Content quality assessment methods, devices, media and electronic equipment

Abstract

This disclosure relates to a content quality evaluation method, apparatus, medium, and electronic device. The method includes: acquiring a first evaluation result and evaluation text from a target user regarding target content; determining a second evaluation result for the target content based on the evaluation text; determining a matching degree between the target user and the target content; and determining a quality evaluation result from the target user regarding the target content based on the first evaluation result, the second evaluation result, and the matching degree. Therefore, combining these three factors to determine the quality evaluation result fully considers the user-input evaluation score and comment text, ensuring consistency between the determined quality evaluation result and the user-input evaluation text. Furthermore, considering the matching degree between the user and the content can reduce the influence of user subjectivity to a certain extent, ensuring the accuracy and objectivity of the quality evaluation result and improving the matching degree between the quality evaluation result and the target content.

Description

Technical Field

[0001] This disclosure relates to the field of data processing, and more specifically, to a content quality evaluation method, apparatus, medium, and electronic device. Background Technology

[0002] Text sentiment analysis refers to the process of analyzing, processing, and extracting sentiment from subjective texts that carry emotional connotations, using methods such as natural language processing, text mining, and computational linguistics.

[0003] For content quality evaluation systems such as those for movies and TV series, viewers provide their own evaluation text and score, thus assessing the content's quality and recommending it to other users. However, in these evaluation methods, the scoring system typically uses a five-point scale. Users can easily give a perfect score by directly entering a rating, but the corresponding evaluation text may not be entirely positive. In other words, the user-entered rating and the evaluation text often do not match, resulting in inflated scores for many contents and making it difficult to provide users with accurate content viewing references. Summary of the Invention

[0004] The purpose of this disclosure is to provide an accurate and objective method, apparatus, medium, and electronic device for evaluating content quality.

[0005] To achieve the above objectives, according to a first aspect of this disclosure, a content quality evaluation method is provided, the method comprising:

[0006] Obtain the target user's initial evaluation results and evaluation text for the target content;

[0007] Based on the evaluation text, determine the second evaluation result of the target content;

[0008] Determine the matching degree between the target user and the target content;

[0009] Based on the first evaluation result, the second evaluation result, and the matching degree, the quality evaluation result of the target user for the target content is determined.

[0010] Optionally, determining the matching degree between the target user and the target content includes:

[0011] Based on the target user's historical evaluation text, determine the user vector corresponding to the target user;

[0012] Based on multiple evaluation texts corresponding to the target content, determine the content vector corresponding to the target content;

[0013] The matching degree is determined based on the user vector and the content vector.

[0014] Optionally, determining the user vector corresponding to the target user based on the target user's historical evaluation text includes:

[0015] Cluster the historical evaluation texts to obtain multiple clusters corresponding to the historical evaluation texts;

[0016] For each cluster, historical evaluation texts in the cluster whose text length is less than a preset length threshold are concatenated to obtain at least one concatenated text.

[0017] Based on the concatenated text in each cluster, historical evaluation text with a text length not less than the specified length threshold, and the topic generation model, the topic words corresponding to the target user are determined, and the user vector is determined based on the vector corresponding to the topic words.

[0018] Optionally, determining the content vector corresponding to the target content based on multiple evaluation texts corresponding to the target content includes:

[0019] Determine the word frequency and inverse document frequency of each word in the multiple evaluation texts corresponding to the target content, as well as the text length ratio of the word segment, wherein the text length ratio of the word segment is the ratio of the length of the evaluation text to which the word segment belongs to the average text length of the multiple evaluation texts;

[0020] For each word segment, the product of the word frequency, the inverse document frequency, and the proportion of the text length corresponding to the word segment is determined as the target parameter corresponding to the word segment.

[0021] Based on the target parameters corresponding to each segmented word, the target segmented word corresponding to the target content is determined, and the content vector is determined based on the vector corresponding to the target segmented word.

