A training method of a source evaluation model, a source evaluation method, and related products

Abstract

This application discloses a training method for a source evaluation model, a source evaluation method, and related products. Text sample sequences and frequency sample sequences are input into the evaluation model to be trained. The model encodes the text sample sequences and frequency sample sequences to obtain text sample vectors corresponding to the text sample sequences and frequency sample vectors corresponding to the frequency sample sequences. The evaluation model then performs prediction and evaluation processing on the text sample vectors and frequency sample vectors to obtain source evaluation prediction results. Based on the difference between the source evaluation result labels and the source evaluation prediction results, the parameters of the evaluation model are adjusted until the adjusted model meets the model training cutoff condition, and training ends to obtain the source evaluation model. Thus, this application can construct a source evaluation model based on the source sample name and the title, keywords, and publication frequency of the published sample text as features, thereby achieving the evaluation of the source.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a training method for a source evaluation model, a source evaluation method, and related products. Background Technology

[0002] Map text intelligence can provide efficient and accurate navigation information for objects, and it is mainly constructed based on text published by reliable sources. However, with the explosive growth of objects on internet platforms, the quality of the text published by these objects varies greatly, resulting in inaccurate map text intelligence. Furthermore, there is currently no relevant technical solution to evaluate the sources and select reliable ones for constructing map text intelligence.

[0003] Therefore, how to evaluate information sources has become a pressing technical problem that needs to be solved in the field. Summary of the Invention

[0004] This application provides a training method for a source evaluation model, a source evaluation method, and related products, aiming to achieve the evaluation of sources.

[0005] The first aspect of this application provides a method for training a source evaluation model, including:

[0006] Obtain the source sample name and source evaluation result label corresponding to the source sample, and obtain the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample;

[0007] The source sample name, the sample text titles corresponding to the multiple sample texts, and the sample text keywords corresponding to the multiple sample texts are concatenated to obtain a text sample sequence.

[0008] The frequency of publication of the sample texts corresponding to the multiple sample texts is concatenated to obtain the frequency sample sequence.

[0009] The text sample sequence and the frequency sample sequence are input into the evaluation model to be trained, so that the text sample sequence and the frequency sample sequence are encoded by the evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence.

[0010] The text sample vector and the frequency sample vector are predicted and evaluated by the evaluation model to be trained to obtain the source evaluation prediction result.

[0011] Based on the difference between the source evaluation result label and the source evaluation prediction result, the parameters of the evaluation model to be trained are adjusted until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

[0012] The second aspect of this application provides a method for source assessment, including:

[0013] Obtain the target information source name corresponding to the target information source, and obtain the information source text title, information source text keywords and information source text publication frequency corresponding to the multiple information source texts published by the target information source;

[0014] The target source name, the source text titles corresponding to the multiple source texts, and the source text keywords corresponding to the multiple source texts are concatenated to obtain a source text sequence.

[0015] The frequency of source text publication corresponding to the multiple source texts is concatenated to obtain a source frequency sequence.

[0016] The source text sequence and the source frequency sequence are input into the source evaluation model to encode the source text sequence and the source frequency sequence through the source evaluation model, thereby obtaining the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence, wherein the source evaluation model is a model trained according to the training method of the source evaluation model described in the first aspect;

[0017] The source text vector and the source frequency vector are evaluated by the source evaluation model to obtain a target evaluation result, wherein the target evaluation result is used to indicate whether the target source is valid or invalid.

[0018] A third aspect of this application provides a training apparatus for a source evaluation model, comprising:

[0019] The source sample text acquisition unit is used to acquire the source sample name and source evaluation result label corresponding to the source sample, as well as the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample;

[0020] The text sample sequence acquisition unit is used to concatenate the source sample name, the sample text titles corresponding to the plurality of sample texts, and the sample text keywords corresponding to the plurality of sample texts to obtain a text sample sequence.

[0021] The frequency sample sequence acquisition unit is used to concatenate the publication frequencies of the sample texts corresponding to the plurality of sample texts to obtain the frequency sample sequence.

[0022] The sample sequence encoding processing unit is used to input the text sample sequence and the frequency sample sequence into the training and evaluation model, so as to encode the text sample sequence and the frequency sample sequence through the training and evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence.

[0023] The evaluation prediction result acquisition unit is used to perform prediction evaluation processing on the text sample vector and the frequency sample vector by the evaluation model to be trained, and obtain the source evaluation prediction result.

[0024] The source evaluation model acquisition unit is used to adjust the parameters of the evaluation model to be trained based on the difference between the source evaluation result label and the source evaluation prediction result, until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

[0025] A fourth aspect of this application provides a source evaluation device, comprising:

[0026] The target information source text acquisition unit is used to acquire the target information source name corresponding to the target information source, and to acquire the information source text title, information source text keywords and information source text publication frequency corresponding to the multiple information source texts published by the target information source;

[0027] The source text sequence acquisition unit is used to concatenate the target source name, the source text titles corresponding to the plurality of source texts, and the source text keywords corresponding to the plurality of source texts to obtain a source text sequence.

[0028] The source frequency sequence acquisition unit is used to concatenate the release frequencies of the source texts corresponding to the multiple source texts to obtain a source frequency sequence.

[0029] The source sequence encoding processing unit is used to input the source text sequence and the source frequency sequence into the source evaluation model, so as to encode the source text sequence and the source frequency sequence through the source evaluation model to obtain the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence, wherein the source evaluation model is a model trained according to the above-mentioned source evaluation model training method;

[0030] The target evaluation result acquisition unit is used to evaluate the source text vector and the source frequency vector by the source evaluation model to obtain the target evaluation result, wherein the target evaluation result is used to indicate whether the target source is valid or invalid.

[0031] A fifth aspect of this application provides a computer device, the device comprising a processor and a memory:

[0032] The memory is used to store computer programs and to transfer the computer programs to the processor;

[0033] The processor is configured to execute the steps of the training method for the source evaluation model provided in the first aspect, or the steps of the source evaluation method provided in the second aspect, according to the instructions in the computer program.

[0034] The sixth aspect of this application provides a computer-readable storage medium for storing a computer program, which, when executed by a computer device, implements the steps of the training method for the source evaluation model provided in the first aspect, or the steps of the source evaluation method provided in the second aspect.

[0035] The seventh aspect of this application provides a computer program product, including a computer program that, when executed by a computer device, implements the steps of the training method for the source evaluation model provided in the first aspect, or the steps of the source evaluation method provided in the second aspect.

[0036] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0037] The technical solution of this application first obtains the source sample name and source evaluation result label corresponding to the source sample, as well as the sample text title, sample text keywords, and sample text publication frequency corresponding to multiple sample texts published by the source sample. Then, the source sample name, the sample text titles, and the sample text keywords corresponding to multiple sample texts are concatenated to obtain a text sample sequence, and the sample text publication frequencies corresponding to multiple sample texts are concatenated to obtain a frequency sample sequence. During model training, the text sample sequence and frequency sample sequence can be input into the evaluation model to be trained. The evaluation model encodes the text sample sequence and frequency sample sequence to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence. The evaluation model then performs prediction and evaluation processing on the text sample vector and frequency sample vector to obtain the source evaluation prediction result. Finally, based on the difference between the source evaluation result label and the source evaluation prediction result, the parameters of the evaluation model to be trained are adjusted until the adjusted model meets the model training cutoff condition, and training ends, obtaining the source evaluation model.

[0038] As can be seen, in the technical solution of this application, before model training, a text sample sequence and a frequency sample sequence can be obtained based on the source sample name, the sample text titles, sample text keywords, and sample text publication frequency corresponding to multiple sample texts. During model training, the text sample sequence and frequency sample sequence can be processed by the evaluation model to be trained to obtain the source evaluation prediction result. Therefore, based on the difference between the source evaluation result label and the source evaluation prediction result, the model can be adjusted to obtain the final source evaluation model. Thus, in this application, a source evaluation model can be constructed based on the source sample name and the sample text titles, sample text keywords, and sample text publication frequency of the published sample texts, thereby enabling the evaluation of the source. Attached Figure Description

[0039] Figure 1 A scenario architecture diagram of a training method for a source evaluation model provided in an embodiment of this application;

[0040] Figure 2 A schematic diagram illustrating a training method for a source evaluation model in a practical application scenario, provided by an embodiment of this application;

[0041] Figure 3 A flowchart illustrating a training method for a source evaluation model provided in this application embodiment;

[0042] Figure 4 A schematic diagram illustrating a method for obtaining text sample sequences in the training of a source evaluation model provided in this application embodiment;

[0043] Figure 5 This diagram illustrates the publication frequency of multiple sample texts in a training method for a source evaluation model provided in this application embodiment.

