Multimedia recommendation method, recommendation device, vehicle system and storage medium

Abstract

The present application discloses a multimedia recommendation method, a recommendation device, a vehicle system and a storage medium. The multimedia recommendation method includes: obtaining a query word; based on spelling and similar words, correcting the query word and obtaining a corrected word; obtaining multiple multimedia contents related to the corrected word; calculating the matching scores of the multiple multimedia contents and the corrected word and the user behavior scores of the multiple multimedia contents according to a trained predetermined model, wherein the user behavior scores are determined according to the historical playback of the media content; and sorting the multiple multimedia contents according to the matching scores and the user behavior scores and outputting the recommendation results. The multimedia recommendation method of the embodiment of the present application corrects the query word input by the user to obtain multiple corrected multimedia contents, and then outputs the recommended content after sorting according to the matching scores and user behavior scores calculated according to the trained predetermined model. This can improve the output results obtained when the user inputs fuzzy query words and recommend content that is more suitable for the user.

Description

Technical Field

[0001] The present application relates to the technical field of multimedia content search, and in particular to a multimedia recommendation method, a recommendation device, a vehicle system, and a storage medium. Background Art

[0002] In-car media services include content from multiple media sources, including music, audiobooks, and news. Users can search for content through voice or the car console. In related technologies, in-car systems perform natural language understanding based on user-entered query terms, identify the song and artist names in the music query terms, and generate search results based on the specific artist and song names. However, when the song and artist names are not clearly defined, only fuzzy matching searches are performed, resulting in low output accuracy. Summary of the Invention

[0003] The present application provides a multimedia recommendation method, a recommendation device, a vehicle system, and a storage medium.

[0004] The multimedia recommendation method of the embodiment of the present application includes:

[0005] Get the query term;

[0006] Based on spelling and similar words, modify the query word and obtain a modified word;

[0007] acquiring a plurality of multimedia contents related to the modified word;

[0008] Calculating, based on the trained predetermined model, matching scores between the plurality of multimedia contents and the modified words and user behavior scores of the plurality of multimedia contents, wherein the user behavior scores are determined based on historical playback of the multimedia contents;

[0009] The plurality of multimedia contents are sorted according to the matching score and the user behavior score, and a recommendation result is output.

[0010] The multimedia recommendation method of the embodiment of the present application modifies the query words input by the user to obtain multiple modified multimedia contents, and then outputs the recommended contents after sorting according to the matching scores and user behavior scores calculated according to the trained predetermined model. This can improve the output results when the user inputs fuzzy query words and recommend content that is more suitable for the user.

[0011] In some embodiments, obtaining a query word includes:

[0012] Using the input text as the query term; and / or,

[0013] The input speech-parsed text is used as the query word.

[0014] In some embodiments, modifying the query term to obtain a modified term includes:

[0015] Based on the pronunciation of the spelling, query the word library to confirm the correctness of the query word;

[0016] If the query word is incorrect, correct the query word and obtain a corrected word;

[0017] Based on semantics, query a similar word library to identify similar words to the query word;

[0018] In the case where the query word has similar words, the similar words are used as the modified words.

[0019] In certain embodiments, the calculation of the matching score includes:

[0020] Calculating the data length in the multimedia content;

[0021] Determine the length of the text to be matched;

[0022] Calculating the weight of the target word in the predetermined model output sentence;

[0023] The matching score is calculated based on the data length, the text length and the weight of the target word.

[0024] In some embodiments, calculating the weight of the target word in the predetermined model output sentence includes:

[0025] Calculating the ratio of the number of times the target word appears in the output sentence to the number of times all words appear in the output sentence;

[0026] Calculate the frequency of occurrence of the target word in the document;

[0027] The product of the ratio and the frequency of occurrence is used as the weight of the target word.

[0028] In certain embodiments, the matching score is calculated using the following formula:

[0029]

[0030] Among them, total_w word The sum of the weights of the matched target words; mappingWord len The length of the data in the matched multimedia content; Text1 len The length of the query text; Text2 len The length of the text of the target word.

[0031] In some embodiments, the calculation of the user behavior score includes:

[0032] Calculating a normalized value of characteristic data in the multimedia content, the characteristic data including at least one of the number of plays, the number of people who play the content, the average play time, the number of plays by the same user, and the play time;

[0033] Calculating the weight of the feature data;

[0034] The product of the normalized value and the weight of the feature data is accumulated as the user behavior score.

