to determine a test result; a sixth processing module, configured to, when the test result does not meet the requirements, input the first training data into a weight determination model to determine a fourth weight, and update the third weight according to the fourth weight to update the first word-of-mouth analysis model to obtain a target word-of-mouth analysis model; when the test result meets the requirements, determine the first word-of-mouth analysis model as the target word-of-mouth analysis model.

[0034] In a fifth aspect, an embodiment of the present application provides an electronic device, including: a memory, configured to store a computer program, where the computer program includes program instructions; a processor, configured to execute the program instructions so that the electronic device executes the vehicle word-of-mouth determination method provided in the first aspect and / or any possible implementation manner of the first aspect, or the word-of-mouth analysis model training method provided in the second aspect and / or any possible implementation manner of the second aspect.

[0035] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the program instructions are run by an electronic device to execute the vehicle word-of-mouth determination method provided in the first aspect and / or any possible implementation manner of the first aspect, or the word-of-mouth analysis model training method provided in the second aspect and / or any possible implementation manner of the second aspect.

[0036] In a seventh aspect, an embodiment of the present application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the vehicle word-of-mouth determination method provided by the above first aspect and / or any possible implementation manner of the first aspect, or the word-of-mouth analysis model training method provided by the above second aspect and / or any possible implementation manner of the second aspect.

[0037] For the relevant beneficial effects of the above third aspect to the seventh aspect, reference may be made to the relevant descriptions of the above first aspect or the second aspect, which will not be elaborated here.

[0038] Advantages of the present application:

[0039] For the vehicle word-of-mouth determination method provided by the present application, the first information related to the evaluation information of the target vehicle by the user is input into the word-of-mouth analysis model, and the word-of-mouth analysis model determines the first keyword vector corresponding to the first information. In this process, the word-of-mouth analysis model can effectively extract all the keywords related to the vehicle word-of-mouth in the first information, and then, according to the first keyword vector and the first weight corresponding to the first keyword vector, obtain the first score representing the word-of-mouth of the target vehicle. Compared with the prior art method of obtaining the vehicle word-of-mouth only through the evaluation classification and / or evaluation score data in the user evaluation information, the word-of-mouth analysis model in the present application is a model based on big data, which can analyze the model by extracting all the keywords in the first information and according to the weights related to the keywords, and then comprehensively analyze the user evaluation information, increasing the accuracy of vehicle word-of-mouth determination.

[0040] Furthermore, the word-of-mouth analysis model is trained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers. The model training data not only includes the data provided by vehicle users, but also the professional data provided by vehicle engineers, making the output results of the trained word-of-mouth analysis model more comprehensive and objective, and also increasing the accuracy of vehicle word-of-mouth determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the present application, the drawings used in the description of the embodiments will be briefly introduced below.

[0042] Figure 1 is a schematic flowchart of a vehicle word-of-mouth determination method shown according to some implementation manners of the present application;

[0043] Figure 2 is a schematic flowchart of determining a first keyword vector corresponding to the first information according to the first information shown according to some implementation manners of the present application;

[0044] Figure 3According to some implementations of the present application, it shows a schematic flow diagram for determining a first score based on a first keyword vector and a first weight;

[0045] Figure 4 According to some implementations of the present application, it shows a schematic flow diagram for obtaining a word-of-mouth analysis model;

[0046] Figure 5 According to some implementations of the present application, it shows a schematic flow diagram for determining a first weight;

[0047] Figure 6 According to some implementations of the present application, it shows a schematic flow diagram for another method of determining vehicle word-of-mouth;

[0048] Figure 7 According to some implementations of the present application, it shows a schematic flow diagram for a method of training a word-of-mouth analysis model;

[0049] Figure 8 According to some implementations of the present application, it shows a schematic flow diagram for obtaining first training data and first test data;

[0050] Figure 9 According to some implementations of the present application, it shows a schematic flow diagram for another method of determining vehicle word-of-mouth;

[0051] Figure 10 According to some implementations of the present application, it shows a schematic structural diagram of a vehicle word-of-mouth determination device;

[0052] Figure 11 According to some implementations of the present application, it shows a schematic structural diagram of a word-of-mouth analysis model training device;

[0053] Figure 12 According to some implementations of the present application, it shows a schematic structural diagram of an electronic device. Detailed implementation manners

[0054] The technical solutions of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0055] As mentioned above, in the prior art, the method of determining vehicle word-of-mouth based on the evaluation classification and / or scoring data provided by users has strong individual subjectivity, and there are problems such as being prone to overgeneralization and obtaining inaccurate vehicle word-of-mouth.

[0056] That is to say, the existing methods for determining vehicle word-of-mouth rely too much on user participation and feedback, and only focus on limited and perceivable aspects (such as only focusing on the evaluation classification and / or scoring data in user evaluation information), which limits the objective, comprehensive and professional analysis of vehicle word-of-mouth data. Exemplarily, the evaluation information of a user for a certain vehicle is that the space is a little small and the score is 6 points; another user's evaluation of the vehicle is that the space is extremely small and the score is 6 points. The existing method directly extracts 6 points and determines the word-of-mouth of the vehicle based on 6 points. It does not analyze other evaluation information of the user. The user emotions expressed by "the space is a little small" and "the space is extremely small" are completely different. Therefore, the word-of-mouth of the vehicle determined only according to the user's evaluation score is not necessarily accurate. Similarly, there are similar problems for determining vehicle word-of-mouth only based on the evaluation classification (such as A, B, etc.) in the evaluation information.