[0022] Optionally, determining the second evaluation result of the target content based on the evaluation text includes:

[0023] Based on the evaluation text and the text classification model, the classification corresponding to the evaluation text is determined, and the score indicated by the classification is determined as the second evaluation result;

[0024] In the training process of the text classification model, feature extraction is performed based on the feature extraction sub-model and target features are obtained based on the fully connected layer. The prediction result of the text classification model is obtained by predicting some features in the target features.

[0025] Optionally, determining the quality evaluation result of the target user for the target content based on the first evaluation result, the second evaluation result, and the matching degree includes:

[0026] The weighted sum of the first evaluation result and the second evaluation result is determined as the initial evaluation result;

[0027] The initial evaluation result is adjusted based on the matching degree to obtain the quality evaluation result.

[0028] Optionally, the method further includes:

[0029] Based on the quality evaluation results corresponding to each piece of content in the content library, the recommended content corresponding to the target user is determined;

[0030] Output the recommended content.

[0031] According to a second aspect of this disclosure, a content quality evaluation apparatus is provided, the apparatus comprising:

[0032] The acquisition module is used to acquire the target user's first evaluation result and evaluation text for the target content;

[0033] The first determining module is used to determine the second evaluation result of the target content based on the evaluation text;

[0034] The second determining module is used to determine the matching degree between the target user and the target content;

[0035] The third determining module is used to determine the quality evaluation result of the target user for the target content based on the first evaluation result, the second evaluation result, and the matching degree.

[0036] According to a third aspect of this disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of any of the methods described in the first aspect.

[0037] According to a fourth aspect of this disclosure, an electronic device is provided, comprising:

[0038] A memory on which computer programs are stored;

[0039] A processor for executing the computer program in the memory to implement the steps of any of the methods described in the first aspect.

[0040] Therefore, through the above technical solution, the first evaluation result is the rating input by the target user, and the second evaluation result is the rating determined based on the evaluation text input by the target user. The matching degree between the target user and the target content can, to a certain extent, characterize the possibility of subjective evaluation by the user. Combining these three factors to determine the quality evaluation result fully considers the user's input rating score and comment text, ensuring consistency between the determined quality evaluation result and the user's input evaluation text. Simultaneously, considering the matching degree between the user and the content can, to a certain extent, reduce the influence of user subjectivity, thereby ensuring the accuracy and objectivity of the quality evaluation result, improving the matching degree between the quality evaluation result and the target content, providing users with accurate content viewing references, and enhancing the user experience.

[0041] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0042] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0043] Figure 1 This is a flowchart of a content quality evaluation method provided according to one embodiment of the present disclosure;

[0044] Figure 2 This is a schematic diagram of the structure of a text classification model provided according to one embodiment of the present disclosure;

[0045] Figure 3 This is a flowchart illustrating an exemplary implementation of determining the matching degree between target users and target content;

[0046] Figure 4 This is a block diagram of a content quality evaluation device provided according to one embodiment of the present disclosure;

[0047] Figure 5 This is a block diagram illustrating an electronic device according to an exemplary embodiment;

[0048] Figure 6 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0049] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0050] Figure 1This is a flowchart illustrating a content quality evaluation method according to one embodiment of the present disclosure. Figure 1 As shown, the method may include:

[0051] In step 11, the first evaluation result and evaluation text of the target user regarding the target content are obtained. The target content can be any content from multimedia content such as movies, TV series, and animations. For example, the first evaluation result can be a rating value entered by the target user for the target content, and the evaluation text can be the comment text entered by the target user for the target content.

[0052] In step 12, a second evaluation result for the target content is determined based on the evaluation text.

[0053] The evaluation text is the target user's comment on the target content. Therefore, the evaluation text can reflect the target user's evaluation bias towards the target content to a certain extent. Thus, in this embodiment, another evaluation result of the target user on the target content can be further determined based on the evaluation text. For example, the second evaluation result can be represented by a rating value, and the range of the rating value is the same as the range of the rating value corresponding to the first evaluation result. In this case, the second evaluation result can be used to represent the rating determined based on the evaluation text input by the target user.