[0035] In certain embodiments, the normalized value is calculated using the following formula:

[0036] X norm =(X-Xmin) / (Xmax-Xmin)

[0037] Wherein, X is the feature data, Xmin is the minimum value of the feature data, and Xmax is the maximum value of the feature data.

[0038] The recommended devices for the implementation of this application include:

[0039] A first acquisition module is used to acquire a query word;

[0040] A correction module, configured to correct the query word based on spelling and similar words and obtain a corrected word;

[0041] A second acquisition module is used to acquire a plurality of multimedia contents related to the modified word;

[0042] a calculation module, configured to calculate, based on a trained predetermined model, a matching score between the plurality of multimedia contents and the modified word and a user behavior score of the plurality of multimedia contents, wherein the user behavior score is determined based on historical playback of the multimedia contents;

[0043] The push module is used to sort the plurality of multimedia contents according to the matching score and the user behavior score and output a recommendation result.

[0044] The recommendation device of the embodiment of the present application modifies the query word input by the user to obtain a plurality of modified multimedia contents, and then outputs the recommended content after sorting according to the matching score and user behavior score calculated by the trained predetermined model. This can improve the output results obtained when the user inputs fuzzy query words and recommend content that is more suitable for the user.

[0045] The vehicle system of the embodiment of the present application includes a memory and a processor, the memory stores a computer program, and the processor implements the multimedia recommendation method as described in any of the above embodiments when executing the computer program.

[0046] The vehicle system of the embodiment of the present application can make recommendations and displays based on the user's behavior data based on the text matching of the user's search content through a multimedia recommendation method.

[0047] The non-volatile computer-readable storage medium containing a computer program according to an embodiment of the present application, when the computer program is executed by one or more processors, enables the processors to implement the multimedia recommendation method according to any one of the above embodiments.

[0048] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0050] Figure 1 Schematic diagram of the multimedia recommendation method according to the embodiment of the present application;

[0051] Figure 2 It is a module schematic diagram of a recommended device according to an embodiment of the present application;

[0052] Figure 3 This is a schematic diagram of the overall architecture of the multimedia recommendation method according to the embodiment of the present application;

[0053] Figure 4 This is a schematic diagram of the process from data collection to model processing in the multimedia recommendation method according to the embodiment of the present application;

[0054] Figure 5 yes Figure 3 Schematic diagram of the online search process in;

[0055] Figure 6 Schematic diagram of the multimedia recommendation method according to the embodiment of the present application;

[0056] Figure 7 Schematic diagram of the multimedia recommendation method according to the embodiment of the present application;

[0057] Figure 8 Schematic diagram of the multimedia recommendation method according to the embodiment of the present application;

[0058] Figure 9 Schematic diagram of the multimedia recommendation method according to the embodiment of the present application;

[0059] Figure 10 It is a structural diagram of the vehicle system of the embodiment of the present application.

[0060] Description of main component symbols:

[0061] Recommendation device 1000, first acquisition module 110, correction module 120, second acquisition module 130, calculation module 140;

[0062] Vehicle system 2000, memory 220, processor 220. DETAILED DESCRIPTION

[0063] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and should not be understood as limiting the present application.

[0064] The disclosure below provides many different embodiments or examples for realizing different structures of the present application. In order to simplify the disclosure of the present application, the components and settings of specific examples are described below. Of course, they are merely examples and are not intended to limit the present application. In addition, the present application may repeat reference numbers and / or reference letters in different examples, and such repetition is for the purpose of simplicity and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed. In addition, the present application provides examples of various specific processes and materials, but those of ordinary skill in the art will appreciate the application of other processes and / or the use of other materials.

[0065] See also Figure 1 The multimedia recommendation method of the embodiment of the present application includes:

[0066] S10: Obtain query words;

[0067] S20: Based on spelling and similar words, modify the query word and obtain a modified word;

[0068] S30: Acquire multiple multimedia contents related to the modified word;

[0069] S40: Calculating, based on the trained predetermined model, matching scores between the plurality of multimedia contents and the correction words and user behavior scores of the plurality of multimedia contents, where the user behavior scores are determined based on historical playback of the multimedia contents;

[0070] S50: Sort the multiple multimedia contents according to the matching scores and user behavior scores and output recommendation results.