[0057] Based on this, the present application provides a method for determining vehicle word-of-mouth, which can obtain a first score representing the word-of-mouth of the target vehicle according to information related to the evaluation information of the target vehicle by the user and a word-of-mouth analysis model. The word-of-mouth analysis model can comprehensively analyze the user evaluation information, increasing the accuracy of vehicle word-of-mouth determination.

[0058] Next, with reference to the accompanying drawings, the implementation process and advantages of the vehicle word-of-mouth determination method provided by the present application will be described in detail.

[0059] In one implementation manner of the present application, as Figure 1 shown, the vehicle word-of-mouth determination method includes the following steps:

[0060] S100: Obtain first information, where the first information is the evaluation information of the user for the target vehicle.

[0061] The vehicle word-of-mouth determination method can be applied to a server, a cloud, or a vehicle terminal. The obtained first information can be, during the execution of this method, the evaluation information of the user on various aspects such as vehicle performance and appearance during the use of the target vehicle, which is captured from vehicle-related websites, Tieba, etc. by an information capture tool (such as the BeautifulSoup tool). It can also be the evaluation information of the user for the target vehicle actively input by an engineer (such as a vehicle word-of-mouth analysis engineer) obtained from vehicle-related websites, Tieba for analyzing the vehicle word-of-mouth.

[0062] S200: Input the first information into the word-of-mouth analysis model. The word-of-mouth analysis model determines a first keyword vector corresponding to the first information and a first weight corresponding to the first keyword vector according to the first information, and determines a first score according to the first keyword vector and the first weight. The first score represents the word-of-mouth of the target vehicle.

[0063] Among them, the word-of-mouth analysis model is a model trained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers.

[0064] Determine the first keyword vector corresponding to the first information according to the first information, that is, extract the keywords in the evaluation information of the target vehicle by the user, and determine the word-of-mouth description of the vehicle by the user through the keywords. For example, words such as "loud noise" and "slow start" extracted from the user's evaluation information are all some keywords used to describe the vehicle performance in the vehicle industry.

[0065] In the process of keyword extraction, the word-of-mouth analysis model stores pre-trained word vectors, that is, the words commonly used to evaluate vehicles in the vehicle industry. By comparing the first information with the pre-trained word vectors, the keywords in the user's current evaluation information (i.e., the first information) can be obtained, such as noise, chassis, etc.

[0066] Furthermore, a user's evaluation information may describe various aspects of vehicle performance and appearance. The keywords extracted related to vehicle word-of-mouth are often more than one. The performance corresponding to some keywords has a general impact on vehicle word-of-mouth, and the performance corresponding to some keywords has a great impact on vehicle word-of-mouth. Therefore, in the word-of-mouth analysis model, it includes the weights (i.e., the first weights) corresponding to each keyword obtained by training through big data in advance. Different weights indicate the impact degree of the performance corresponding to the keyword on vehicle word-of-mouth. For example, "noise" has a relatively large impact degree, so its corresponding weight is larger, and "color" has a relatively small impact degree, so its corresponding weight is smaller. Moreover, for the same keyword, the emotional degree expressed by the user further affects the evaluation of vehicle word-of-mouth. For example, "the noise is very loud" indicates that the user is very dissatisfied with the vehicle noise, and the corresponding vehicle word-of-mouth should be very low. "The noise is slightly loud" indicates that the user only has a slight opinion on the vehicle noise and does not account for a large proportion in the determination of vehicle word-of-mouth. Therefore, in the process of determining the weight corresponding to the keyword, the degree adverb corresponding to the keyword can also be determined first, such as: "very", "especially", "generally", etc., and then the corresponding weight is determined according to the degree adverb. "Very" and "especially" indicate that the user's emotional experience of using the function corresponding to the keyword of the vehicle is relatively strong, and the corresponding weight can be set to 2, for example. "Generally" indicates that the user's emotional experience of using the function corresponding to the keyword of the vehicle is not very intense, and the corresponding weight can be set to 1, for example. Specifically, it can be set according to needs.

[0067] It should be noted that the first weight is pre-trained and preset in the keyword of the word-of-mouth analysis model or the weight corresponding to the adverb of degree corresponding to the keyword. There are often many keywords or adverbs of degree corresponding to the keyword. Therefore, the first weight is not just a single number. The first weight can be represented in the form of a weight vector or in the form of a weight matrix.

[0068] The word-of-mouth analysis model determines the first score representing the word-of-mouth of the target vehicle according to the first keyword vector and the first weight. According to the first score, the word-of-mouth of the target vehicle can be accurately known. For example, if the first score is 90, it means that the word-of-mouth of the target vehicle is very good. If the first score is 60, it means that the word-of-mouth of the target vehicle is average. If the first score is 30, it means that the word-of-mouth of the target vehicle is very poor. Engineers can improve the performance, appearance, etc. of the vehicle according to the word-of-mouth score of the target vehicle to increase the word-of-mouth of the vehicle.