[0054] In step 13, the matching degree between the target user and the target content is determined.

[0055] Each target user may have a corresponding area of ​​interest. The target content within that area might elicit a more favorable evaluation from that user, indicating a potential bias. Therefore, in this embodiment, the matching degree between the target user and the target content can be determined to characterize the likelihood of subjective evaluation by the user; a higher matching degree indicates a higher probability that the user's evaluation is subjective.

[0056] In step 14, the quality evaluation result of the target user for the target content is determined based on the first evaluation result, the second evaluation result, and the matching degree.

[0057] Therefore, through the above technical solution, the first evaluation result is the rating input by the target user, and the second evaluation result is the rating determined based on the evaluation text input by the target user. The matching degree between the target user and the target content can, to a certain extent, characterize the possibility of subjective evaluation by the user. Combining these three factors to determine the quality evaluation result fully considers the user's input rating score and comment text, ensuring consistency between the determined quality evaluation result and the user's input evaluation text. Simultaneously, considering the matching degree between the user and the content can, to a certain extent, reduce the influence of user subjectivity, thereby ensuring the accuracy and objectivity of the quality evaluation result, improving the matching degree between the quality evaluation result and the target content, providing users with accurate content viewing references, and enhancing the user experience.

[0058] As an example, a rating prediction can be made on the evaluation text based on a rating model to obtain a second evaluation result. For example, the rating model can be implemented based on CNN (Convolutional Neural Networks), RNN (Recurrent Neural Networks), etc., and trained using the rated evaluation text as training samples. The training method can adopt the common methods in this field, which will not be elaborated here.

[0059] As another example, an exemplary implementation of determining the second evaluation result of the target content based on the evaluation text in step 12 is as follows:

[0060] Based on the evaluation text and the text classification model, the classification corresponding to the evaluation text is determined, and the score indicated by the classification is determined as the second evaluation result.

[0061] In the training process of the text classification model, feature extraction is performed based on the feature extraction sub-model and target features are obtained based on the fully connected layer. The prediction result of the text classification model is obtained by predicting some features in the target features.

[0062] For example, this feature extraction sub-model can be implemented based on the Transformer model. Correspondingly, the structural diagram of this text classification model is shown below. Figure 2 As shown, the number of layers in each feature layer of the text classification model is merely illustrative and does not limit the scope of this disclosure.

[0063] like Figure 2As shown, the text classification model includes a feature extraction sub-model, which can be composed of a Transformer model, followed by a fully-connected layer and a dropout layer. It can also combine an activation function and a normalization layer to derive the prediction result. For example, the activation function can be GELU. For example, if the score is set to 1-5 points, then 5 categories can be set in the normalization layer, corresponding to scores 1, 2, 3, 4, and 5 respectively.

[0064] Therefore, in this embodiment, the evaluation text can be segmented to obtain the corresponding segmented word sequence. Each segmented word is then vectorized and input into the text classification model. During feature extraction in the text classification model, an attention mechanism can be used to focus more on features related to the score. After feature extraction, a fully connected layer is used to obtain the target features, ensuring the accuracy of feature extraction and providing reliable data support for subsequent predictions. Furthermore, the prediction result of the text classification model is obtained by predicting a subset of the target features. For example, a Dropout Layer can be used to set the probability of dropout for each neural network layer. For neural network training units, they are removed from the network according to this probability to obtain prediction results and for training. Thus, during the training of the text classification model, predictions can be made randomly based on a subset of the target features. For stochastic gradient descent, the selection of subset features is random, ensuring that each mini-batch of the text classification model trains a different network. This effectively avoids overfitting during training, thereby improving the accuracy of the trained text classification model and ensuring the accuracy of the determined second evaluation result, providing support for objective and accurate evaluation of the target content.

[0065] In one possible embodiment, an exemplary implementation of determining the matching degree between the target user and the target content in step 13 is as follows: Figure 3 As shown, this step may include:

[0066] In step 31, the user vector corresponding to the target user is determined based on the target user's historical evaluation text.