[0044] Figure 6 A flowchart illustrating the encoding of sample sequences in a training method for a source evaluation model provided in this application embodiment;

[0045] Figure 7 A flowchart illustrating the prediction and evaluation process of sample vectors in a training method for a source evaluation model provided in this application embodiment;

[0046] Figure 8 A flowchart illustrating the entire process of model training in a training method for a source evaluation model provided in this application embodiment;

[0047] Figure 9 A flowchart of a source evaluation method provided in this application embodiment;

[0048] Figure 10 A flowchart illustrating the application of a source evaluation method provided in this application embodiment;

[0049] Figure 11 A schematic diagram of the structure of a training device for a source evaluation model provided in an embodiment of this application;

[0050] Figure 12 This is a schematic diagram of the structure of a source evaluation device provided in an embodiment of this application;

[0051] Figure 13 This is a schematic diagram of the server structure in an embodiment of this application;

[0052] Figure 14 This is a schematic diagram of the structure of a terminal device in an embodiment of this application. Detailed Implementation

[0053] The embodiments of this application will now be described with reference to the accompanying drawings.

[0054] As mentioned earlier, map text intelligence can efficiently and accurately provide navigation information to objects, primarily based on text published by reliable sources. However, with the explosive growth of objects on internet platforms, the quality of the text published by these objects varies greatly, leading to inaccuracies in the constructed map text intelligence. Understandably, the varying backgrounds, knowledge levels, and language abilities of objects can result in inconsistent text quality; some texts may be logically clear and correctly expressed, while others may be logically flawed, incorrectly expressed, or even contain false information. Furthermore, there is currently no technical solution to evaluate sources and select reliable ones for constructing map text intelligence. Therefore, how to evaluate sources has become a pressing technical problem in this field.

[0055] In view of the above problems, this application provides a training method for a source evaluation model, a source evaluation method, and related products, with the aim of evaluating sources. In the technical solution provided in this application, firstly, a text sample sequence and a frequency sample sequence are obtained based on the source sample name, the corresponding sample text titles, sample text keywords, and sample text publication frequency. Subsequently, during model training, the text sample sequence and frequency sample sequence are processed by the evaluation model to be trained to obtain source evaluation prediction results. Based on the difference between the source evaluation result label and the source evaluation prediction result, the model is adjusted to obtain the final source evaluation model. Thus, in this application, a source evaluation model can be constructed based on the source sample name and the sample text titles, sample text keywords, and sample text publication frequency of the published sample texts, thereby enabling the evaluation of sources and the selection of effective sources to construct map text intelligence.

[0056] The execution entity of the training method for the information source evaluation model provided in this application embodiment can be a terminal device. For example, the terminal device can obtain the information source sample name and information source evaluation result label corresponding to the information source sample. Alternatively, the execution entity of the information source evaluation method provided in this application embodiment can be a terminal device, such as obtaining the target information source name corresponding to the target information source on the terminal device. As an example, the terminal device may include, but is not limited to, mobile phones, desktop computers, tablet computers, laptops, handheld computers, smart voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc. The execution entity of the training method for the information source evaluation model provided in this application embodiment can also be a server. That is, the server can obtain the information source sample name and information source evaluation result label corresponding to the information source sample. Alternatively, the execution entity of the information source evaluation method provided in this application embodiment can also be a server. That is, the server can obtain the target information source name corresponding to the target information source on the server. Furthermore, the training method for the information source evaluation model provided in this application embodiment can also be executed collaboratively by the terminal device and the server, or the information source evaluation method provided in this application embodiment can also be executed collaboratively by the terminal device and the server. The terminal and the server can be directly or indirectly connected through wired or wireless communication, which is not limited herein. Therefore, the implementation subject of the technical solution of this application is not limited in the embodiments of this application.

[0057] Figure 1 An exemplary scenario architecture diagram of a training method for a source evaluation model is shown. The diagram includes a server and various types of terminal devices. First, the terminal devices can obtain the source sample name and source evaluation result label corresponding to the source sample, as well as the sample text titles, keywords, and publication frequencies corresponding to the multiple sample texts published by the source sample. Then, the server can construct the source evaluation model based on the source sample name, source evaluation result label, and the sample text titles, keywords, and publication frequencies corresponding to the multiple sample texts, thus enabling the evaluation of the source. Figure 1 The server shown can be a standalone physical server, a server cluster consisting of multiple physical servers, or a distributed system. Additionally, the server can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0058] To facilitate understanding of the technical solutions provided in the embodiments of this application, the training method of a source evaluation model provided in the embodiments of this application will be introduced next, in conjunction with a practical application scenario. See also Figure 2 , Figure 2This is a schematic diagram of a training method for a source evaluation model in a practical application scenario provided by an embodiment of this application. In this practical application scenario, the processing device is a server 200 with model training function.

[0059] First, server 200 can obtain the source sample name and source evaluation result label corresponding to the source sample, as well as the sample text title, sample text keywords, and sample text publication frequency corresponding to multiple sample texts published by the source sample. In the initial stage of model training, server 200 can concatenate the source sample name, the sample text titles corresponding to multiple sample texts, and the sample text keywords corresponding to multiple sample texts to obtain a text sample sequence. Similarly, server 200 can concatenate the sample text publication frequencies corresponding to multiple sample texts to obtain a frequency sample sequence. Subsequently, in the model training stage, server 200 can encode the text sample sequence and frequency sample sequence using the evaluation model to be trained to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence. The evaluation model then performs prediction and evaluation processing on the text sample vector and frequency sample vector to obtain the source evaluation prediction result. Finally, server 200 can adjust the parameters of the evaluation model to be trained based on the difference between the source evaluation result label and the source evaluation prediction result until the adjusted model meets the model training cutoff condition, at which point training ends and the source evaluation model is obtained. Therefore, in this application, the source sample name and the sample text title, sample text keywords, and sample text publication frequency of the published sample text can be used as features to construct a source evaluation model, thereby enabling the evaluation of the source.

[0060] See Figure 3 This figure is a flowchart of a training method for a source evaluation model provided in an embodiment of this application. Figure 3 The training method for the information source evaluation model shown includes the following steps:

[0061] S301: Obtain the source sample name and source evaluation result label corresponding to the source sample, and obtain the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample.

[0062] In this step, the source evaluation result label can be understood as the training label of the evaluation model to be trained. This label characterizes whether the source sample is valid, and it is a manually labeled label. Understandably, in this application, a source database can first be obtained, which includes the source sample names corresponding to the source samples. The source sample names corresponding to the source samples are then retrieved from this database. Here, the source sample can be understood as an object in the application that can publish text, such as a public account, and the source sample name can be understood as the name of the object in the application that can publish text, such as the name of a public account being XXX. It should be noted that the source samples and their corresponding sample names existing in the source database in this application are only partial examples, and can be further determined in practical applications.

[0063] Furthermore, this application utilizes web crawling technology to obtain multiple sample texts published by the source sample. These multiple sample texts can be sample texts published by the source sample within a preset time period, where the preset time period can be one month. The predicted time period shown in this application is only a partial example, thus enabling a more accurate construction of the source evaluation model. The multiple sample texts published by the source sample can be multiple sample texts on the same topic, or multiple sample texts on different topics. For example, multiple sample texts can be multiple sample texts under topic A, or multiple sample texts under topic A and topic B. It should be noted that the sample texts shown in this application are only partial examples, and the number of sample texts is not limited here.

[0064] It should also be noted that the data collection and processing described in this application should strictly comply with the requirements of relevant national laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0065] Subsequently, in this application, multiple sample texts published by the source sample can be parsed to obtain the sample text titles, sample text keywords, and sample text publication times corresponding to the multiple sample texts published by the source sample. That is, each sample text has a corresponding sample text title, sample text keywords, and sample text publication time. The sample text title can be understood as the title of the sample text, the sample text publication time can be understood as the publication time of the sample text, and the sample text keywords can be understood as keywords that can characterize the core content of the sample text.