[0069] The vehicle word-of-mouth determination method provided by this application inputs the first information related to the user's evaluation information of the target vehicle into the word-of-mouth analysis model. The word-of-mouth analysis model determines the first keyword vector corresponding to the first information. In this process, the word-of-mouth analysis model can effectively extract all keywords related to the vehicle word-of-mouth in the first information. Then, according to the first keyword vector and the first weight corresponding to the first keyword vector, the first score representing the word-of-mouth of the target vehicle is obtained. Compared with the method of obtaining the vehicle word-of-mouth only through the evaluation classification and / or evaluation score data in the user evaluation information in the prior art, the word-of-mouth analysis model in this application is a model based on big data. It can perform model analysis by extracting all keywords in the first information and according to the weights related to the keywords, and then comprehensively analyze the user evaluation information, increasing the accuracy of vehicle word-of-mouth determination.

[0070] Furthermore, the word-of-mouth analysis model is trained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers. The model training data not only includes the data provided by vehicle users, but also the professional data provided by vehicle engineers. By comprehensively combining the data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers, the output results of the trained word-of-mouth analysis model are more comprehensive and objective, and also increase the accuracy of vehicle word-of-mouth determination.

[0071] In an implementation manner of this application, as Figure 2 shown, determining the first keyword vector corresponding to the first information according to the first information includes the following steps:

[0072] S210: Preprocess the first information to obtain the preprocessed first information.

[0073] Preprocess the first information. The specific preprocessing method can be tokenizing the first information, removing stop words, etc., and then obtain the preprocessed first information.

[0074] S220: Determine all keywords in the preprocessed first information, the corresponding degree adverbs for each keyword, and the corresponding negative words for each keyword.

[0075] S230: Determine the first keyword vector according to each keyword, the corresponding degree adverbs for each keyword, and the corresponding negative words for each keyword.

[0076] Traditional algorithms for feature extraction of sentences generally use vector space for calculation. The method of using vector space for calculation can only extract the features expressed in the sentence, but lacks the consideration of text length and semantic connection, resulting in certain limitations in the results of sentiment analysis based on the extracted features.

[0077] For example, the user's evaluation information of the target vehicle is "I don't think this car is particularly noisy", or "I feel this car is very, very noisy". In the general way of calculating with spatial vectors, the extracted feature information will be "noisy". However, if we consider the degree words and negative words mentioned in each sentence, it will be found that for the first evaluation, it actually means the noise is not big, and for the second user evaluation, it actually means the noise is very big.

[0078] Therefore, in the implementation manner of the present application, in the process of extracting the keyword vector, not only can the feature "noisy" in the first information be extracted, but also degree adverbs such as "not", "particularly", "very, very" and negative words can be extracted according to the semantic connection of the context, and then the first keyword vector can be determined, so as to comprehensively analyze the first information in the process of using the model analysis and obtain the accurate word-of-mouth score of the target vehicle.

[0079] In one implementation manner of the present application, as Figure 3 shown, determining the first score according to the first keyword vector and the first weight includes the following steps:

[0080] S240: Determine the number of negative words corresponding to each keyword included in the first keyword vector.

[0081] In the process of some users evaluating the vehicle, there will be cases of using negative sentences. For example, "I don't think this car is not good", which actually means this car is good. Therefore, in the process of analyzing keywords, it is necessary to extract the number of negative words corresponding to the negative words to further determine the true meaning expressed by the user evaluation information.

[0082] S250: Determine a first score based on the number of negative words and the first weight.

[0083] In an implementation manner of the present application, determining the first weight corresponding to the first keyword vector includes: determining the degree adverbs corresponding to each keyword included in the first keyword vector, and determining the first weight according to each degree adverb.

[0084] As described above in detail, different degree adverbs express different emotions of users. Therefore, the weights corresponding to different degree adverbs are different. For example, degree adverbs such as "extremely extremely" and "very very" indicating strong emotions of users generally have relatively large weights, which can be set to 3. For "very" and "extremely", the weights can be set to 2. For words such as "generally" and "a bit", they can be set to 1. The specific numerical values of the above weight settings are only for illustration, and the size of the weight can be set as needed.

[0085] In an implementation manner of the present application, as Figure 4 shown, training the word-of-mouth analysis model based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers includes the following steps:

[0086] S201: Obtain a vehicle sentiment dictionary based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers.

[0087] At present, natural language analysis is based on machine learning and deep learning technologies. By training on large-scale text data and building models, computers can understand and process human language. Since the terms in the vehicle industry are highly targeted, for example, the term "slow" has a specific meaning in the vehicle industry. Therefore, when sentiment analysis involves the understanding of natural language and ambiguity processing, the sentiment analysis results of many vehicle reviews are contrary to the original intention of users. Deep learning models may have difficulties in processing complex semantics and context relevance because they usually learn based on statistical patterns and lack in-depth understanding of semantics and logic.

[0088] Therefore, in the present application, obtaining a vehicle sentiment dictionary based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers can count all the commonly used words in the vehicle industry and obtain a keyword dictionary for vehicle industry use, that is, the vehicle sentiment dictionary.