[0067] In this step, with the user's authorization, the various evaluation texts of the target user can be obtained, thereby extracting the user's characteristics from the user's historical evaluations, i.e., obtaining the user vector.

[0068] As an example, keywords can be extracted based on multiple historical review texts of the target user, and the extracted keywords can be used as feature words of the target user. Then, the user vector can be obtained by vectorizing the feature words.

[0069] As another example, determining the user vector corresponding to the target user based on the target user's historical evaluation text may include:

[0070] The historical evaluation texts are clustered to obtain multiple clusters corresponding to the historical evaluation texts.

[0071] Among these methods, historical evaluation texts can be clustered using common clustering algorithms such as K-means or KNN to obtain multiple clusters.

[0072] For each cluster, historical evaluation texts in that cluster whose text length is less than a preset length threshold are concatenated to obtain at least one concatenated text.

[0073] In this context, users often rate content using short texts, which can hinder keyword generation. Therefore, in this embodiment, historical rating texts can be clustered together, grouping those with similar characteristics together.

[0074] For example, the preset length threshold can be set based on the actual application scenario, and this disclosure does not limit it. For each cluster, if the text length of the historical evaluation text is less than the length threshold, it means that the historical evaluation text is a short text. In this case, short texts in the same cluster can be concatenated, so that the length of the historical evaluation text can be extended based on texts with similar characteristics, thereby achieving text extension while preserving the characteristics of each historical evaluation text.

[0075] As an example, all historical evaluation texts in a cluster whose length is less than a preset length threshold can be concatenated to obtain a concatenated text. As another example, to avoid excessively long concatenated texts, a text length limit can be set for the concatenated text. That is, when concatenating short texts in the current cluster, if the length of the currently concatenated text S is less than this limit, and if the length of the concatenated text S' obtained after adding a new historical evaluation text A exceeds this limit, then no concatenation is performed. Instead, the currently concatenated text S is used as the completed concatenated text, and the newly added historical evaluation text A is used as a new concatenated text for further concatenation with other historical evaluation texts to obtain multiple concatenated texts.

[0076] Subsequently, based on the concatenated text in each cluster, historical evaluation text with a text length not less than the specified length threshold, and the topic generation model, the topic words corresponding to the target user are determined, and the user vector is determined based on the vector corresponding to the topic words.

[0077] In this embodiment, the text length of historical evaluation text is not less than the length threshold, indicating that the historical evaluation text can be directly applied to topic word generation. As mentioned above, historical evaluation texts with a text length less than the length threshold in each cluster are concatenated to obtain concatenated text, which results in a longer text length. Therefore, in this embodiment, topic words can be obtained based on the concatenated text and the unconcatenated historical evaluation texts in the clusters using a topic generation model. For example, a preset number of words can be selected as topic words according to the probability distribution output by the topic generation model in descending order of probability. Then, the topic words are vectorized to obtain a user vector representing the user's characteristics.

[0078] For example, the topic generation model can be implemented based on LDA (Latent Dirichlet Allocation), and its training method is an existing technology, which will not be elaborated here. The LDA model is an unsupervised model. In this embodiment, the LDA model can be pre-trained based on existing prior text, such as user reviews of horror movies or comedy movies. This can improve the topic generation model's ability to determine keywords for customized scenarios relevant to the user's interests.

[0079] Therefore, short texts with similar characteristics can be merged and spliced ​​together, allowing for the identification of keywords based on longer texts. This improves the accuracy of keyword identification to some extent, thereby enhancing the accuracy of identified user characteristics and providing accurate data support for user profiling.

[0080] In step 32, the content vector corresponding to the target content is determined based on multiple evaluation texts corresponding to the target content.

[0081] Multiple users can rate the same content. Therefore, based on the rating texts of multiple users on the same target content, the true comprehensive features of the target content can be obtained, that is, the content vector.

[0082] As an example, keywords can be extracted from multiple evaluation texts, and the extracted keywords can be used as feature words of the target content. Then, the content vector can be obtained by vectorizing the feature words.