[0066] Specifically, in this application, for each sample text, keyword extraction techniques can be used to extract keywords from the sample text to obtain the corresponding sample keywords. These keyword extraction techniques include the TFIDF algorithm and the TextRank algorithm. It should be noted that, in this application, for each sample text, the number of sample keywords obtained can be one, two, or three. The examples shown here are only partial examples, and the selection of sample keywords can be based on the highest score and ranking.

[0067] In this application, for each sample text, the publication time of the sample text can be statistically analyzed to obtain the publication frequency of the sample text. It is understood that in this application, based on the publication time of the sample text of the source sample within a preset time period, the publication time of the previous sample text of the source sample can be determined. Thus, the publication frequency of the sample text can be determined based on the publication time of the sample text and the publication time of the previous sample text. That is, the publication frequency of the sample text is the publication time interval of the sample text, where a publication time interval less than one day is calculated as one day.

[0068] S302: The source sample name, the sample text titles corresponding to the multiple sample texts, and the sample text keywords corresponding to the multiple sample texts are concatenated to obtain a text sample sequence.

[0069] In this step, the multiple sample texts include a first sample text and a second sample text. The sample text titles corresponding to each of the multiple sample texts include the first sample text title corresponding to the first sample text and the second sample text title corresponding to the second sample text. The sample text keywords corresponding to each of the multiple sample texts include the first sample text keyword corresponding to the first sample text and the second sample text keyword corresponding to the second sample text. It should be noted that, in this application, for the purpose of more concisely illustrating the construction process of the information source evaluation model, the number of sample texts and the number of sample text keywords corresponding to each sample text shown here are only partial examples. In the actual model construction process, the number of sample texts and the number of sample text keywords can be adjusted according to actual needs.

[0070] Specifically, in this application, the source sample name, the first sample text title, and the first sample text keywords can be horizontally concatenated sequentially according to a preset concatenation order to obtain the first sample sequence. Then, the first sample sequence, the second sample text title, and the second sample text keywords can be horizontally concatenated sequentially to obtain the text sample sequence. The preset concatenation order can be name-title1-keyword1-…-titlen-keywordn. This method of generating the text sample sequence according to the preset concatenation order in this application facilitates faster data understanding and processing by the subsequent model.

[0071] It should be noted that this application allows for the addition of delimiters and classifiers to each unit during the horizontal concatenation process. The source sample name can be considered a unit, the first sample text title can be considered a unit, the first sample text keywords can be considered a unit, the second sample text title can be considered a unit, and the second sample text keywords can be considered a unit. Delimiters, including [SEP], are used to separate different units and are typically located at the end of the unit. Classifiers, including [CLS], are used to classify different units and are typically located at the beginning of the unit. By adding delimiters and classifiers, subsequent model extraction of data can be facilitated more quickly.

[0072] like Figure 4 As shown, Figure 4 A schematic diagram illustrating the method for obtaining text sample sequences in the training of a source evaluation model provided in this application embodiment (as needed, Figure 4 (For example only). Figure 4 The source sample name corresponding to the source sample is sample name K. The multiple sample texts published by the source sample are sample text A, sample text B, sample text ..., sample text N. The sample text title corresponding to sample text A is sample text title A, the sample text title corresponding to sample text B is sample text title B, the sample text keyword corresponding to sample text A is sample text keyword A, and the sample text keyword corresponding to sample text B is sample text keyword B.

[0073] At this point, the sample name K, sample text title A, sample text keyword A, sample text title ..., sample text keyword ..., sample text title N, and sample text keyword N can be horizontally concatenated in a preset concatenation order to obtain a text sample sequence. Each unit in the text sample sequence can be added with a delimiter and a classifier, which facilitates the construction of the model in the subsequent process.

[0074] S303: The frequency of publication of the sample texts corresponding to the multiple sample texts is concatenated to obtain the frequency sample sequence.

[0075] In this step, the publication frequency of the sample texts corresponding to the multiple sample texts includes the publication frequency of the first sample text corresponding to the first sample text and the publication frequency of the second sample text corresponding to the second sample text. For example... Figure 5 As shown, Figure 5 A schematic diagram illustrating the publication frequency of multiple sample texts in a training method for a source evaluation model provided in this application embodiment (as needed, Figure 5 (For example only). Figure 5 The multiple sample texts released by CITIC Source are Sample Text A, Sample Text B, Sample Text ..., Sample Text N. The sample text corresponding to Sample Text A is released once a day, and the sample text corresponding to Sample Text B is released twice a day.

[0076] It should be noted that before performing the concatenation process on the publication frequency of multiple sample texts to obtain the frequency sample sequence, this application may also record the concatenation order of the title and keywords of the first sample text in the text sample sequence as the first sample order, and record the concatenation order of the title and keywords of the second sample text in the text sample sequence as the second sample order.

[0077] Specifically, in this application, the publication frequency of the first sample text can be encoded using one-hot encoding to obtain a first initial sample vector corresponding to the publication frequency of the first sample text, and the publication frequency of the second sample text can be encoded using one-hot encoding to obtain a second initial sample vector corresponding to the publication frequency of the second sample text. Afterwards, the first and second initial sample vectors can be sorted according to the first sample order and the second sample order to obtain a frequency sample sequence. It can be understood that the first sample order represents the sorting order of the first sample text, and the second sample order represents the sorting order of the second sample text. Thus, by sorting the initial sample vectors according to the first and second sample orders in this application, the sorting of sample text in the frequency sample sequence is the same as the sorting of sample text in the text sample sequence, thereby facilitating rapid data extraction during subsequent model construction and ensuring the accuracy of the extracted data.

[0078] S304: Input the text sample sequence and the frequency sample sequence into the training evaluation model to encode the text sample sequence and the frequency sample sequence through the training evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence.

[0079] In this step, the model to be trained and evaluated includes a vector transformation model and a frequency embedding layer. The vector transformation model can be a BERT pre-trained model (specifically, chinese-roberta-wwm-ext), a word2vec model, or a T5 model. The choice of vector transformation model can be determined in practical applications and is not specifically limited here. Figure 6 As shown, Figure 6 The flowchart of the sample sequence encoding in the training method of the information source evaluation model provided in this application embodiment is as follows: Figure 6 Steps S3041 and S3042 are shown below. The specific details of steps S3041 and S3042 are as follows:

[0080] S3041: Input the text sample sequence into the vector transformation model of the model to be trained and evaluated, so as to transform the text sample sequence through the vector transformation model of the model to be trained and evaluated, and obtain the text sample vector corresponding to the text sample sequence.

[0081] In this step, the vector transformation model of the model to be trained and evaluated can be used to perform vector transformation on the source sample name, the first sample text title, the first sample text keyword, the second sample text title and the second sample text keyword in the text sample sequence, respectively, to obtain the sample name vector corresponding to the source sample name, the first title vector corresponding to the first sample text title, the first keyword vector corresponding to the first sample text keyword, the second title vector corresponding to the second sample text title and the second keyword vector corresponding to the second sample text keyword.

[0082] Understandably, in this application, the vector transformation model of the model to be trained and evaluated can perform vector transformation on each character in the source sample name to obtain the name vector corresponding to each character in the source sample name, and use the name vector corresponding to each character in the source sample name as the sample name vector corresponding to the source sample name. It should be noted that the vector transformation process of the first sample text title, the first sample text keyword, the second sample text title and the second sample text keyword is similar to the vector transformation process of the source sample name, and will not be repeated here.

[0083] Subsequently, in this application, the sample name vector corresponding to the source sample name, the first title vector corresponding to the first sample text title, the first keyword vector corresponding to the first sample text keyword, the second title vector corresponding to the second sample text title, and the second keyword vector corresponding to the second sample text keyword can be used as the text sample vector corresponding to the text sample sequence. It should be noted that after obtaining the text sample vector corresponding to the text sample sequence, this application can also record the position of each vector in the text sample vector to facilitate rapid vector extraction by the subsequent model. Thus, by using a vector transformation model to perform vector transformation operations in this application, the deep semantic features of the text can be captured more accurately, thereby improving the model's predictive performance.

[0084] S3042: Input the frequency sample sequence into the frequency embedding layer of the model to be trained and evaluated, so as to map the frequency sample sequence through the frequency embedding layer of the model to be trained and evaluated, and obtain the frequency sample vector corresponding to the frequency sample sequence.