[0089] Furthermore, according to the specific uses of words, the words in the vehicle sentiment dictionary can be classified to obtain multiple sub-dictionaries. The sub-dictionaries can include a vehicle word segmentation vocabulary dictionary, a vehicle stop word dictionary, a vehicle positive word dictionary, a vehicle negative word dictionary, a vehicle degree adverb dictionary, a vehicle negative word dictionary, etc.

[0090] In an implementation of the present application, the BeautifulSoup tool is used to scrape the vehicle word-of-mouth data of the vehicle network (i.e., the vehicle word-of-mouth data provided by vehicle users), and an engineering evaluation report of vehicle industry engineers is collected to establish a vehicle sentiment dictionary (which may include a vehicle industry jieba segmentation word library, a stop word library, and a vehicle sentiment analysis word library).

[0091] S202: Train a word-of-mouth analysis model based on the vehicle sentiment dictionary.

[0092] Taking the word information in the vehicle sentiment dictionary as the input and the corresponding vehicle word-of-mouth score as the output, a word-of-mouth analysis model for determining the score corresponding to the vehicle word-of-mouth can be trained.

[0093] In an implementation of the present application, as Figure 5 shown, the vehicle sentiment dictionary includes a vehicle degree adverb dictionary; and, determining the first weight according to each degree adverb includes the following steps:

[0094] S211: Determine the adverb gear of each degree adverb in the vehicle degree adverb dictionary.

[0095] S212: Determine the first weight according to the adverb gear corresponding to each degree adverb.

[0096] Specifically, for example, there are a total of 5 adverb gears in the vehicle degree adverb dictionary, and the weights of the adverbs corresponding to each adverb gear are different. After analyzing the first information to obtain the corresponding first word vector, determine the gear where the degree adverb corresponding to each keyword is located. For example, the adverb gears where "especially especially" and "very very" are located are the 5th gear, and their corresponding weight is 3, and the adverb gears where "general" and "slightly" are located are the 3rd gear, and their corresponding weight is 1, etc. The degree adverbs recorded in the vehicle degree adverb dictionary may include positive degree adverbs and negative degree adverbs.

[0097] In an implementation of the present application, both the positive degree adverbs and the negative degree adverbs can have their own corresponding vehicle degree adverb dictionaries, that is, the vehicle degree adverb dictionary can include, for example, a vehicle positive degree adverb dictionary, a vehicle negative degree adverb dictionary, etc. The vehicle positive degree adverb dictionary and the vehicle negative degree adverb dictionary can both have multiple adverb gears, such as 5 adverb gears, and each adverb gear has multiple word segmentation positions.

[0098] In addition, in another implementation of the present application, the number of adverb gears included in the aforementioned vehicle degree adverb dictionary, vehicle positive degree adverb dictionary, and vehicle negative degree adverb dictionary can also be other numbers, such as 4 adverb gears, 6 adverb gears, or 7 adverb gears, which can be set according to needs.

[0099] In one implementation of the present application, the adverb gear includes five gears. The first score is determined according to the number of negative words and the first weight, including determining the first score through the following formula:

[0100]

[0101] where Sentiment Score is the first score, x P is the positive degree adverb in the degree adverbs, x N is the negative degree adverb in the degree adverbs, is the first weight corresponding to the positive degree adverb x P ; is the first weight corresponding to the negative degree adverb x N , i is the adverb gear corresponding to the degree adverb, j is the word segmentation position of the degree adverb in the corresponding adverb gear, c j is a parameter related to the number of negative words. When the number of negative words corresponding to the degree adverb is even, c j takes the value of 1. When the number of negative words corresponding to the degree adverb is odd, c j takes the value of -1.

[0102] In one implementation of the present application, the above formula can also be expressed as

[0103]

[0104] where Sentiment Score is the first score, x P is the positive degree adverb in the degree adverbs, x N is the negative degree adverb in the degree adverbs, is the first weight corresponding to the positive degree adverb x P ; is the first weight corresponding to the negative degree adverb x N , n ij is the position information of the degree adverb, where i is the adverb gear corresponding to the degree adverb, j is the word segmentation position of the degree adverb in the corresponding adverb gear, c j is a parameter related to the number of negative words. When the number of negative words corresponding to the degree adverb is even, c j takes the value of 1. When the number of negative words corresponding to the degree adverb is odd, c j takes the value of -1.

[0105] In one implementation of the present application, as Figure 6 shown, the method further includes the following steps:

[0106] S300: When the reputation of the target vehicle does not match its actual reputation, input the first information into the weight determination model to determine the second weight, and update the first weight according to the second weight, so as to update the reputation analysis model.

[0107] The adjustment of the first weight uses the attention mechanism adjustment method. For example, the first information can be input into the weight determination model, and in the weight determination model, for a specific attention head or layer, a new second weight (which can also be called a new attention weight matrix) is obtained, that is, the weight corresponding to the intensifier is adjusted (for example, the weights of "extremely" and "very" are adjusted to 4, etc.). Replace the original weight matrix with the new attention weight matrix (that is, the first weight, which can also be called the initial weight before adjustment). This is used to adjust the attention degree of the reputation analysis model to the intensifiers corresponding to different keywords, so as to meet the needs of evaluating vehicle reputation in the vehicle field.