[0071] See also Figure 2The recommendation device 1000 for recommending multimedia content in the embodiment of the present application includes a first acquisition module 110, a correction module 120, a second acquisition module 130 and a calculation module 130. The first acquisition module 110 is used to obtain a query word; the correction module 120 is used to correct the query word based on spelling and similar words and obtain a corrected word; the second acquisition module 130 is used to obtain multiple multimedia contents related to the corrected word; the calculation module 130 is used to calculate the matching scores between the multiple multimedia contents and the corrected word and the user behavior scores of the multiple multimedia contents based on the trained predetermined model, and the user behavior scores are determined according to the historical playback status of the media content; the push module is used to sort the multiple multimedia contents according to the matching scores and the user behavior scores and output recommendation results.

[0072] The multimedia recommendation method and recommendation device 1000 of the embodiment of the present application obtain a plurality of corrected multimedia contents by correcting the query words input by the user, and then output the recommended contents after sorting according to the matching scores and user behavior scores calculated according to the trained predetermined model. This can improve the output results obtained when the user inputs fuzzy query words and recommend content that is more suitable for the user.

[0073] Specifically, the recommendation device 1000 implements the multimedia recommendation method by first performing step S10, where a query term can be obtained through the first acquisition module 110. The query term can be a term that the user enters into the vehicle-mounted terminal to search for. After obtaining the query term entered by the user, step S20 can be performed to correct the query term based on spelling and similar words to obtain a corrected term. Spelling correction and similar word correction can be performed by correcting pinyin errors and replacing similar words, thereby obtaining a corrected corrected term. Furthermore, step S30 can be performed to obtain multiple multimedia contents related to the corrected term through the second acquisition device. The multimedia contents can be various media sources in the multimedia services provided by the vehicle-mounted terminal to the user, such as audio, video, songs, and news.

[0074] The calculation module 130 can then perform step S40 to calculate the matching scores between the multiple multimedia contents and the modified word and the user behavior scores of the multiple multimedia contents based on the trained predetermined model. The user behavior scores are determined based on the historical playback history of the media contents. The matching scores can be understood as the degree of matching between the multiple multimedia contents obtained by searching for the modified word and the modified word. The user behavior scores can be understood as the historical user playback history of the multiple multimedia contents obtained by searching for the modified word. For example, the historical playback history can be the number of historical playbacks, the historical playback duration, etc. The predetermined model can use a TF-IDF model, a weighted model, or other model methods to calculate the matching scores and user behavior scores.

[0075] Then, the recommendation module may take step S50 to sort the multiple multimedia contents and output a recommendation result according to the matching score and the user behavior score calculated in step S40.

[0076] The overall architecture of the multimedia recommendation method can be shown as follows Figure 3 As shown:

[0077] Users enter a query through online search. After obtaining the query term, they can first perform text retrieval of multimedia content based on spelling correction and then similar word replacement. Spelling correction, similar word replacement, and text retrieval can be performed based on interactive data combined with offline training.

[0078] For example, spelling correction, similar word replacement, and text retrieval can use the vocabulary pinyin table, similar vocabulary library, content library, etc. corresponding to the interactive data in the figure; the training methods such as vocabulary extraction, similar vocabulary extraction, user content statistics, data cleaning, data integration, and content statistics in the big data platform corresponding to the offline training in the figure can be used for spelling correction, similar word replacement, and text retrieval.

[0079] Furthermore, after the text search results are obtained, a matching score can be calculated, followed by a user behavior score. Finally, a recommendation result can be output based on the matching score and user behavior score calculation results. The corresponding matching score calculation and user behavior score calculation can also be performed based on offline training of the interactive data structure.

[0080] For example, matching score calculations can use the TF-IDF model corresponding to the interaction data in the graph, as well as the TF-IDF model and weight model within the training platform corresponding to the offline training in the graph. User behavior score calculations can use the content popularity and user content behavior corresponding to the interaction data in the graph, as well as the TF-IDF model and weight model within the training platform corresponding to the offline training in the graph.

[0081] The process of building a predetermined model can be as follows Figure 4 As shown:

[0082] First, data collection can be performed. The vehicle computer can report the embedded data to the cloud and forward it to Kafka (a message channel). The big data platform consumes the Kafka (message channel) data and enters the data warehouse. Business data in the cloud, including basic information about users and content, is synchronized to the big data platform.