[0083] As another example, determining the content vector corresponding to the target content based on multiple evaluation texts corresponding to the target content may include:

[0084] Determine the word frequency and inverse document frequency of each word in the multiple evaluation texts corresponding to the target content, as well as the text length ratio corresponding to the word, wherein the text length ratio corresponding to the word is the ratio of the length of the evaluation text to which the word belongs to the average text length of the multiple evaluation texts.

[0085] Each evaluation text can be treated as a document, and the evaluation text can be segmented into words to obtain the individual words corresponding to multiple evaluation texts. From this, the term frequency (TF), inverse document frequency (IDF), and text length ratio (DOC_LEN) corresponding to each word segment can be further determined, as shown in the following formula:

[0086]

[0087]

[0088]

[0089] For each segmented word, the product of the word frequency, the inverse document frequency, and the proportion of the text length corresponding to the segmented word is determined as the target parameter corresponding to the segmented word; based on the target parameter corresponding to each segmented word, the target segmented word corresponding to the target content is determined, and the content vector is determined based on the vector corresponding to the target segmented word.

[0090] Specifically, the top N word segments ranked from largest to smallest according to their corresponding target parameters can be identified as the target word segments corresponding to the target content. These target word segments can then be vectorized to obtain a content vector. The vectorization of the target user's topic words and the vectorization of the target word segments for the target content are performed in the same way, and commonly used vectorization methods in this field, such as word2vec, can be selected. This disclosure does not limit the specific vectorization method used.

[0091] Therefore, when determining target words from evaluation text, in addition to considering the word frequency and inverse document frequency of the segmented words, the text length of the evaluation text to which the segmented words belong is also taken into account. As shown above, it is difficult to accurately extract keywords from evaluation texts with shorter text lengths. Therefore, this disclosure considers the text length ratio to increase the importance of word segmentation in evaluation texts with longer text lengths, ensuring the accuracy of the determined target words, thereby improving the accuracy and comprehensiveness of features in the content vector.

[0092] In step 33, the matching degree is determined based on the user vector and the content vector.

[0093] For example, the cosine similarity between the user vector and the content vector can be calculated as the matching degree, for example, using the following formula:

[0094]

[0095] Where β1 represents the user vector, β2 represents the content vector, Ai represents the i-th feature in the user vector, Bi represents the i-th feature in the content vector, and n represents the dimension of the features in the user vector and the content vector, with the same number of dimensions for both.

[0096] Therefore, through the above technical solution, the interest characteristics of the target user can be obtained based on the user's historical evaluation text, and the characteristics of the target content itself can be obtained based on multiple evaluation texts of the target content. Thus, the matching degree between the two can characterize whether the user can give an objective evaluation of the target content, providing data parameters for determining the final quality evaluation result based on the matching degree, and ensuring the objectivity of the determined quality evaluation result.

[0097] In one possible embodiment, an exemplary implementation of determining the target user's quality evaluation result for the target content based on the first evaluation result, the second evaluation result, and the matching degree in step 14 is as follows: This step may include:

[0098] The weighted sum of the first evaluation result and the second evaluation result is determined as the initial evaluation result.

[0099] For example, the weights corresponding to the first evaluation result and the second evaluation result can be preset, wherein the sum of the weights corresponding to the first evaluation result and the second evaluation result is 1. The weights corresponding to the two can be set according to the actual application scenario, and this disclosure does not limit them. As an example, the weights corresponding to the two are 0.5, and the average value of the first evaluation result and the second evaluation result can be determined as the initial evaluation result, thereby comprehensively considering the consistency between the user input rating and the text.

[0100] The initial evaluation result is adjusted based on the matching degree to obtain the quality evaluation result.