[0085] In this step, the first initial sample vector in the frequency sample sequence can be vector-mapped through the frequency embedding layer of the model to be trained and evaluated to obtain the first frequency vector corresponding to the first initial sample vector, and the second initial sample vector in the frequency sample sequence can be vector-mapped through the frequency embedding layer of the model to be trained and evaluated to obtain the second frequency vector corresponding to the second initial sample vector.

[0086] Understandably, since the vector space of each vector executed in the model to be trained and evaluated is set to 768 dimensions, and the first and second initial sample vectors are obtained based on one-hot encoding, the vector spaces of the first and second initial sample vectors may be sparse during model training. In this case, the frequency embedding layer of the model to be trained and evaluated can be used to transform the first and second initial sample vectors respectively, so that the vector spaces of the transformed first and second frequency vectors are both 768 dimensions.

[0087] Subsequently, the first frequency vector corresponding to the first initial sample vector and the second frequency vector corresponding to the second initial sample vector can be used as the frequency sample vector corresponding to the frequency sample sequence, wherein the order of the sample text in the frequency sample vector is the same as the order of the sample text in the frequency sample sequence. Thus, by using the frequency embedding layer to perform vector transformation operations in this application, the identical vector space of the vectors in the model can be guaranteed, thereby facilitating the prediction performance of subsequent models.

[0088] S305: The evaluation model to be trained performs prediction and evaluation processing on the text sample vector and the frequency sample vector to obtain the source evaluation prediction result.

[0089] In this step, the model to be trained and evaluated also includes a max pooling layer and an evaluation classification layer. For example... Figure 7 As shown, Figure 7 This application provides a flowchart of the sample vector prediction and evaluation process in a training method for a source evaluation model, as illustrated in an embodiment of the present application. Figure 7 Steps S3051 and S3052 are shown below. The specific details of steps S3051 and S3052 are as follows:

[0090] S3051: The evaluation model to be trained performs dimensional transformation processing on the first title vector, the first keyword vector and the first frequency vector in the frequency sample vector of the text sample vector to obtain the first sample vector, and performs dimensional transformation processing on the second title vector, the second keyword vector and the second frequency vector in the frequency sample vector of the text sample vector to obtain the second sample vector.

[0091] It should be noted that the model to be trained and evaluated also includes a dimension transformation layer, which can be understood as a fully connected layer. In this step, the model to be trained and evaluated can horizontally concatenate the first title vector, the first keyword vector, and the first frequency vector from the frequency sample vector in the text sample vector to obtain the first concatenated vector, and horizontally concatenate the second title vector, the second keyword vector, and the second frequency vector from the frequency sample vector in the text sample vector to obtain the second concatenated vector. Understandably, at this point, the model to be trained and evaluated can perform concatenation processing on the title vector, keyword vector, and frequency vector related to each sample text to achieve the combination of relevant information for each sample text.

[0092] Subsequently, in this application, the dimension transformation layer in the model to be trained and evaluated can perform dimension transformation on the first concatenated vector to obtain the first sample vector, and the dimension transformation layer in the model to be trained and evaluated can perform dimension transformation on the second concatenated vector to obtain the second sample vector. It is understood that since the vector spaces corresponding to the first title vector, first keyword vector, first frequency vector, second title vector, second keyword vector, and second frequency vector obtained through the vector transformation model are each 768-dimensional, and the vector spaces corresponding to the first concatenated vector and the second concatenated vector are both 2304-dimensional (i.e., 768 × 3-dimensional), the dimension transformation layer in the model to be trained and evaluated can convert the first concatenated vector and the second concatenated vector into a first sample vector and a second sample vector with a vector space of 768 dimensions, respectively. Thus, in this application, the dimension transformation of the vector space can be achieved under the action of the dimension transformation layer, ensuring that the vector dimension of each concatenated vector is 768 dimensions.

[0093] S3052: The evaluation model to be trained performs prediction and evaluation processing on the sample name vector, the first sample vector, and the second sample vector in the text sample vector to obtain the source evaluation prediction result.

[0094] It should be noted that the evaluation model to be trained also includes a max pooling layer and an evaluation classification layer, where the evaluation classification layer can be understood as a fully connected layer containing a softmax activation function. In this step, the max pooling layer in the evaluation model first aggregates the first and second sample vectors to obtain an aggregated sample vector. Then, the evaluation classification layer in the evaluation model horizontally concatenates the sample name vector from the text sample vector with the aggregated sample vector to obtain the vector to be predicted. Finally, the evaluation classification layer in the evaluation model predicts and evaluates this vector to obtain the source evaluation prediction result, which can be understood as the source evaluation prediction probability. Thus, in this application, the source sample name and the sample text title, sample text keywords, and sample text publication frequency of the published sample text can be used as features to output the model's prediction result, thereby achieving the evaluation of the source.

[0095] S306: Based on the difference between the source evaluation result label and the source evaluation prediction result, adjust the parameters of the evaluation model to be trained until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

[0096] Since the source evaluation result label indicates whether the source sample is valid, if the source evaluation result label indicates that the source sample is valid, then the source evaluation result label can be represented as 1; if the source evaluation result label indicates that the source sample is invalid, then the source evaluation result label can be represented as 0. The source evaluation prediction result can be understood as the source evaluation prediction probability. Based on the difference between the source evaluation result label and the source evaluation prediction result, the parameters of the evaluation model to be trained can be adjusted until the adjusted model meets the model training cutoff condition, and training ends, obtaining the source evaluation model. Thus, this application achieves model construction in a supervised manner, enabling the trained source evaluation model to better adapt to map road-related business scenarios, thereby realizing the evaluation of sources.

[0097] In one feasible implementation, the difference between the source assessment result label and the source assessment prediction result can be represented by a loss calculation formula, which is specifically described as follows:

[0098] L=-∑y i logp i

[0099] Where L represents the cross-entropy loss function (i.e., the difference between the source evaluation result label and the source evaluation prediction result), y i Indicates the label of the information source assessment result, p i represents the prediction result of the information source evaluation (i.e. the probability that the evaluation model to be trained predicts class i), where i represents valid (1) or invalid (0).

[0100] like Figure 8 As shown, Figure 8 A flowchart illustrating the entire process of model training in a training method for a source evaluation model provided in this application (as needed, ...). Figure 8 (For example only). Figure 8 First, the text sample sequence (i.e., sample name K, sample text title A, sample text keyword A, sample text title ..., sample text keyword ..., sample text title N, sample text keyword N) can be input into the vector transformation model of the model to be trained and evaluated. The vector transformation model of the model to be trained and evaluated will then perform a transformation operation to obtain the text sample vector corresponding to the text sample sequence. Simultaneously, the frequency sample sequence can be input into the frequency embedding layer of the model to be trained and evaluated. The frequency embedding layer of the model to be trained and evaluated will then perform a mapping operation to obtain the frequency sample vector corresponding to the frequency sample sequence.

[0101] Taking sample text A as an example, the title vector, keyword vector, and frequency vector corresponding to sample text A can be concatenated to obtain the concatenated vector corresponding to sample text A. The concatenated vector corresponding to sample text A can then be converted into the sample vector corresponding to sample text A through the dimension transformation layer of the model to be trained and evaluated. It should be noted that the execution process of sample text ... and sample text N is similar to that of sample text A, and will not be elaborated here.

[0102] Subsequently, in this application, the sample vectors corresponding to multiple sample texts can be aggregated through the max pooling layer of the evaluation model to be trained, resulting in an aggregated sample vector. The sample name vector of the source sample and the aggregated sample vector are then horizontally concatenated to obtain the vector to be predicted. Finally, the evaluation classification layer in the evaluation model to be trained can predict and evaluate this vector to obtain the source evaluation prediction result. (Alternatively, in this application, the evaluation classification layer in the evaluation model to be trained can horizontally concatenate the sample name vector and the aggregated sample vector in the text sample vector to obtain the vector to be predicted, and then the evaluation classification layer in the evaluation model to be trained can predict and evaluate this vector to obtain the source evaluation prediction result.) Thus, in this application, a source evaluation model can be constructed based on the source sample name and the sample text title, sample text keywords, and sample text publication frequency of the published sample text as features, thereby enabling the evaluation of the source.

[0103] See Figure 9 This figure is a flowchart of a source evaluation method provided in an embodiment of this application. Figure 9 The source assessment method shown includes the following steps:

[0104] S901: Obtain the target source name corresponding to the target source, and obtain the source text title, source text keywords and source text publication frequency corresponding to the multiple source texts published by the target source.