[0108] Furthermore, the weight determination model can be a trained model obtained based on the RoBERTa model, or a trained model obtained based on the BERT (full name: Bidirectional Encoder Representations from Transformers) model and lightweight models derived from the BERT model. Lightweight models derived from the BERT model can be, for example, the Albert model, the Xlnet model, etc. It can also be a trained model obtained based on a model with a different architecture from the BERT model, such as the GPT (full name: Generative Pre-trained Transformer) model and models derived from the GPT model. Models derived from the GPT model can be, for example, GPT3, GPT5, etc.

[0109] In an implementation manner of the present application, the method further includes: updating the vehicle sentiment dictionary when the reputation of the target vehicle does not match its actual reputation.

[0110] In this implementation manner, statements that do not meet the expected results are screened and checked, and then the vehicle sentiment dictionary is updated according to the results, such as adding vocabulary content to the dictionary or adjusting the vocabulary content in the dictionary. Considering that there are reverse sentiment words in the context in reality, such as "loud noise" and "too soft chassis" which are opposite to the general cognitive sentiment vectors, the present application uses a combination of precise screening and fuzzy screening. In addition to phrases like "slow start", retrieval methods for phrases like "noise... loud" are also added to the vehicle sentiment dictionary. This method requires splitting the data using two terminators (such as punctuation marks, prepositions, etc.), and then combining operations with the jieba word segmentation method, the vehicle word segmentation vocabulary dictionary, and the vehicle positive / negative dictionary.

[0111] In one implementation of the present application, as Figure 7 shown, the present application provides a method for training a word-of-mouth analysis model, and the method includes the following steps:

[0112] S10: Determine first training data and first test data, where the first training data and the first test data are obtained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers.

[0113] S20: Input the first training data into an initial word-of-mouth analysis model for model training to obtain a first word-of-mouth analysis model, where the first word-of-mouth analysis model includes a third weight.

[0114] S30: Input the first test data into the first word-of-mouth analysis model for testing to determine the test result.

[0115] S40: In the case where the test result does not meet the requirements (for example, the vehicle word-of-mouth determined according to the test result does not match the real vehicle word-of-mouth), input the first training data into a weight determination model to determine a fourth weight, and update the third weight according to the fourth weight to update the first word-of-mouth analysis model to obtain a target word-of-mouth analysis model. In the case where the test result meets the requirements (for example, the vehicle word-of-mouth determined according to the test result matches the real vehicle word-of-mouth), determine the first word-of-mouth analysis model as the target word-of-mouth analysis model.

[0116] That is, during the training process of the model, the weights corresponding to the word vectors will be trained in advance and then embedded into the word-of-mouth analysis model. During use, the weights embedded in the word-of-mouth analysis model are used as the initial weights.

[0117] The first test data is the prepared test data, and its corresponding user evaluation information and real vehicle word-of-mouth are both known. Input the first test data into the first word-of-mouth analysis model for testing. By comparing the test result with the real vehicle word-of-mouth, it is possible to know the accuracy of the first word-of-mouth analysis model in determining the vehicle word-of-mouth, and then adjust and optimize the first word-of-mouth analysis model.

[0118] In one implementation of the present application, as Figure 8 shown, obtaining the first training data and the first test data based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers includes the following steps:

[0119] S11: Obtain a vehicle sentiment dictionary based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers.

[0120] S12: Obtain the first training data and the first test data based on the vehicle sentiment dictionary.

[0121] In one implementation of the present application, the method further includes: updating the vehicle sentiment dictionary when the test result does not meet the requirements.

[0122] In one implementation of the present application, as Figure 9 shown, the vehicle word-of-mouth determination method provided by the present application includes the steps of:

[0123] Determine the training corpus (such as user evaluations of the vehicle, i.e., the first information), and train a deep learning sentiment analysis model (i.e., the word-of-mouth analysis model) based on the vehicle sentiment dictionary, and determine the keyword vector or the initial weight (i.e., the first weight) of the degree adverb corresponding to the keyword word vector according to the vehicle sentiment dictionary, and load the initial weight into the embedding layer of the deep learning sentiment analysis model.

[0124] Perform preprocessing such as word segmentation and stop word removal on the training corpus according to the vehicle segmentation vocabulary dictionary and vehicle stop word vocabulary dictionary included in the vehicle sentiment dictionary to obtain the preprocessed training corpus.

[0125] Pre-train word vectors according to the preprocessed training corpus to obtain the keyword vector corresponding to the training corpus.

[0126] Input the keyword vector into the deep learning sentiment analysis model, and the model performs multi-task learning, taking the sentiment analysis task as the main task and the dictionary matching task as the auxiliary task for multi-task learning to obtain the first score representing the vehicle word-of-mouth output by the model.

[0127] Randomly sample the vehicle word-of-mouth determined by the first score output by the model and export the result, and determine whether the result meets the expectation, that is, determine whether the word-of-mouth of the vehicle output by the model is consistent with the actual word-of-mouth of the vehicle.

