Title Generation Method, Device, Electronic Device and Readable Storage Medium

Abstract

An embodiment of the present application provides a method, apparatus, electronic device, and readable storage medium for generating a title, which relates to the field of computer technology. The method includes: obtaining set information of a target set, where the target set includes at least one object, and the set information includes query click logs of at least one object respectively, and the query click logs include query texts; performing frequency statistical analysis on the query texts in the query click logs of at least one object, and selecting at least one target query text from the query texts in the query click logs of at least one object in the order of decreasing frequency; generating a target title of the target set according to at least one target query text. In this way, the target title of the target set can be automatically generated according to the query texts with high usage frequencies corresponding to the target set, and this method can obtain a title with high accuracy, thereby improving the recommendation efficiency.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more specifically, to a method and device for generating a title, an electronic device, and a readable storage medium. Background Art

[0002] In actual search scenarios, recommendation results are generally generated based on the query description text entered by the user and the titles of various information data (such as audio and video). However, since users often find it difficult to provide precise query keywords in the query description text, the recommendation results often contain a certain degree of error, which affects the user experience. As a direct reflection of the user's perspective on the reason for the recommendation, the title is an important feature in the recommendation scenario, and its description accuracy has a significant impact on the effectiveness of the recommendation. Summary of the Invention

[0003] The embodiments of the present application provide a title generation method, device, electronic device and readable storage medium, which can automatically generate target titles for the target set based on the frequently used query texts included in the query click logs of each object in the target set, thereby improving the accuracy of the title and the recommendation efficiency.

[0004] The embodiments of the present application can be implemented as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for generating a title, the method comprising:

[0006] Obtaining set information of a target set, wherein the target set includes at least one object, the set information includes a query click log of each of the at least one object, and the query click log includes a query text;

[0007] Performing frequency statistical analysis on query texts in the query click log of the at least one object, and selecting at least one target query text from the query texts in the query click log of the at least one object in descending order of the frequencies;

[0008] Generate a target title of the target set according to the at least one target query text.

[0009] In a second aspect, an embodiment of the present application provides a title generation device, the device comprising:

[0010] An information acquisition module, configured to obtain set information of a target set, wherein the target set includes at least one object, the set information includes a query click log of each of the at least one object, and the query click log includes a query text;

[0011] An analysis module for performing frequency statistical analysis on the query texts in the query click logs of the at least one object, and selecting at least one target query text from the query texts in the query click logs of the at least one object in descending order of the frequency.

[0012] A generation module for generating a target title of the target set according to the at least one target query text.

[0013] In a third aspect, an embodiment of the present application provides an electronic device, including a processor and a memory, where the memory stores machine-executable instructions that can be executed by the processor, and the processor can execute the machine-executable instructions to implement the title generation method in the foregoing implementation manner.

[0014] In a fourth aspect, the present application provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the title generation method in the foregoing implementation manner.

[0015] The title generation method, device, electronic device, and readable storage medium provided by the embodiments of the present application first obtain the set information of the target set, where the target set includes at least one object, and the set information includes the query click logs of the at least one object respectively, and the query click logs include query texts; based on the set information, perform frequency statistical analysis on the query texts, and select at least one target query text from the query texts in the query click logs of the at least one object in descending order of the frequency; finally, generate a target title of the target set according to the obtained at least one target query text. In this way, a title for facilitating the user to retrieve the target set can be generated based on the target query texts with higher frequencies when searching and clicking on the objects in the target set, and this method can improve the accuracy of the title and the recommendation efficiency based on the title. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a block diagram of the electronic device provided by the embodiment of the present application;

[0018] Figure 2 It is a flowchart of the title generation method provided by the embodiment of the present application;

[0019] Figure 3 ForFigure 2 Flow diagram of the sub-steps included in step S120 in

[0020] Figure 4 For Figure 2 Flow diagram of the sub-steps included in step S130 in

[0021] Figure 5 Flow diagram of obtaining the rewriting model provided by the embodiment of the present application;

[0022] Figure 6 For Figure 4 Flow diagram of the sub-steps included in sub-step S131 in

[0023] Figure 7 One of the block diagrams of the title generation device provided by the embodiment of the present application;

[0024] Figure 8 Another block diagram of the title generation device provided by the embodiment of the present application.