[0105] In this step, the target information source can be understood as the information source to be predicted and evaluated. It should be noted that the process of obtaining the information source text title, information source text keywords, and information source text publication frequency in this step is similar to the process of obtaining the sample text title, sample text keywords, and sample text publication frequency in the steps above, and will not be described in detail here.

[0106] S902: The target source name, the source text titles corresponding to the multiple source texts, and the source text keywords corresponding to the multiple source texts are concatenated to obtain a source text sequence.

[0107] It should be noted that the process of concatenating the target source name, multiple source text titles, and multiple source text keywords in this step is similar to the process of concatenating the source sample name, multiple sample text titles, and multiple sample text keywords in the steps above, and will not be described in detail here.

[0108] S903: The frequency of the source texts corresponding to the multiple source texts are concatenated to obtain a source frequency sequence.

[0109] It should be noted that the process of splicing the publication frequencies of multiple source texts in this step is similar to the process of splicing the publication frequencies of multiple sample texts in the above steps, and will not be described in detail here.

[0110] S904: Input the source text sequence and the source frequency sequence into the source evaluation model, so as to encode the source text sequence and the source frequency sequence through the source evaluation model to obtain the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence.

[0111] It should be noted that the source evaluation model is the model trained using the training method described above. The encoding process for the source text sequence and the source frequency sequence in this step is similar to the encoding process for the text sample sequence and the frequency sample sequence in the above steps, and will not be described in detail here.

[0112] S905: The source text vector and the source frequency vector are evaluated by the source evaluation model to obtain the target evaluation result.

[0113] It should be noted that the target evaluation result is used to indicate whether the target information source is valid or invalid. The evaluation process for the information source text vector and information source frequency vector in this step is similar to the prediction and evaluation process for the text sample vector and frequency sample vector in the above steps, and will not be described in detail here.

[0114] like Figure 10 As shown, Figure 10This application provides a flowchart of an information source evaluation method. First, the target information source name is obtained from the information source database. Then, based on the target information source name, the titles, keywords, and publication frequencies of multiple source texts published by the target information source are determined. Based on these parameters, a source text sequence and a source frequency sequence are determined. Finally, an information source evaluation model is used to evaluate the source text sequence and the source frequency sequence to obtain the target evaluation result. This achieves the evaluation of information sources, enabling the selection of effective information sources and thus facilitating the construction of map text intelligence.

[0115] It should also be noted that if the target evaluation result indicates that the target information source is invalid, the remaining information sources can continue to be evaluated. Conversely, if the target evaluation result indicates that the target information source is valid, the multiple source texts published by the target information source can be filtered to obtain source map texts containing map elements. These multiple source texts published by the target information source are those published by the target information source within a preset time period, which can be one month. It is understood that in this application, if the target evaluation result indicates that the target information source is valid, the multiple source texts obtained within the preset time period can be filtered to obtain texts related to map roads (i.e., source map texts) from among the multiple source texts.

[0116] Furthermore, this application can obtain information such as traffic restrictions, closures, and openings of a road at a certain time from the source map text. Subsequently, the target map can be updated based on this source map text to obtain an updated target map, which can provide effective navigation information for the target audience (such as drivers). Thus, by evaluating and processing the source information in this application, effective sources can be effectively selected, thereby constructing map text intelligence and providing accurate navigation services for the target audience's travel, improving their experience.

[0117] Furthermore, it should be noted that related technologies include methods for evaluating the profiles of objects and methods for evaluating the profiles of objects on e-commerce platforms. The object profiling method primarily constructs a profile based on information such as the object's age, school, and occupation. The e-commerce platform profiling method primarily constructs a profile based on information such as the object's product purchases, product logistics, and product reviews on the e-commerce platform. It is evident that the two evaluation methods used in the aforementioned related technologies can only be used in their specific domains to evaluate object profiles. The technical solution of this application differs from these related technical solutions. The information source evaluation method proposed in this application is applied to the field of map information source evaluation to support the construction of map text intelligence. Specifically, it evaluates information sources by analyzing information such as the text title, keywords, and frequency of the information source. This allows for the evaluation of effective information sources to further update map data, thereby effectively constructing map text intelligence. In summary, in the embodiments of this application, at the beginning of model training, a text sample sequence and a frequency sample sequence can be obtained based on the source sample name, the sample text titles, sample text keywords, and sample text publication frequency corresponding to multiple sample texts. Subsequently, during model training, the text sample sequence and frequency sample sequence can be processed by the evaluation model to obtain the source evaluation prediction result. Based on the difference between the source evaluation result label and the source evaluation prediction result, the model can be adjusted to obtain the final source evaluation model. Thus, in this application, a source evaluation model can be constructed based on the source sample name and the sample text titles, sample text keywords, and sample text publication frequency of the published sample texts. This enables the evaluation of source information and, in the process of model application, the selection of effective source information, thereby achieving the construction of map text intelligence.

[0118] Based on the training method of the source evaluation model provided in the preceding embodiments, this application also provides a training device for the source evaluation model. The training device for the source evaluation model provided in the embodiments of this application will be described in detail below.

[0119] See Figure 11 This figure is a schematic diagram of the structure of a training device for a source evaluation model provided in an embodiment of this application. Figure 11 As shown, the training device for this information source evaluation model specifically includes:

[0120] The source sample text acquisition unit 1101 is used to acquire the source sample name and source evaluation result label corresponding to the source sample, as well as the sample text title, sample text keywords and sample text release frequency corresponding to the multiple sample texts published by the source sample.

[0121] The text sample sequence obtaining unit 1102 is used to concatenate the source sample name, the sample text titles corresponding to the plurality of sample texts, and the sample text keywords corresponding to the plurality of sample texts to obtain a text sample sequence.

[0122] The frequency sample sequence acquisition unit 1103 is used to concatenate the publication frequencies of the sample texts corresponding to the plurality of sample texts to obtain the frequency sample sequence.

[0123] The sample sequence encoding processing unit 1104 is used to input the text sample sequence and the frequency sample sequence into the training evaluation model, so as to encode the text sample sequence and the frequency sample sequence through the training evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence.

[0124] The evaluation prediction result acquisition unit 1105 is used to perform prediction evaluation processing on the text sample vector and the frequency sample vector by the evaluation model to be trained, and obtain the source evaluation prediction result.

[0125] The source evaluation model acquisition unit 1106 is used to adjust the parameters of the evaluation model to be trained based on the difference between the source evaluation result label and the source evaluation prediction result, until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

[0126] In one feasible implementation, the text sample sequence obtaining unit 1102 is specifically used for:

[0127] The source sample name, the first sample text title, and the first sample text keywords are horizontally concatenated to obtain the first sample sequence;

[0128] The first sample sequence, the second sample text title, and the second sample text keywords are horizontally concatenated to obtain a text sample sequence.

[0129] In one feasible implementation, the device further includes:

[0130] The sample order recording unit is used to record the concatenation order of the first sample text title and the first sample text keyword in the text sample sequence as the first sample order, and to record the concatenation order of the second sample text title and the second sample text keyword in the text sample sequence as the second sample order;

[0131] The frequency sample sequence acquisition unit 1103 is specifically used for:

[0132] The publication frequency of the first sample text and the publication frequency of the second sample text are encoded to obtain a first initial sample vector corresponding to the publication frequency of the first sample text and a second initial sample vector corresponding to the publication frequency of the second sample text.

[0133] Based on the first sample order and the second sample order, the first initial sample vector and the second initial sample vector are sorted to obtain a frequency sample sequence.

[0134] In one feasible implementation, the sample sequence encoding processing unit 1104 includes:

[0135] The text sample vector acquisition unit is used to input the text sample sequence into the vector transformation model of the model to be trained and evaluated, so as to transform the text sample sequence through the vector transformation model of the model to be trained and evaluated, and obtain the text sample vector corresponding to the text sample sequence.

[0136] The frequency sample vector acquisition unit is used to input the frequency sample sequence into the frequency embedding layer of the model to be trained and evaluated, so as to map the frequency sample sequence through the frequency embedding layer of the model to be trained and evaluated to obtain the frequency sample vector corresponding to the frequency sample sequence.