[0083] Data processing then begins by modeling the embedded data and business data, calculating user and music features based on demand. For multimedia content text data, we extract vocabulary and associated vocabulary, perform custom word segmentation on the content text based on the extracted vocabulary, remove stop words, and consolidate and store the album, content name, and author name after word segmentation.

[0084] Then, the word weight model is trained for the query words. The TF-IDF model can be trained on the text data of the user's voice search to calculate the weight of the words in the query words for the subsequent text matching.

[0085] The TF model calculation method is the ratio of the number of times a word appears in a text to the number of times all words appear in the text during prediction.

[0086]

[0087] Among them, tf i,j represents the i-th word in the j-th sentence of the query; n i,j Indicates the number of times the i-th word in the j-th sentence of the query word appears in the j-th sentence; ∑ k n k,j Represents the sum of the number of times the i-th word in the j-th sentence of the query word appears in all the words in the j-th sentence.

[0088] For example, the user enters two query words: "Play, fire truck," and "Oh, a song. I love the country but I love the beauty." Take the first sentence "Play, fire truck" as an example. The first sentence is segmented to obtain the words "Play" and "Fire truck." Taking "Play" as an example, the TF of "Play" is 1 / 2. The principle is as follows:

[0089] tf i,j =tf 1,1 Indicates the first word in the first sentence of the query. i,j =n 1,1 =1 means the word "play" appears once in the first sentence of the query. k n k,j =n 1,1 +n 2,1 =1+1=2, that is, the first sentence segmentation results in the word "play" and the word "fire truck". The word "play" appears once in the first sentence of the query, and "fire truck" appears once in the first sentence of the query. Therefore, the first sentence segmentation results in the two words "play" and "fire truck", that is, the total number of words in the first sentence is 2.

[0090] The IDF calculation method can learn the number of times each word appears in the query word during model training, and perform dictionary queries based on the predicted text words during prediction.

[0091]

[0092] Among them, {j:t i ∈d j} represents the number of sentences in which a single word appears in the query. || represents the total number of sentences in the query.

[0093] The model data for prediction is the product of TF and IDF, which serves as the weight for each target word. The output of the model file can be saved in PMML format.

[0094] The overall process of online search can be as follows Figure 5 As shown, the user inputs voice request to the voice service on the vehicle side, and requests the search service for the text query. The search service processes the text and queries the data within the search service, then calls the model from the model repository and returns the result to the search service, and then returns the final data to the vehicle side.

[0095] In some embodiments, obtaining a query term includes:

[0096] Use the input text as a query; and / or,

[0097] The input speech parsed text is used as the query word.

[0098] In this way, the text input and voice parsed as query words can be used by users in different scenarios.

[0099] Specifically, the input text is used as a query word, or the input speech-parsed text is used as a query word, or the input text and the input speech-parsed text are used as query words at the same time.

[0100] See also Figure 6 In some embodiments, modifying the query term and obtaining the modified term (step S20) includes:

[0101] S21: Based on the pronunciation of the spelling, query the word library to confirm the correctness of the query word;

[0102] S22: If the query term is incorrect, correct the query term and obtain a corrected term;

[0103] S23: Based on semantics, query a similar word library to identify similar words to the query word;

[0104] S24: When the query word has similar words, the similar words are used as correction words.

[0105] The correction module 120 is used to query the word library based on the spelling of the pronunciation to confirm the correctness of the query word; and to correct the query word and obtain the corrected word when the query word is incorrect; and to query the similar word library based on semantics to confirm the similar words of the query word; and to use the similar word as the corrected word when the query word has similar words.

[0106] In this way, the query word is corrected by confirming the pronunciation, spelling and semantics of the query word from the word library.

[0107] Specifically, to implement step S20, the spelling of the query term can be corrected first and then similar words can be replaced. Step S21 is taken to query the vocabulary based on the spelling of the pronunciation to confirm the correctness of the query term; then step S22 is taken to correct the query term and obtain the corrected word if the query term is incorrect; then step S23 is taken to query the similar vocabulary based on semantics to confirm the similar words of the query term; then step S24 is taken to use the similar word as the corrected word if the query term has similar words.