[0025] Icons: 100 - electronic device; 110 - memory; 120 - processor; 130 - communication unit; 200 - title generation device; 201 - training module; 210 - information acquisition module; 220 - analysis module; 230 - generation module. Detailed implementation manners

[0026] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0028] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0029] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features therein may be combined with each other.

[0030] Please refer to Figure 1 , Figure 1 This is a block diagram of an electronic device 100 provided in an embodiment of the present application. The electronic device 100 may be, but is not limited to, a computer, a server, or the like. The electronic device 100 may include a memory 110, a processor 120, and a communication unit 130. The memory 110, the processor 120, and the communication unit 130 are electrically connected to each other, directly or indirectly, to enable data transmission or interaction. For example, these components may be electrically connected to each other via one or more communication buses or signal lines.

[0031] The memory 110 is used to store programs or data. The memory 110 may be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc.

[0032] The processor 120 is used to read / write the data or programs stored in the memory 110 and perform corresponding functions. For example, a title generation device 200 is stored in the memory 110, and the title generation device 200 includes at least one software function module that can be stored in the memory 110 in the form of software or firmware. The processor 120 executes various functional applications and data processing by running the software programs and modules stored in the memory 110, such as the title generation device 200 in the embodiments of the present application, that is, the title generation method in the embodiments of the present application is implemented.

[0033] The communication unit 130 is used to establish a communication connection between the electronic device 100 and other communication terminals through a network, and is used to transmit and receive data through the network.

[0034] It should be understood that Figure 1 The structure shown is only a schematic diagram of the structure of the electronic device 100, and the electronic device 100 may further include Figure 1 more or fewer components than those shown, or have a different configuration from Figure 1 that shown. Figure 1 Each component shown can be implemented by hardware, software, or a combination thereof.

[0035] Please refer to Figure 2 , Figure 2 which is a schematic flowchart of the title generation method provided by the embodiments of the present application. The method can be applied to the above-mentioned electronic device 100. The specific process of the title generation method will be elaborated in detail below. In this embodiment, the method may include step S110 to step S130.

[0036] Step S110, obtain the set information of the target set.

[0037] In this embodiment, the target set is the set for which a title needs to be generated, and can be specifically determined according to actual requirements. The target set includes at least one object, and the object can be audio, video, document, etc. Correspondingly, the target set can be an audio set, a video set, a document set, etc. The set information of the target set can be obtained from other devices or in other ways. The set information may include the query click logs of each of the at least one object, and the query click logs include query texts. The query text in the query click log can be the query description text input by the user when searching before clicking on an object, or the text obtained after performing preset preprocessing (such as removing sensitive words, garbled codes, etc.) on the query description text input by the user, etc.

[0038] Optionally, the set information may further include each object and the title of each object, etc. For example, the set information of an audio set may include each piece of audio and the title of each piece of audio.

[0039] Step S120, perform frequency statistical analysis on the query texts in the query click logs of the at least one object, and select at least one target query text from the query texts in the query click logs of the at least one object in the order of decreasing frequency.

[0040] In the case of obtaining the set information, based on the set information, frequency statistical analysis can be performed on the query texts included in the set information according to a preset frequency statistical method, and at least one target query text can be selected from the query texts in the query click logs of the at least one object in the order of decreasing frequency. In this way, target query texts with relatively high frequencies of use when searching for and clicking on objects in the search click target set can be obtained. Among them, the frequency statistical method can be specifically set according to actual needs.

[0041] Step S130, generate a target title for the target set according to the at least one target query text.

[0042] In the case of selecting the at least one target query text based on frequency statistics, a title that facilitates the user to retrieve the target set can be generated based on the at least one target query text. In this way, the accuracy of the title and the recommendation efficiency based on the title can be improved.