[0137] In one feasible implementation, the text sample vector acquisition unit is specifically used for:

[0138] Using the vector transformation model of the model to be trained and evaluated, vector transformation is performed on the source sample name, first sample text title, first sample text keyword, second sample text title and second sample text keyword in the text sample sequence to obtain the sample name vector corresponding to the source sample name, the first title vector corresponding to the first sample text title, the first keyword vector corresponding to the first sample text keyword, the second title vector corresponding to the second sample text title and the second keyword vector corresponding to the second sample text keyword.

[0139] The sample name vector corresponding to the source sample name, the first title vector corresponding to the first sample text title, the first keyword vector corresponding to the first sample text keyword, the second title vector corresponding to the second sample text title, and the second keyword vector corresponding to the second sample text keyword are used as the text sample vector corresponding to the text sample sequence.

[0140] In one feasible implementation, the frequency sample vector acquisition unit is specifically used for:

[0141] Through the frequency embedding layer of the model to be trained and evaluated, the first initial sample vector and the second initial sample vector in the frequency sample sequence are vector mapped to obtain the first frequency vector corresponding to the first initial sample vector and the second frequency vector corresponding to the second initial sample vector.

[0142] The first frequency vector corresponding to the first initial sample vector and the second frequency vector corresponding to the second initial sample vector are used as the frequency sample vectors corresponding to the frequency sample sequence.

[0143] In one feasible implementation, the evaluation prediction result obtaining unit 1105 includes:

[0144] The dimension transformation processing unit is used to perform dimension transformation processing on the first title vector, the first keyword vector and the first frequency vector in the frequency sample vector of the text sample vector by the evaluation model to be trained, to obtain the first sample vector, and to perform dimension transformation processing on the second title vector, the second keyword vector and the second frequency vector in the frequency sample vector of the text sample vector, to obtain the second sample vector.

[0145] The prediction and evaluation processing unit is used to perform prediction and evaluation processing on the sample name vector, the first sample vector, and the second sample vector in the text sample vector by the evaluation model to be trained, so as to obtain the source evaluation prediction result.

[0146] In one feasible implementation, the prediction and evaluation processing unit is specifically used for:

[0147] The first sample vector and the second sample vector are aggregated by the max pooling layer in the model to be trained and evaluated to obtain an aggregated sample vector.

[0148] The evaluation classification layer in the evaluation model to be trained predicts and evaluates the sample name vector and the aggregated sample vector in the text sample vector to obtain the source evaluation prediction result.

[0149] In one feasible implementation, the dimension transformation processing unit is specifically used for:

[0150] The training evaluation model concatenates the first title vector, the first keyword vector, and the first frequency vector in the frequency sample vector of the text sample vector to obtain a first concatenated vector; and concatenates the second title vector, the second keyword vector, and the second frequency vector in the frequency sample vector of the text sample vector to obtain a second concatenated vector.

[0151] The dimension transformation layer in the model to be trained and evaluated performs dimension transformation on the first concatenated vector to obtain a first sample vector, and performs dimension transformation on the second concatenated vector to obtain a second sample vector.

[0152] The training apparatus for the source evaluation model provided in this application embodiment has the same beneficial effects as the training method for the source evaluation model provided in the above embodiments, and therefore will not be described again.

[0153] Based on the source evaluation method provided in the preceding embodiments, this application also provides a source evaluation device. The source evaluation device provided in the embodiments of this application will be described in detail below.

[0154] See Figure 12 This figure is a schematic diagram of the structure of a source evaluation device provided in an embodiment of this application. Figure 11 As shown, the information source evaluation device specifically includes:

[0155] The target information source text acquisition unit 1201 is used to acquire the target information source name corresponding to the target information source, and to acquire the information source text title, information source text keywords and information source text publication frequency corresponding to the multiple information source texts published by the target information source;

[0156] The source text sequence obtaining unit 1202 is used to concatenate the target source name, the source text titles corresponding to the plurality of source texts, and the source text keywords corresponding to the plurality of source texts to obtain a source text sequence.

[0157] The source frequency sequence acquisition unit 1203 is used to concatenate the source text publication frequencies corresponding to the multiple source texts to obtain a source frequency sequence.

[0158] The source sequence encoding processing unit 1204 is used to input the source text sequence and the source frequency sequence into the source evaluation model, so as to encode the source text sequence and the source frequency sequence through the source evaluation model to obtain the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence, wherein the source evaluation model is a model trained according to the above-mentioned source evaluation model training method;

[0159] The target evaluation result obtaining unit 1205 is used to evaluate the source text vector and the source frequency vector by the source evaluation model to obtain the target evaluation result, wherein the target evaluation result is used to indicate whether the target source is valid or invalid.

[0160] In one feasible implementation, the device further includes:

[0161] A text filtering and processing unit is used to filter multiple source texts published by the target source if the target evaluation result indicates that the target source is valid, and to obtain source map texts containing map elements in the multiple source texts.

[0162] The map update processing unit is used to update the target map based on the source map text to obtain an updated target map, wherein the updated target map is used to provide effective navigation information. The source evaluation apparatus provided in this application embodiment has the same beneficial effects as the source evaluation method provided in the above embodiments, and therefore will not be described again.

[0163] This application provides a computer device, which can be a server. Figure 13 This is a schematic diagram of a server structure provided in an embodiment of this application. The server 900 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 922 (e.g., one or more processors) and memory 932, and one or more storage media 930 (e.g., one or more mass storage devices) for storing application programs 942 or data 944. The memory 932 and storage media 930 can be temporary or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the server. Furthermore, the CPU 922 may be configured to communicate with the storage media 930 and execute the series of instruction operations in the storage media 930 on the server 900.

[0164] Server 900 may also include one or more power supplies 926, one or more wired or wireless network interfaces 950, one or more input / output interfaces 958, and / or one or more operating systems 941, such as Windows Server. TM Mac OS X TM Unix TM Linux TM FreeBSD TM etc.

[0165] CPU 922 is used to perform the following steps:

[0166] Obtain the source sample name and source evaluation result label corresponding to the source sample, and obtain the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample;

[0167] The source sample name, the sample text titles corresponding to the multiple sample texts, and the sample text keywords corresponding to the multiple sample texts are concatenated to obtain a text sample sequence.

[0168] The frequency of publication of the sample texts corresponding to the multiple sample texts is concatenated to obtain the frequency sample sequence.

[0169] The text sample sequence and the frequency sample sequence are input into the evaluation model to be trained, so that the text sample sequence and the frequency sample sequence are encoded by the evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence.

[0170] The text sample vector and the frequency sample vector are predicted and evaluated by the evaluation model to be trained to obtain the source evaluation prediction result.

[0171] Based on the difference between the source evaluation result label and the source evaluation prediction result, the parameters of the evaluation model to be trained are adjusted until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

[0172] or;

[0173] Obtain the target information source name corresponding to the target information source, and obtain the information source text title, information source text keywords and information source text publication frequency corresponding to the multiple information source texts published by the target information source;

[0174] The target source name, the source text titles corresponding to the multiple source texts, and the source text keywords corresponding to the multiple source texts are concatenated to obtain a source text sequence.

[0175] The frequency of source text publication corresponding to the multiple source texts is concatenated to obtain a source frequency sequence.

[0176] The source text sequence and the source frequency sequence are input into the source evaluation model to encode the source text sequence and the source frequency sequence through the source evaluation model, thereby obtaining the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence. The source evaluation model is a model trained according to the above-described source evaluation model training method.

[0177] The source text vector and the source frequency vector are evaluated by the source evaluation model to obtain a target evaluation result, wherein the target evaluation result is used to indicate whether the target source is valid or invalid.

[0178] This application also provides another computer device, which can be a terminal device. For example... Figure 14 As shown, for ease of explanation, only the parts related to the embodiments of this application are shown. For specific technical details not disclosed, please refer to the method section of the embodiments of this application. Taking a mobile phone as an example:

[0179] Figure 14 The diagram shown is a block diagram of a portion of the structure of a mobile phone provided in an embodiment of this application. (Reference) Figure 14 The mobile phone includes: a radio frequency (RF) circuit 1010, a memory 1020, an input unit 1030, a display unit 1040, a sensor 1050, an audio circuit 1060, a wireless fidelity (WiFi) module 1070, a processor 1080, and a power supply 1090, etc. Those skilled in the art will understand that... Figure 14 The mobile phone structure shown does not constitute a limitation on the mobile phone and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0180] The following is combined with Figure 14 A detailed introduction to each component of a mobile phone:

[0181] The RF circuit 1010 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink information from the base station and processes it with the processor 1080; additionally, it transmits uplink data to the base station. Typically, the RF circuit 1010 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low-noise amplifier (LNA), a duplexer, etc. Furthermore, the RF circuit 1010 can also communicate wirelessly with networks and other devices. The aforementioned wireless communications may use any communication standard or protocol, including but not limited to Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, and Short Messaging Service (SMS).