[0108] For example, if the query word "California Hotel" entered by the user has a spelling error, "California Hotel" can be corrected to "California Hotel" through steps S21 and S22. Further, through steps S23 and S24, similar words can be replaced on the corrected "California Hotel", and "California Hotel" can be replaced with "Hotel California".

[0109] See also Figure 7 In some embodiments, the calculation of the matching score (step S40) includes:

[0110] S41: Calculate the data length in the multimedia content;

[0111] S42: Determine the length of the text to be matched;

[0112] S43: Calculate the weight of the target word in the predetermined model output sentence;

[0113] S44: Calculate the matching score based on the data length, the text length and the weight of the target word.

[0114] The calculation module 130 is used to calculate the data length in the multimedia content; to determine the length of the text to be matched; to calculate the weight of the target word in the predetermined model output sentence; and to calculate the matching score based on the data length, the length of the text and the weight of the target word.

[0115] In this way, the matching score is calculated using the data length, the text length and the weight of the target word to reflect the matching degree of the target word in the query word to the multimedia content.

[0116] Specifically, the matching score calculation in step S40 may first be performed by calculating the data length in the multimedia content in step S41; then, step S42 may be performed to determine the length of the text to be matched; then, step S43 may be performed to calculate the weight of the target word in the sentence output by the predetermined model; and finally, step S44 may be performed to calculate the matching score based on the data length, the length of the text, and the weight of the target word. The predetermined model may be a TF-IDF model or a weighted model.

[0117] See also Figure 8In some embodiments, calculating the weight of the target word in the predetermined model output sentence (step S43) includes:

[0118] S431: Calculate the ratio of the number of times the target word appears in the output sentence to the number of times all words appear in the output sentence;

[0119] S432: Calculate the frequency of the target word in the document;

[0120] S433: The product of the ratio and the frequency of occurrence is used as the weight of the target word.

[0121] The calculation module 130 is used to calculate the ratio of the number of times the target word appears in the output sentence to the number of times all words appear in the output sentence; and to calculate the frequency of the target word in the document; and to use the product of the ratio and the frequency of occurrence as the weight of the target word.

[0122] In this way, the weight of the target word in the sentence is calculated by multiplying the number of times and the frequency of occurrence.

[0123] Specifically, in step S431, the ratio of the number of times the target word appears in the output sentence to the number of times all words appear in the output sentence can be calculated using the TF model. In step S432, the frequency of the target word in the document can be calculated using the IDF model. Then, in step S433, the product of the ratio calculated in step S431 and the frequency of occurrence in step S432 can be used as the weight of the target word.

[0124] In certain embodiments, the matching score is calculated using the following formula:

[0125]

[0126] Among them, total_w word The sum of the weights of the matched target words; mappingWord len The length of the data in the matched multimedia content; Text1 len The length of the query text; Text2 len The length of the text of the target word.

[0127] In this way, the calculation formula used for the matching score can more accurately obtain the matching degree between the target word and the multimedia content.

[0128] Specifically, the calculation formula for total_w word The sum of the weights of the matched target words, where the weight can be the weight of the target word in the sentence predicted by the TF-IDF model in the prediction model; mappingWord lenThe length of the data in the matched multimedia content; Text1 len The length of the query text, that is, the length of the query text entered by the user; Text2 len The length of the text of the target word, that is, the length of the text of the target word in the query word.

[0129] See also Figure 9 In some embodiments, the calculation of user behavior scores includes:

[0130] S45: Calculating normalized values ​​of characteristic data in the multimedia content, the characteristic data including at least one of the number of plays, number of people who play the content, average play time, number of plays by the same user, and play time;

[0131] S46: Calculate the weight of the feature data;

[0132] S47: Accumulate the product of the normalized value and the weight of the feature data as the user behavior score.

[0133] The calculation module 130 is used to calculate the normalized value of the feature data in the multimedia content, where the feature data includes at least one of the number of plays, the number of players, the average play time, the number of plays by the same user, and the play time; and the weight for calculating the feature data; and the value for accumulating the product of the normalized value and the weight of the feature data as the user behavior score.

[0134] In this way, the calculation of user behavior scores can show behavioral data such as user preferences for multimedia content.