[0128] When the output result of the model does not meet the expectation, that is, when the word-of-mouth of the vehicle output by the model is not consistent with the actual word-of-mouth of the vehicle, optimize and adjust the initial weight loaded into the embedding layer of the deep learning sentiment analysis model, load the new weight into the embedding layer of the deep learning sentiment analysis model, and optimize the vehicle sentiment dictionary. When the output result of the model meets the expectation, that is, when the word-of-mouth of the vehicle output by the model is consistent with the actual word-of-mouth of the vehicle, save the test result.

[0129] The vehicle word-of-mouth determination method provided by the present application enables the deep learning model to simultaneously learn the ability to learn sentiment information from the context and perform sentiment analysis using the dictionary by introducing the vehicle sentiment dictionary. By jointly training the two tasks of the deep learning model and the vehicle sentiment dictionary, the performance of sentiment analysis and the vehicle text analysis ability are improved. Continuously iterate to calculate a more comprehensive, objective and accurate vehicle word-of-mouth analysis result.

[0130] This application improves the accuracy of vehicle word-of-mouth determination by collecting pre-processed user evaluation data for the vehicle industry and performing multi-dimensional feature extraction. After excluding user scoring data, it uses the word-of-mouth data of a large number of vehicle models and trains a sentiment analysis model based on the automotive industry sentiment dictionary through methods such as natural language processing, machine learning, and statistics to perform fine-grained sentiment classification and scoring on user evaluations. By applying big data analysis to dilute the subjective interference of users, it improves the objectivity of vehicle word-of-mouth analysis. This application uses three methods (pre-trained vector words, introducing a vehicle sentiment dictionary, and deep learning) and a fusion model (deep learning and introducing a dictionary) to analyze the feature correlation during the sentiment analysis process, increasing the accuracy of vehicle word-of-mouth determination.

[0131] Please refer to Figure 10 , Figure 10 shown in the following is a vehicle word-of-mouth determination device of this application, including:

[0132] A first processing module, configured to obtain first information, where the first information is the evaluation information of a user on a target vehicle.

[0133] A second processing module, configured to input the first information into a word-of-mouth analysis model. The word-of-mouth analysis model determines a first keyword vector corresponding to the first information and a first weight corresponding to the first keyword vector according to the first information, and determines a first score according to the first keyword vector and the first weight. The first score represents the word-of-mouth of the target vehicle; among them, the word-of-mouth analysis model is a model trained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers.

[0134] The specific operation content that each processing module can perform can be referred to the above Figure 1 corresponding vehicle word-of-mouth determination method. Moreover, according to the specific operation steps of the above vehicle word-of-mouth determination method, the vehicle word-of-mouth determination device may include more or fewer processing modules for processing the content in the above vehicle word-of-mouth determination method.

[0135] Please refer to Figure 11 , Figure 11 shown in the following is a word-of-mouth analysis model training device of this application, including:

[0136] A third processing module, configured to determine first training data and first test data, where the first training data and the first test data are obtained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers.

[0137] A fourth processing module, configured to input the first training data into an initial word-of-mouth analysis model for model training to obtain a first word-of-mouth analysis model, where the first word-of-mouth analysis model includes a third weight.

[0138] The fifth processing module is configured to input the first test data into the first word-of-mouth analysis model for testing and determine the test result.

[0139] The sixth processing module is configured to, when the test result does not meet the requirements, input the first training data into the weight determination model to determine the fourth weight, and update the third weight according to the fourth weight, so as to update the first word-of-mouth analysis model to obtain the target word-of-mouth analysis model; when the test result meets the requirements, determine the first word-of-mouth analysis model as the target word-of-mouth analysis model.

[0140] The specific operation content that each processing module can perform can be referred to the above Figure 7 corresponding word-of-mouth analysis model training method. Moreover, according to the specific operation steps of the above word-of-mouth analysis model training method, the word-of-mouth analysis model training device may include more or fewer processing modules for processing the content in the above word-of-mouth analysis model training method.

[0141] Please refer to Figure 12 , Figure 12 Shown is a block diagram of a structure of an electronic device provided by an implementation manner of the present application. The electronic device may include one or more processors 1002, a system control logic 1008 connected to at least one of the processors 1002, a system memory 1004 connected to the system control logic 1008, a non-volatile memory (NVM) 1006 connected to the system control logic 1008, and a network interface 1010 connected to the system control logic 1008.

[0142] The processor 1002 may include one or more single-core or multi-core processors. The processor 1002 may include any combination of a general-purpose processor and a dedicated processor (such as a graphics processor, an application processor, a baseband processor, etc.). In the implementation manner herein, the processor 1002 may be configured to execute the aforementioned vehicle word-of-mouth determination method or the word-of-mouth analysis model training method.

[0143] In some implementation manners, the system control logic 1008 may include any suitable interface controller to provide any suitable interface to at least one of the processors 1002 and / or any suitable device or component communicating with the system control logic 1008.

[0144] In some implementations, the system control logic 1008 may include one or more memory controllers to provide an interface to the system memory 1004. The system memory 1004 may be used to load and store data and / or instructions. In some implementations, the system memory 1004 of the electronic device may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).

[0145] The NVM / memory 1006 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some implementations, the NVM / memory 1006 may include any suitable non-volatile memory such as flash memory and / or any suitable non-volatile storage device, such as at least one of a hard disk drive (HDD), a compact disc (CD) drive, and a digital versatile disc (DVD) drive.