[0101] As mentioned above, the matching degree represents the subjectivity of a target user's evaluation of the target content. A higher matching degree indicates stronger subjectivity, meaning the target user is more likely to give a biased evaluation. Therefore, when determining the quality evaluation result requires offsetting the influence of subjective preferences, the initial evaluation result is adjusted based on the matching degree. This can be achieved by subtracting the matching degree from the initial evaluation result to obtain the quality evaluation result. Alternatively, a pre-set adjustment weight can be used, where the product of the matching degree and its adjustment weight is used as the adjustment value, and the initial evaluation result is subtracted from this adjustment value to obtain the quality evaluation result.

[0102] Therefore, the above technical solution ensures that the determined quality evaluation results are consistent with the user-input rating scores and evaluation text, while also reducing the influence of user subjective preferences to a certain extent, thus guaranteeing the accuracy and objectivity of the determined quality evaluation results. Furthermore, it makes the quality evaluation results more dispersed compared to the user's initial evaluation, avoiding situations where there is a high degree of overlap in content evaluation results, making it difficult to distinguish content quality, and providing users with more accurate data references to clarify the actual quality of each piece of content.

[0103] In one possible embodiment, the method may further include:

[0104] Based on the quality evaluation result corresponding to each piece of content in the content library, recommended content corresponding to the target user is determined; the recommended content is then output. The quality evaluation results corresponding to all or part of the content in the content library are determined according to the content quality evaluation method described above.

[0105] As an example, the top P items in the content library can be selected as recommended content for the target user, ranked from highest to lowest quality rating. Here, P can be set according to the actual user needs. This allows for the recommendation of high-quality content to the target user, ensuring a positive viewing experience.

[0106] As another example, the top Q content items in the content library can be selected as candidate content for the target user, ranked from highest to lowest quality rating. Here, Q can be set according to the actual user needs. Then, based on the selection matching degree between the candidate content and the target user, the top P content items are selected as recommended content in descending order of selection matching degree and displayed accordingly, where P is less than or equal to Q. This selection matching degree can be determined by calculating the similarity between the candidate content's tag vector and the target user's interest vector, indicating whether the candidate content matches the user's interests. For example, the tag vector can be derived from the vectorization of the candidate content's type tag, which could be a genre (e.g., feature film, documentary), type (e.g., comedy, tragedy, urban, rural), etc. The interest vector can be obtained by vectorizing the user's interest tags with user authorization, such as idol drama, youth, urban, etc. Therefore, while recommending high-quality content to target users, the system can also display recommendations based on the target users' interests and preferences, ensuring a good match between the recommended content and the target users, while increasing the diversity and personalization of the recommended content, and further enhancing the user experience.

[0107] Based on the same inventive concept, this disclosure also provides a content quality evaluation device, such as... Figure 4 As shown, the device 10 includes:

[0108] The acquisition module 100 is used to acquire the first evaluation result and evaluation text of the target user for the target content;

[0109] The first determining module 200 is used to determine a second evaluation result of the target content based on the evaluation text;

[0110] The second determining module 300 is used to determine the matching degree between the target user and the target content;

[0111] The third determining module 400 is used to determine the quality evaluation result of the target user for the target content based on the first evaluation result, the second evaluation result and the matching degree.

[0112] Optionally, the second determining module includes:

[0113] The first determining submodule is used to determine the user vector corresponding to the target user based on the target user's historical evaluation text;

[0114] The second determining submodule is used to determine the content vector corresponding to the target content based on multiple evaluation texts corresponding to the target content;

[0115] The third determining submodule is used to determine the matching degree based on the user vector and the content vector.

[0116] Optionally, the first determining submodule includes:

[0117] The clustering submodule is used to cluster the historical evaluation texts to obtain multiple clusters corresponding to the historical evaluation texts.

[0118] The splicing submodule is used to splice historical evaluation texts in each cluster whose text length is less than a preset length threshold to obtain at least one spliced ​​text.

[0119] The fourth determination submodule is used to determine the topic words corresponding to the target user based on the concatenated text in each cluster, the historical evaluation text with a text length not less than the length threshold, and the topic generation model, and to determine the user vector based on the vector corresponding to the topic words.