[0043] Optionally, as a possible implementation manner, the target set may include only one object. In this case, a title for the object can be regenerated according to the query text in the query click log of the object. In this way, when the user expects to retrieve the object subsequently, the probability of the user retrieving the object can be increased. Compared with the method of manually renaming the title of the object, this method can save a large amount of manpower and improve the retrieval efficiency at the same time.

[0044] Optionally, as another possible implementation manner, the target set may include multiple objects. When the set includes multiple objects, usually, a summary is made manually according to the titles of each object itself and the content included in each object to determine the title of the set. However, when the number of objects included in the set is very large and / or the content data volume of each object is relatively large, the above method cannot complete the task. In view of this situation, in this embodiment, a title with high accuracy can be automatically generated for the target set corresponding to the multiple objects according to the query texts in the query click logs of the multiple objects. In this way, the task of determining the title of the target set can be completed, and the accuracy is high, which is convenient for improving the retrieval efficiency.

[0045] In some scenarios, objects may have uneven qualities. In such cases, if all the searched objects are directly pushed to the user, low-quality objects will appear in the recommended results. For example, in a large audio sharing platform, users can freely upload and share original album works. Due to the uncertainty of user quality, the album works naturally have uneven qualities. To avoid the above situation, low-quality objects can be filtered first so that the qualities of the objects included in the target set all meet the preset requirements. In this way, high-quality objects can be recommended to the user, and low-quality objects can be avoided in the recommended results.

[0046] For example, the target set includes multiple audio files, and the audio instructions of each audio file meet the preset requirements. When the user retrieves this target set, since all the audio files in the target set are of high quality, there will be no situation where the recommended target set includes low-quality audio.

[0047] Optionally, the multiple objects included in the target set can be objects of the same content category. In this way, when the user retrieves, it is convenient to recommend objects of the same category to the user, which can reduce the user's retrieval cost while helping the user find personal preferences and improve the recommendation effect. Among them, the specific classification method of the object can be determined in combination with actual needs. For example, it can be aggregated according to multi-dimensional strategies such as the main content, usage scenario, and user behavior, which will not be specifically limited here.

[0048] The following is an example of the above target set including multiple objects.

[0049] In a large audio sharing platform, to avoid low-quality audio in the recommended results and ensure that high-quality audio can be pushed to the target user group as much as possible in the search scenario, high-quality audio can be aggregated according to a fixed strategy to form a listening list (or called a sound set, audio set, etc.), thereby optimizing the user recommendation results. That is, a listening list includes multiple audio files of the same content category, and the audio quality of each audio file meets the preset requirements. Specifically, the introduction of the listening list concept, on the one hand, filters out low-quality sounds and overall improves the quality of the recommended results returned; on the other hand, aggregating according to multi-dimensional strategies such as the theme content, usage scenario, and user behavior can reduce the user's retrieval cost while helping the user find personal preferences and improve the recommendation effect.

[0050] The following explains how to generate the title of the target set.

[0051] Optionally, as a possible implementation, when obtaining the set information, the usage frequency of each query text included in the set information can be statistically obtained. Then, the query texts can be sorted in descending order of usage frequency. Finally, according to the obtained sorting result, at least one query text is selected from the sorting result as the target query text. For example, the top-5 in the sorting result are selected as the target query texts, or the query texts with a usage frequency greater than a certain preset value are used as the target query texts. At this time, the usage frequency of the target query text is not less than that of the unselected query texts. In this way, the target query text can be quickly obtained.

[0052] Optionally, as another possible implementation, the target query text can be obtained in the manner Figure 3 shown. Please refer to Figure 3 , Figure 3 which Figure 2 is a schematic flowchart of the sub-steps included in step S120 in

[0053] Sub-step S121, for each of the objects, according to the query click log of the object, statistically obtain the first occurrence frequency of each query text corresponding to the object, and select the initial target query text from the query texts corresponding to the object in descending order of the first occurrence frequency.