[0146] The NVM / memory 1006 may include a portion of the storage resources mounted on the device of the electronic device, or it may be accessible by the device but not necessarily part of the device. For example, the NVM / memory 1006 may be accessed via the network interface 1010 over a network.

[0147] Specifically, the system memory 1004 and the NVM / memory 1006 may respectively include: a temporary copy and a permanent copy of the instructions 1020. The instructions 1020 may include: instructions that cause the electronic device to implement the foregoing vehicle word-of-mouth determination method or word-of-mouth analysis model training method when executed by at least one of the processors 1002. In some implementations, the instructions 1020, hardware, firmware, and / or its software components may additionally / alternatively be disposed in the system control logic 1008, the network interface 1010, and / or the processor 1002.

[0148] The network interface 1010 may include a transceiver for providing a radio interface for the electronic device to communicate with any other suitable device (such as a front-end module, an antenna, etc.) over one or more networks. In some implementations, the network interface 1010 may be integrated with other components of the electronic device. For example, the network interface 1010 may be integrated with at least one of the processor 1002, the system memory 1004, the NVM / memory 1006, and a firmware device (not shown) having instructions, and when at least one of the processors 1002 executes the instructions, the electronic device implements the foregoing vehicle word-of-mouth determination method or word-of-mouth analysis model training method.

[0149] The network interface 1010 may further include any suitable hardware and / or firmware to provide a multiple-input multiple-output radio interface. For example, the network interface 1010 may be a network adapter, a wireless network adapter, a telephone modem, and / or a wireless modem.

[0150] In one implementation, at least one of the processors 1002 may be logically encapsulated with one or more controllers for the system control logic 1008 to form a System In a Package (SiP). In one implementation, at least one of the processors 1002 may be integrated with the logic of one or more controllers for the system control logic 1008 on the same die to form a System on Chip (SoC).

[0151] The electronic device may further include an input / output (I / O) device 1012. The I / O device 1012 may include a user interface that enables a user to interact with the electronic device; the design of the peripheral component interface enables peripheral components to also interact with the electronic device. In some implementations, the electronic device further includes sensors for determining at least one of environmental conditions and location information related to the electronic device.

[0152] In some implementations, the user interface may include, but is not limited to, a display (e.g., a liquid crystal display, a touch screen display, etc.), a speaker, a microphone, one or more cameras (e.g., a still image camera and / or a video camera), a flashlight (e.g., a light-emitting diode flash), and a keyboard.

[0153] In some implementations, the peripheral component interface may include, but is not limited to, a non-volatile memory port, an audio jack, and a power interface.

[0154] In some implementations, the sensors may include, but are not limited to, a gyroscope sensor, an accelerometer, a proximity sensor, an ambient light sensor, and a positioning unit. The positioning unit may also be part of the network interface 1010 or interact with the network interface 1010 to communicate with components of a positioning network (e.g., Global Positioning System (GPS) satellites).

[0155] It can be understood that the structure schematically shown in the implementations of the present invention does not constitute a specific limitation on the electronic device. In other implementations of the present application, the electronic device may include more or fewer components than shown, or combine certain components, or split certain components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0156] Program code can be applied to input instructions to perform the various functions described herein and generate output information. The output information can be applied to one or more output devices in a known manner. For purposes of the implementations of this application, a processing system includes any system having a processor such as, for example, a Digital Signal Processor (DSP), a microcontroller, an Application Specific Integrated Circuit (ASIC), or a microprocessor.

[0157] The program code can be implemented in a high-level procedural or object-oriented programming language so as to communicate with the processing system. When needed, the program code can also be implemented in assembly language or machine language. In fact, the mechanisms described herein are not limited to the scope of any particular programming language. In any case, the language can be a compiled language or an interpreted language.

[0158] One or more aspects of at least one implementation can be realized by representative instructions stored on a computer-readable storage medium, the instructions representing various logic in a processor, the instructions causing the machine, when read by the machine, to fabricate logic for performing the techniques described herein. These representations, referred to as “IP cores,” can be stored on a tangible computer-readable storage medium and provided to multiple customers or production facilities to be loaded into a manufacturing machine that actually fabricates the logic or processor.

[0159] It should be noted that in the drawings, some structural or method features may be shown in a particular arrangement and / or order. However, it should be understood that such a particular arrangement and / or ordering may not be required. Rather, in some implementations, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. Additionally, the inclusion of a structural or method feature in a particular figure does not imply that such a feature is required in all implementations, and in some implementations, these features may not be included or may be combined with other features.

[0160] It should be noted that the terms “first,” “second,” etc. are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0161] It should be noted that in the drawings, some structural or method features may be shown in a particular arrangement and / or order. However, it should be understood that such a particular arrangement and / or ordering may not be required. Rather, in some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. Additionally, the inclusion of a structural or method feature in a particular figure does not imply that such a feature is required in all embodiments, and in some embodiments, these features may not be included or may be combined with other features.