[0120] Optionally, the second determining submodule includes:

[0121] The fifth determination submodule is used to determine the word frequency and inverse document frequency of each word in the multiple evaluation texts corresponding to the target content, as well as the text length ratio of the word segment, wherein the text length ratio of the word segment is the ratio of the length of the evaluation text to which the word segment belongs to the average text length of the multiple evaluation texts;

[0122] The sixth determining submodule is used to determine the target parameter corresponding to each word segment by multiplying the word frequency, the inverse document frequency and the proportion of the text length corresponding to the word segment.

[0123] The seventh determination submodule is used to determine the target word corresponding to the target content based on the target parameters corresponding to each word segmentation, and to determine the content vector based on the vector corresponding to the target word segmentation.

[0124] Optionally, the first determining module includes:

[0125] The eighth determining submodule is used to determine the category corresponding to the evaluation text based on the evaluation text and the text classification model, and to determine the score indicated by the category as the second evaluation result;

[0126] In the training process of the text classification model, feature extraction is performed based on the feature extraction sub-model and target features are obtained based on the fully connected layer. The prediction result of the text classification model is obtained by predicting some features in the target features.

[0127] Optionally, the third determining module includes:

[0128] The ninth determining submodule is used to determine the initial evaluation result by weighting the first evaluation result and the second evaluation result;

[0129] The adjustment submodule is used to adjust the initial evaluation result according to the matching degree to obtain the quality evaluation result.

[0130] Optionally, the device further includes:

[0131] The fourth determination module is used to determine the recommended content corresponding to the target user based on the quality evaluation result of each piece of content in the content library;

[0132] The output module is used to output the recommended content.

[0133] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0134] Figure 5 This is a block diagram illustrating an electronic device 700 according to an exemplary embodiment. Figure 5 As shown, the electronic device 700 may include a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input / output (I / O) interface 704, and a communication component 705.

[0135] The processor 701 controls the overall operation of the electronic device 700 to complete all or part of the steps in the content quality evaluation method described above. The memory 702 stores various types of data to support the operation of the electronic device 700. This data may include, for example, instructions for any application or method operating on the electronic device 700, and application-related data such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 702 or transmitted via communication component 705. The audio component also includes at least one speaker for outputting audio signals. I / O interface 704 provides an interface between processor 701 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IoT, eMTC, or other 5G technologies, or combinations thereof, is not limited here. Therefore, the corresponding communication component 705 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0136] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the content quality evaluation method described above.

[0137] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the content quality evaluation method described above. For example, the computer-readable storage medium may be the memory 702 including the program instructions described above, which may be executed by the processor 701 of the electronic device 700 to complete the content quality evaluation method described above.

[0138] Figure 6 This is a block diagram illustrating an electronic device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. (Refer to...) Figure 6 The electronic device 1900 includes a processor 1922, which may be one or more, and a memory 1932 for storing computer programs executable by the processor 1922. The computer program stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor 1922 may be configured to execute the computer program to perform the aforementioned content quality assessment method.

[0139] Additionally, the electronic device 1900 may also include a power supply component 1926 and a communication component 1950. The power supply component 1926 can be configured to perform power management of the electronic device 1900, and the communication component 1950 can be configured to enable communication of the electronic device 1900, such as wired or wireless communication. Furthermore, the electronic device 1900 may also include an input / output (I / O) interface 1958. The electronic device 1900 can operate on an operating system, such as Windows Server, stored in memory 1932. TM Mac OS X TM Unix TM Linux TM etc.

[0140] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the content quality evaluation method described above. For example, the non-transitory computer-readable storage medium may be the memory 1932 including the program instructions, which may be executed by the processor 1922 of the electronic device 1900 to complete the content quality evaluation method described above.

[0141] In another exemplary embodiment, a computer program product is also provided, which includes a computer program executable by a programmable device, the computer program having a code portion for performing the above-described content quality evaluation method when executed by the programmable device.