[0182] The memory 1020 can be used to store software programs and modules. The processor 1080 executes various mobile phone functions and data processing by running the software programs and modules stored in the memory 1020. The memory 1020 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 1020 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0183] The input unit 1030 can be used to receive input numerical or character information, and to generate key signal inputs related to user settings and function control of the mobile phone. Specifically, the input unit 1030 may include a touch panel 1031 and other input devices 1032. The touch panel 1031, also known as a touch screen, can collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel 1031), and drive the corresponding connection devices according to a pre-set program. Optionally, the touch panel 1031 may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch position and the signal generated by the touch operation, and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends it to the processor 1080, and can also receive and execute commands sent by the processor 1080. In addition, the touch panel 1031 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch panel 1031, the input unit 1030 may also include other input devices 1032. Specifically, other input devices 1032 may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0184] The display unit 1040 can be used to display information input by the user or information provided to the user, as well as various menus of the mobile phone. The display unit 1040 may include a display panel 1041, which may optionally be configured as a Liquid Crystal Display (LCD), Organic Light-Emitting Diode (OLED), or similar display panel 1041. Further, a touch panel 1031 may cover the display panel 1041. When the touch panel 1031 detects a touch operation on or near it, it transmits the information to the processor 1080 to determine the type of touch event. Subsequently, the processor 1080 provides corresponding visual output on the display panel 1041 according to the type of touch event. Although in Figure 14 In this embodiment, the touch panel 1031 and the display panel 1041 are two separate components to realize the input and output functions of the mobile phone. However, in some embodiments, the touch panel 1031 and the display panel 1041 can be integrated to realize the input and output functions of the mobile phone.

[0185] The mobile phone may also include at least one sensor 1050, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display panel 1041 according to the ambient light level, and the proximity sensor can turn off the display panel 1041 and / or the backlight when the phone is moved to the ear. As a type of motion sensor, an accelerometer sensor can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity and can be used for applications that recognize the phone's posture (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition-related functions (such as pedometer, taps), etc. Other sensors that may be configured in the mobile phone, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be described in detail here.

[0186] The audio circuit 1060, speaker 1061, and microphone 1062 provide an audio interface between the user and the mobile phone. The audio circuit 1060 converts the received audio data into electrical signals and transmits them to the speaker 1061, where the speaker 1061 converts them into sound signals for output. On the other hand, the microphone 1062 converts the collected sound signals into electrical signals, which are then received by the audio circuit 1060, converted into audio data, and then processed by the processor 1080 before being transmitted via the RF circuit 1010 to, for example, another mobile phone, or the audio data can be output to the memory 1020 for further processing.

[0187] WiFi is a short-range wireless transmission technology. Through the WiFi module 1070, mobile phones can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 14 The WiFi module 1070 is shown, but it is understood that it is not an essential component of a mobile phone and can be omitted as needed without changing the essence of the invention.

[0188] The processor 1080 is the control center of the mobile phone, connecting various parts of the phone through various interfaces and lines. It executes software programs and / or modules stored in the memory 1020 and calls data stored in the memory 1020 to perform various functions and process data, thereby collecting overall data and information from the phone. Optionally, the processor 1080 may include one or more processing units; preferably, the processor 1080 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 1080.

[0189] The mobile phone also includes a power supply 1090 (such as a battery) that supplies power to various components. Preferably, the power supply can be logically connected to the processor 1080 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.

[0190] Although not shown, mobile phones may also include a camera, Bluetooth module, etc., which will not be described in detail here.

[0191] In this embodiment of the application, the processor 1080 included in the mobile phone also has the following functions:

[0192] Obtain the source sample name and source evaluation result label corresponding to the source sample, and obtain the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample;

[0193] The source sample name, the sample text titles corresponding to the multiple sample texts, and the sample text keywords corresponding to the multiple sample texts are concatenated to obtain a text sample sequence.

[0194] The frequency of publication of the sample texts corresponding to the multiple sample texts is concatenated to obtain the frequency sample sequence.

[0195] The text sample sequence and the frequency sample sequence are input into the evaluation model to be trained, so that the text sample sequence and the frequency sample sequence are encoded by the evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence.

[0196] The text sample vector and the frequency sample vector are predicted and evaluated by the evaluation model to be trained to obtain the source evaluation prediction result.

[0197] Based on the difference between the source evaluation result label and the source evaluation prediction result, the parameters of the evaluation model to be trained are adjusted until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

[0198] or;

[0199] Obtain the target information source name corresponding to the target information source, and obtain the information source text title, information source text keywords and information source text publication frequency corresponding to the multiple information source texts published by the target information source;

[0200] The target source name, the source text titles corresponding to the multiple source texts, and the source text keywords corresponding to the multiple source texts are concatenated to obtain a source text sequence.

[0201] The frequency of source text publication corresponding to the multiple source texts is concatenated to obtain a source frequency sequence.

[0202] The source text sequence and the source frequency sequence are input into the source evaluation model to encode the source text sequence and the source frequency sequence through the source evaluation model, thereby obtaining the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence. The source evaluation model is a model trained according to the above-described source evaluation model training method.

[0203] The source text vector and the source frequency vector are evaluated by the source evaluation model to obtain a target evaluation result, wherein the target evaluation result is used to indicate whether the target source is valid or invalid.

[0204] This application also provides a computer-readable storage medium for storing a computer program that, when run on a computer device, causes the computer device to execute any one of the training methods for a source evaluation model described in the foregoing embodiments, or to execute any one of the source evaluation methods described in the foregoing embodiments.

[0205] This application also provides a computer program product including a computer program, which, when run on a computer device, causes the computer device to execute any one of the implementation methods of the training method for a source evaluation model described in the foregoing embodiments, or to execute any one of the implementation methods of a source evaluation method described in the foregoing embodiments.

[0206] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and equipment described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0207] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of the system is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple systems 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 an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0208] The system described as separate components may or may not be physically separate. 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, depending on actual needs.

[0209] 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.

[0210] If the integrated unit 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. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing computer programs.

[0211] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0212] 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.

Claims

1. A training method for a source evaluation model, characterized in that, include: Obtain the source sample name and source evaluation result label corresponding to the source sample, and obtain the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample; The source sample name, the sample text titles corresponding to the multiple sample texts, and the sample text keywords corresponding to the multiple sample texts are concatenated to obtain a text sample sequence. The frequency of publication of the sample texts corresponding to the multiple sample texts is concatenated to obtain the frequency sample sequence. The text sample sequence and the frequency sample sequence are input into the evaluation model to be trained, so that the text sample sequence and the frequency sample sequence are encoded by the evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence. The text sample vector and the frequency sample vector are predicted and evaluated by the evaluation model to be trained to obtain the source evaluation prediction result. Based on the difference between the source evaluation result label and the source evaluation prediction result, the parameters of the evaluation model to be trained are adjusted until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

2. The method according to claim 1, characterized in that, The sample text titles corresponding to the plurality of sample texts include the first sample text title corresponding to the first sample text and the second sample text title corresponding to the second sample text; the sample text keywords corresponding to the plurality of sample texts include the first sample text keyword corresponding to the first sample text and the second sample text keyword corresponding to the second sample text. The process of concatenating the source sample name, the sample text titles corresponding to the multiple sample texts, and the sample text keywords corresponding to the multiple sample texts to obtain a text sample sequence includes: The source sample name, the first sample text title, and the first sample text keywords are horizontally concatenated to obtain the first sample sequence; The first sample sequence, the second sample text title, and the second sample text keywords are horizontally concatenated to obtain a text sample sequence.