[0162] Although the present application has been illustrated and described by reference to certain preferred embodiments thereof, those of ordinary skill in the art should understand that the above content is a further detailed description of the present application in conjunction with specific embodiments, and it cannot be determined that the specific implementation of the present application is only limited to these descriptions. Those skilled in the art can make various changes in form and detail, including making several simple deductions or substitutions, without departing from the spirit and scope of the present application.

Claims

1. A method for determining vehicle word-of-mouth, characterized in that, The method includes: Obtaining first information, where the first information is evaluation information of a target vehicle by a user; Inputting the first information into a word-of-mouth analysis model, which determines a first keyword vector corresponding to the first information and a first weight corresponding to the first keyword vector according to the first information, and determines a first score according to the first keyword vector and the first weight, where the first score represents the word-of-mouth of the target vehicle; wherein, The word-of-mouth analysis model is a model trained based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers.

2. The vehicle word-of-mouth determination method according to claim 1, wherein Wherein, Determining the first keyword vector corresponding to the first information according to the first information includes: Performing preprocessing on the first information to obtain preprocessed first information; Determining all keywords in the preprocessed first information, the degree adverbs corresponding to each keyword, and the negation words corresponding to each keyword; Determining the first keyword vector according to each keyword, the degree adverbs corresponding to each keyword, and the negation words corresponding to each keyword.

3. The vehicle word-of-mouth determination method according to claim 2, wherein Wherein, Determining the first score according to the first keyword vector and the first weight includes: Determining the number of negation words corresponding to each keyword included in the first keyword vector; Determining the first score according to the number of negation words and the first weight.

4. The vehicle word-of-mouth determination method according to claim 3, characterized in that Determining the first weight corresponding to the first keyword vector includes: Determining the degree adverbs corresponding to each keyword included in the first keyword vector; Determining the first weight according to each degree adverb.

5. The vehicle word-of-mouth determination method according to any one of claims 1-4, characterized in that, Wherein, Training the word-of-mouth analysis model based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers includes: Obtaining a vehicle sentiment dictionary based on vehicle word-of-mouth data provided by vehicle users and vehicle evaluation reports provided by vehicle engineers; Training the word-of-mouth analysis model based on the vehicle sentiment dictionary.

6. The vehicle word-of-mouth determination method according to claim 5, wherein The vehicle sentiment dictionary includes a vehicle degree adverb dictionary; And, determining the first weight according to each degree adverb includes: Determining the adverb gear of each degree adverb in the vehicle degree adverb dictionary; Determining the first weight according to the adverb gear corresponding to each degree adverb.

7. The vehicle word-of-mouth determination method according to claim 6, characterized in that, The adverb gear includes five gears. Determining the first score according to the number of negation words and the first weight includes determining the first score through the following formula: Among them, Sentiment Score is the first score, x P is the positive degree adverb among the degree adverbs, x N is the negative degree adverb among the degree adverbs, is the positive degree adverb x P corresponding to the first weight, is the negative degree adverb x N corresponding to the first weight, i is the adverb gear corresponding to the degree adverb, j is the participle position of the degree adverb in the corresponding adverb gear, c j is a parameter related to the number of negative words. When the number of negative words corresponding to the degree adverb is even, c j takes the value of 1. When the number of negative words corresponding to the degree adverb is odd, c j takes the value of -1.

8. The vehicle word-of-mouth determination method according to claim 7, wherein The method further includes: In the case where the word-of-mouth of the target vehicle does not match the actual word-of-mouth of the target vehicle, inputting the first information into a weight determination model to determine a second weight, and updating the first weight according to the second weight to update the word-of-mouth analysis model.

9. The vehicle word-of-mouth determination method according to claim 8, wherein The method further includes: In the case where the word-of-mouth of the target vehicle does not match the actual word-of-mouth of the target vehicle, updating the vehicle sentiment dictionary.

10. A method for training a word-of-mouth analysis model, characterized in that, The method includes: Determine the first training data and the first test data, where the first training data and the first test data are obtained based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers; Input the first training data into the initial word-of-mouth analysis model for model training to obtain the first word-of-mouth analysis model, where the first word-of-mouth analysis model includes a third weight; Input the first test data into the first word-of-mouth analysis model for testing to determine the test result; When the test result does not meet the requirements, input the first training data into the weight determination model to determine a fourth weight, and update the third weight according to the fourth weight to update the first word-of-mouth analysis model to obtain the target word-of-mouth analysis model; When the test result meets the requirements, determine the first word-of-mouth analysis model as the target word-of-mouth analysis model.

11. The method for training the word-of-mouth analysis model according to claim 10, wherein, Obtaining the first training data and the first test data based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers includes: Obtain a vehicle sentiment dictionary based on the vehicle word-of-mouth data provided by vehicle users and the vehicle evaluation reports provided by vehicle engineers; Obtain the first training data and the first test data based on the vehicle sentiment dictionary.

12. The method for training the word-of-mouth analysis model according to claim 11, wherein, The method further includes: When the test result does not meet the requirements, update the vehicle sentiment dictionary.

13. An electronic device, characterized in that, It includes: A processor and a memory communicatively connected to the processor; The memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to enable the electronic device to implement the vehicle word-of-mouth determination method according to any one of claims 1-9, or to enable the electronic device to implement the word-of-mouth analysis model training method according to any one of claims 10-12.

Images