[0135] Specifically, the user behavior score can be calculated after the user matching score is calculated. Step S45 can be used to calculate the normalized value of the feature data in the multimedia content, where the feature data includes at least one of the number of plays, the number of people who play the content, the average play duration, the number of plays by the same user, and the play duration. Step S46 can then be used to calculate the weight of the feature data. Step S47 can then be used to accumulate the product of the normalized value and the feature data weight to obtain the user behavior score.

[0136] Exemplarily, user clicks and music features are normalized, and then the size of the information entropy of each feature is calculated. The weight of the feature data is determined based on the information entropy, which can then be used to calculate the user behavior score.

[0137] The information entropy calculation formula is:

[0138]

[0139] Among them, pi j Represents the normalized ratio of the data values ​​for each dimension. jRepresents the data value in row i and column j in a data table or a two-dimensional array. Indicates the sum of all rows of data in each column.

[0140] For example, a column of data is: 1, 2, 4, 6. The sum is: 1+2+4+6=13; the calculated values ​​are 1 / 13, 2 / 13, 4 / 13, 6 / 13.

[0141] The information entropy calculation formula also includes:

[0142]

[0143] in, Represents the sum of the data calculated by pij in the above formula. Indicates the number of all rows, Ln is the logarithm with base e, E j Represents the information entropy of the j-th column data.

[0144] The information weight calculation formula is:

[0145]

[0146] in, Indicates that E in the above formula j The sum of the calculated data.

[0147] Referring to the figure, in certain embodiments, the normalized value is calculated using the following formula:

[0148] X norm =(X-Xmin) / (Xmax-Xmin)

[0149] Wherein, X is the feature data, Xmin is the minimum value of the feature data, and Xmax is the maximum value of the feature data.

[0150] In this way, normalization can facilitate the sorting of feature data and the subsequent calculation of user behavior scores.

[0151] Specifically, X norm Indicates the normalized value of feature data. Taking the feature data of the number of times a user plays multimedia content as an example, if the data is 1, 3, 4, and 6, the data obtained by calculating the normalized value is:

[0152] The result of the calculation of data "1" is "(1-1) / (6-1)=0", and the result of the calculation of data "3" is The result of the calculation of "4" is The result of the calculation of "6" is "(6-1) / (6-1)=1", and a normalized value between 0 and 1 can be obtained.

[0153] Then the calculation formula for user behavior score can be:

[0154]

[0155] Among them, w j is the weight of each column of data, X norm Indicates the normalized eigenvalue of each column, score i This means the calculated score for each row of data is the behavior score for each user.

[0156] In summary, in step S50, multiple multimedia content items are sorted and recommended based on the matching scores and user behavior scores. The matching scores and user behavior scores are then reconciled, sorted from high to low, and the content is then output. This reconciliation can be a cumulative sum of a certain ratio, for example, a matching score of 0.6 plus a user behavior score of 0.4. The content output varies for different users.

[0157] For example, the query word input by user 10001 is "play Xiao Ming's songs".

[0158] The output effect of user 10001 is as follows:

[0159] Album Title Song Title Artist Name A rain Xiao Ming B Tree Xiao Ming C Creek Xiao Ming

[0160] The query term input by user 10002 is also "play Xiao Ming's songs".

[0161] The output effect of user 10002 is as follows:

[0162] Album Title Song Title Artist Name C Creek Xiao Ming D Maple Leaf Xiao Ming F piano Xiao Ming

[0163] It can be understood that content that matches the user and is consistent with the user's behavior will be recommended depending on the user.

[0164] See also Figure 10 The vehicle system 2000 of the embodiment of the present application includes a memory 210 and a processor 220. The memory 210 stores a computer program, and when the processor 220 executes the computer program, any multimedia recommendation method in the above-mentioned embodiment is implemented.

[0165] The vehicle system 2000 of the embodiment of the present application can make recommendations and displays based on the user's behavior data based on the text matching of the user's search content through a multimedia recommendation method.

[0166] Specifically, the vehicle system 2000 may be an intelligent vehicle system in a car, which may be used to implement vehicle-machine interaction, media playback, etc. The memory 210 and the processor 220 may be provided on the vehicle system 2000 .

[0167] The non-volatile computer-readable storage medium containing the computer program according to the embodiment of the present application, when the computer program is executed by one or more processors, enables the processors to implement the multimedia recommendation method according to any one of the above-mentioned embodiments.

[0168] The processor of a computer program may be a central processing unit (CPU). The processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of these chips.