3. The method according to claim 2, characterized in that, Before concatenating the publication frequencies of the sample texts corresponding to the plurality of sample texts to obtain the frequency sample sequence, the method further includes: The order in which the title and keywords of the first sample text in the text sample sequence are concatenated is recorded as the first sample order, and the order in which the title and keywords of the second sample text in the text sample sequence are concatenated is recorded as the second sample order; The publication frequency of the sample texts corresponding to the plurality of sample texts includes the publication frequency of the first sample text corresponding to the first sample text and the publication frequency of the second sample text corresponding to the second sample text. The step of concatenating the publication frequencies of the sample texts corresponding to the plurality of sample texts to obtain the frequency sample sequence includes: The publication frequency of the first sample text and the publication frequency of the second sample text are encoded to obtain a first initial sample vector corresponding to the publication frequency of the first sample text and a second initial sample vector corresponding to the publication frequency of the second sample text. Based on the first sample order and the second sample order, the first initial sample vector and the second initial sample vector are sorted to obtain a frequency sample sequence.

4. The method according to claim 3, characterized in that, The training evaluation model includes a vector transformation model and a frequency embedding layer. The step of inputting the text sample sequence and the frequency sample sequence into the training evaluation model to encode the text sample sequence and the frequency sample sequence through the training evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence includes: The text sample sequence is input into the vector transformation model of the model to be trained and evaluated, so as to transform the text sample sequence through the vector transformation model of the model to be trained and evaluated, and obtain the text sample vector corresponding to the text sample sequence. The frequency sample sequence is input into the frequency embedding layer of the model to be trained and evaluated, so that the frequency sample sequence is mapped through the frequency embedding layer of the model to be trained and evaluated to obtain the frequency sample vector corresponding to the frequency sample sequence.

5. The method according to claim 4, characterized in that, The step of transforming the text sample sequence using the vector transformation model of the model to be trained and evaluated to obtain the text sample vector corresponding to the text sample sequence includes: Using the vector transformation model of the model to be trained and evaluated, vector transformation is performed on the source sample name, first sample text title, first sample text keyword, second sample text title and second sample text keyword in the text sample sequence to obtain the sample name vector corresponding to the source sample name, the first title vector corresponding to the first sample text title, the first keyword vector corresponding to the first sample text keyword, the second title vector corresponding to the second sample text title and the second keyword vector corresponding to the second sample text keyword. The sample name vector corresponding to the source sample name, the first title vector corresponding to the first sample text title, the first keyword vector corresponding to the first sample text keyword, the second title vector corresponding to the second sample text title, and the second keyword vector corresponding to the second sample text keyword are used as the text sample vector corresponding to the text sample sequence.

6. The method according to claim 5, characterized in that, The step of mapping the frequency sample sequence through the frequency embedding layer of the model to be trained and evaluated to obtain the frequency sample vector corresponding to the frequency sample sequence includes: Through the frequency embedding layer of the model to be trained and evaluated, the first initial sample vector and the second initial sample vector in the frequency sample sequence are vector mapped to obtain the first frequency vector corresponding to the first initial sample vector and the second frequency vector corresponding to the second initial sample vector. The first frequency vector corresponding to the first initial sample vector and the second frequency vector corresponding to the second initial sample vector are used as the frequency sample vectors corresponding to the frequency sample sequence.

7. The method according to claim 6, characterized in that, The process of predicting and evaluating the text sample vector and the frequency sample vector using the evaluation model to be trained, to obtain the source evaluation prediction result, includes: The evaluation model to be trained performs dimensionality transformation on the first title vector, the first keyword vector, and the first frequency vector in the frequency sample vector of the text sample vector to obtain the first sample vector; and performs dimensionality transformation on the second title vector, the second keyword vector, and the second frequency vector in the frequency sample vector of the text sample vector to obtain the second sample vector. The evaluation model to be trained performs prediction and evaluation processing on the sample name vector, the first sample vector, and the second sample vector in the text sample vector to obtain the source evaluation prediction result.

8. The method according to claim 7, characterized in that, The evaluation model to be trained further includes a max pooling layer and an evaluation classification layer. The process of predicting and evaluating the sample name vector, the first sample vector, and the second sample vector in the text sample vector by the evaluation model to obtain the source evaluation prediction result includes: The first sample vector and the second sample vector are aggregated by the max pooling layer in the model to be trained and evaluated to obtain an aggregated sample vector. The evaluation classification layer in the evaluation model to be trained predicts and evaluates the sample name vector and the aggregated sample vector in the text sample vector to obtain the source evaluation prediction result.

9. The method according to claim 7, characterized in that, The evaluation model to be trained further includes a dimension transformation layer, wherein the evaluation model to be trained performs dimension transformation processing on the first title vector, the first keyword vector, and the first frequency vector in the frequency sample vector of the text sample vector to obtain a first sample vector, and performs dimension transformation processing on the second title vector, the second keyword vector, and the second frequency vector in the frequency sample vector of the text sample vector to obtain a second sample vector, including: The training evaluation model concatenates the first title vector, the first keyword vector, and the first frequency vector in the frequency sample vector of the text sample vector to obtain a first concatenated vector; and concatenates the second title vector, the second keyword vector, and the second frequency vector in the frequency sample vector of the text sample vector to obtain a second concatenated vector. The dimension transformation layer in the model to be trained and evaluated performs dimension transformation on the first concatenated vector to obtain a first sample vector, and performs dimension transformation on the second concatenated vector to obtain a second sample vector.

10. A source evaluation method, characterized in that, include: Obtain the target information source name corresponding to the target information source, and obtain the information source text title, information source text keywords and information source text publication frequency corresponding to the multiple information source texts published by the target information source; The target source name, the source text titles corresponding to the multiple source texts, and the source text keywords corresponding to the multiple source texts are concatenated to obtain a source text sequence. The frequency of source text publication corresponding to the multiple source texts is concatenated to obtain a source frequency sequence. The source text sequence and the source frequency sequence are input into the source evaluation model to encode the source text sequence and the source frequency sequence through the source evaluation model, thereby obtaining the source text vector corresponding to the source text sequence and the source frequency vector corresponding to the source frequency sequence, wherein the source evaluation model is a model trained by the training method of the source evaluation model according to any one of claims 1-9; The source text vector and the source frequency vector are evaluated by the source evaluation model to obtain a target evaluation result, wherein the target evaluation result is used to indicate whether the target source is valid or invalid.

11. The method according to claim 10, characterized in that, Also includes: If the target evaluation result indicates that the target information source is valid, the multiple source texts published by the target information source are filtered to obtain source map texts containing map elements in the multiple source texts; The target map is updated based on the source map text to obtain an updated target map, wherein the updated target map is used to provide effective navigation information.

12. A training device for a source evaluation model, characterized in that, include: The source sample text acquisition unit is used to acquire the source sample name and source evaluation result label corresponding to the source sample, as well as the sample text title, sample text keywords and sample text publication frequency corresponding to the multiple sample texts published by the source sample; The text sample sequence acquisition unit is used to concatenate the source sample name, the sample text titles corresponding to the plurality of sample texts, and the sample text keywords corresponding to the plurality of sample texts to obtain a text sample sequence. The frequency sample sequence acquisition unit is used to concatenate the publication frequencies of the sample texts corresponding to the plurality of sample texts to obtain the frequency sample sequence. The sample sequence encoding processing unit is used to input the text sample sequence and the frequency sample sequence into the training and evaluation model, so as to encode the text sample sequence and the frequency sample sequence through the training and evaluation model to obtain the text sample vector corresponding to the text sample sequence and the frequency sample vector corresponding to the frequency sample sequence. The evaluation prediction result acquisition unit is used to perform prediction evaluation processing on the text sample vector and the frequency sample vector by the evaluation model to be trained, and obtain the source evaluation prediction result. The source evaluation model acquisition unit is used to adjust the parameters of the evaluation model to be trained based on the difference between the source evaluation result label and the source evaluation prediction result, until the adjusted model meets the model training cutoff condition, and the training ends to obtain the source evaluation model.

13. A computer device, characterized in that, The device includes a processor and a memory: The memory is used to store computer programs and to transfer the computer programs to the processor; The processor is configured to execute the steps of the training method of the source evaluation model according to any one of claims 1 to 9, or to execute the steps of the source evaluation method according to any one of claims 10 to 11, according to the instructions in the computer program.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, which, when executed by a computer device, implements the steps of the training method for the source evaluation model according to any one of claims 1 to 9, or the steps of the source evaluation method according to any one of claims 10 to 11.

15. A computer program product, characterized in that, It includes a computer program that, when executed by a computer device, implements the steps of the training method for the source evaluation model according to any one of claims 1 to 9, or the steps of the source evaluation method according to any one of claims 10 to 11.

Images