[0142] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0143] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0144] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A content quality evaluation method characterized by comprising: The method comprises: obtaining a first evaluation result and an evaluation text of a target user for target content; determining a second evaluation result of the target content according to the evaluation text; determining a matching degree corresponding to the target user and the target content, the matching degree being used to represent a subjective degree of the target user's evaluation of the target content, and the higher the matching degree, the stronger the subjective degree; determining a quality evaluation result of the target user for the target content according to the first evaluation result, the second evaluation result and the matching degree; The determination of the matching degree corresponding to the target user and the target content comprises: determining a user vector corresponding to the target user based on historical evaluation texts of the target user; determining a content vector corresponding to the target content based on a plurality of evaluation texts corresponding to the target content; determining the matching degree according to the user vector and the content vector; The determination of the quality evaluation result of the target user for the target content according to the first evaluation result, the second evaluation result and the matching degree comprises: determining a weighted sum of the first evaluation result and the second evaluation result as an initial evaluation result; adjusting the initial evaluation result according to the matching degree to obtain the quality evaluation result.

2. The method of claim 1, wherein, The determination of the user vector corresponding to the target user based on the historical evaluation texts of the target user comprises: clustering the historical evaluation texts to obtain a plurality of clustering clusters corresponding to the historical evaluation texts; for each clustering cluster, splicing historical evaluation texts with a text length less than a preset length threshold in the clustering cluster to obtain at least one spliced text; determining a topic word corresponding to the target user based on the spliced texts in each clustering cluster, historical evaluation texts with a text length not less than the length threshold, and a topic generation model, and determining the user vector based on a vector corresponding to the topic word.

3. The method of claim 1, wherein, The determination of the content vector corresponding to the target content based on a plurality of evaluation texts corresponding to the target content comprises: determining a term frequency and an inverse document frequency corresponding to each token in the plurality of evaluation texts corresponding to the target content, and a text length ratio corresponding to the token, wherein the text length ratio corresponding to the token is a ratio of a length of an evaluation text to which the token belongs to an average text length of the plurality of evaluation texts; for each token, determining a product of the term frequency, the inverse document frequency corresponding to the token, and the text length ratio corresponding to the token as a target parameter corresponding to the token; determining a target token corresponding to the target content according to the target parameter corresponding to each token, and determining the content vector based on a vector corresponding to the target token.

4. The method of claim 1, wherein, The determination of the second evaluation result of the target content according to the evaluation text comprises: determining a classification corresponding to the evaluation text according to the evaluation text and a text classification model, and determining a score indicated by the classification as the second evaluation result; In the training process of the text classification model, feature extraction is performed based on a feature extraction sub-model, and target features are obtained based on a full connection layer. The prediction result of the text classification model is obtained by predicting part of the target features.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: determining recommended content corresponding to the target user according to a quality evaluation result corresponding to each content in the content library; outputting the recommended content.

6. A content quality evaluation device characterized by comprising: The device includes: an acquisition module configured to acquire a first evaluation result and an evaluation text of a target user for target content; a first determination module configured to determine a second evaluation result of the target content according to the evaluation text; a second determination module configured to determine a matching degree corresponding to the target user and the target content, the matching degree being used to represent a subjective degree of the target user's evaluation of the target content, and the higher the matching degree, the stronger the subjective degree; a third determination module configured to determine a quality evaluation result of the target user for the target content according to the first evaluation result, the second evaluation result, and the matching degree; the second determination module includes: a first determination sub-module configured to determine a user vector corresponding to the target user based on historical evaluation texts of the target user; a second determination sub-module configured to determine a content vector corresponding to the target content based on a plurality of evaluation texts corresponding to the target content; a third determination sub-module configured to determine the matching degree according to the user vector and the content vector; the third determination module includes: a ninth determination sub-module configured to determine a weighted sum of the first evaluation result and the second evaluation result as an initial evaluation result; an adjustment sub-module configured to adjust the initial evaluation result according to the matching degree to obtain the quality evaluation result.

7. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method of any one of claims 1-5.

8. An electronic device, comprising: includes: a memory having a computer program stored thereon; a processor configured to execute the computer program in the memory to implement the steps of the method of any one of claims 1-5.

Images