[0169] Computer programs can be stored in a memory. Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above-described method embodiments. The processor executes the non-transitory software programs, instructions, and modules stored in the memory to perform various functional applications and data processing of the processor, thereby implementing the methods in the above-described method embodiments.

[0170] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0171] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0172] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0173] It should be understood that various parts of the embodiments of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0174] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0175] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0176] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0177] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are illustrative and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above implementation methods within the scope of the present application.

Claims

1. A multimedia recommendation method, characterized in that: include: Get the query term; Based on spelling and similar words, modify the query word and obtain a modified word; acquiring a plurality of multimedia contents related to the modified word; Calculating, based on the trained predetermined model, matching scores between the plurality of multimedia contents and the modified words and user behavior scores of the plurality of multimedia contents, wherein the user behavior scores are determined based on historical playback of the multimedia contents; sorting the plurality of multimedia contents according to the matching score and the user behavior score and outputting a recommendation result; The calculation of the matching score includes: Calculating the data length in the multimedia content; Determine the length of the text that needs to match the query term; Calculating the weight of the target word in the predetermined model output sentence; Calculating the matching score based on the length of the data, the length of the text, and the weight of the target word; The calculating the weight of the target word in the output sentence of the predetermined model includes: Calculating the ratio of the number of times the target word appears in the output sentence to the number of times all words appear in the output sentence; Calculate the frequency of occurrence of the target word in the document; The product of the ratio and the frequency of occurrence is used as the weight of the target word.

2. The method according to claim 1, characterized in that The acquisition of query words includes: Using the input text as the query term; and / or, The input speech-parsed text is used as the query word.

3. The method according to claim 1, characterized in that The step of modifying the query word based on spelling and similar words to obtain a modified word includes: Based on the pronunciation of the spelling, query the word library to confirm the correctness of the query word; If the query word is incorrect, correct the query word and obtain a corrected word; Based on semantics, query a similar word library to identify similar words to the query word; In the case where the query word has similar words, the similar words are used as the modified words.

4. The method according to claim 1, wherein The matching score is calculated using the following formula: Among them, total_w word The sum of the weights of the matched target words; mappingWord len The length of the data in the multimedia content; Text1 len The length of the query text; Text2 len The length of the text of the target word.

5. The method according to claim 1, wherein The calculation of the user behavior score includes: Calculating a normalized value of characteristic data in the multimedia content, the characteristic data including at least one of the number of plays, the number of people who play the content, the average play time, the number of plays by the same user, and the play time; Calculating the weight of the feature data; The product of the normalized value and the weight of the feature data is accumulated as the user behavior score.

6. The method according to claim 5, characterized in that The normalized value is calculated using the following formula: X norm =(X-Xmin) / (Xmax-Xmin) Wherein, X is the feature data, Xmin is the minimum value of the feature data, and Xmax is the maximum value of the feature data.

7. A recommendation device, characterized in that: The recommended device includes: A first acquisition module is used to acquire a query word; A correction module, configured to correct the query word based on spelling and similar words and obtain a corrected word; A second acquisition module is used to acquire a plurality of multimedia contents related to the modified word; a calculation module, configured to calculate, based on a trained predetermined model, a matching score between the plurality of multimedia contents and the modified word and a user behavior score of the plurality of multimedia contents, wherein the user behavior score is determined based on historical playback of the multimedia contents; a push module, configured to sort the plurality of multimedia contents according to the matching score and the user behavior score and output a recommendation result; The calculation module is also used to calculate the data length in the multimedia content; determine the length of the text of the query word that needs to be matched; calculate the weight of the target word in the output sentence of the predetermined model; obtain the matching score based on the data length, the length of the text and the weight of the target word; calculate the ratio of the number of times the target word appears in the output sentence to the number of times all words in the output sentence appear; calculate the frequency of occurrence of the target word in the document; and use the product of the ratio and the frequency of occurrence as the weight of the target word.

8. A vehicle computer system, characterized in that: The vehicle system includes a memory and a processor, the memory stores a computer program, and the processor implements the multimedia recommendation method according to any one of claims 1 to 6 when executing the computer program.

9. A non-volatile computer-readable storage medium containing a computer program, characterized in that When the computer program is executed by one or more processors, the processors are enabled to implement the multimedia recommendation method according to any one of claims 1 to 6.

Images