[0054] In this embodiment, for one object, based on the query click log of the object, the frequency of each query text included in the query click log of the object that appears in the query click log of the object can be statistically obtained, so as to obtain the first occurrence frequency of each query text corresponding to the object. For other objects, repeat the above process to obtain the first occurrence frequency of each query text corresponding to each of the objects.

[0055] When obtaining the first occurrence frequency of each query text corresponding to an object, the query texts corresponding to the object can be sorted in descending order of the first occurrence frequency to obtain the first sorting result. Then, from the first sorting result, at least one query text is selected in descending order of the first occurrence frequency as the initial target query text obtained from the query click log of the object. For example, the top-3 in the first sorting result are selected as the initial target query texts, or the query texts with a first occurrence frequency greater than the first preset value in the first sorting result are used as the initial target query texts. At this time, the first occurrence frequency of each of the initial target query texts corresponding to an object is not less than the first occurrence frequency of the unselected query texts corresponding to the object.

[0056] For example, if the query click logs of object 1 include query texts M, N, and P, then according to the query click logs of this object, the first occurrence frequencies of the query texts M, N, and P in the query click logs of this object can be statistically obtained. If the first occurrence frequencies corresponding to the query texts M, N, and P of this object are m, n, and p respectively, and m > n > p, the query texts M and N can be selected as the initial target query texts obtained from the query click logs of this object.

[0057] Sub-step S122: Obtain the second occurrence frequencies of the initial target query texts in the query click logs corresponding to the target set according to the initial target query texts corresponding to each object and the first occurrence frequencies of the initial target query texts.

[0058] In the case of obtaining the initial target query texts and their first occurrence frequencies from the query click logs of each object, statistical analysis is performed to obtain the second occurrence frequencies of the initial target query texts in the query click logs corresponding to the target set. For example, if a certain initial target query text only appears in the query click logs of a certain object, then the second occurrence frequency of this initial target query text is the same as its first occurrence frequency.

[0059] If multiple objects correspond to the same initial target query text, it is necessary to accumulate the first occurrence frequencies of this initial target query text in the query click logs of each object to obtain the second occurrence frequency of this initial target query text. For example, continuing with the above example of selecting the initial target query texts M and N, assuming that the initial target query text N is obtained from the query click logs of object 2, and the first occurrence frequency of this initial target query text N in the query click logs of object 2 is n1. Since this text is also the initial target query text selected from the query click logs of object 1 and the first occurrence frequency in the query click logs of object 1 is n, after summary statistics, the second occurrence frequency of the initial target query text in the query click logs corresponding to the target set can be obtained as n + n1.

[0060] Sub-step S123: Select the target query texts from the initial target query texts in descending order of the second occurrence frequencies according to the second occurrence frequencies of the initial target query texts.

[0061] When the second occurrence frequency of each initial target query text is obtained, the initial target query texts can be sorted in descending order of the second occurrence frequency to obtain a second sorting result. Then, at least one initial target query text is selected as the target query text from the second sorting result in descending order of the second occurrence frequency. For example, the top-5 in the second sorting result is selected as the target query text, or the initial target query texts with the second occurrence frequency greater than a second preset value in the second sorting result are used as the target query texts. At this time, the second occurrence frequency of each target query text is not less than that of each unselected initial target query text.

[0062] Optionally, when there are multiple objects in the target set, multiple initial target query texts can be selected from the query click logs of each object, and multiple target query texts can be selected.

[0063] When the target query text is obtained, as a possible implementation, one target query text can be directly selected from the obtained target query texts as the title of the target set. In this way, the target title of the target set can be quickly obtained.

[0064] As another possible implementation, it can be obtained through Figure 4 as shown. Please refer to Figure 4 , Figure 4 is Figure 2 a schematic flowchart of the sub-steps included in step S130 in

[0065] Sub-step S131: Using a preset rewriting model, based on each target query text, obtain at least one rewritten result corresponding to each target query text.

[0066] In this embodiment, the rewriting model can be pre-trained by the electronic device 100 for generating titles, or can be trained by other devices and then sent to the electronic device 100. Optionally, the rewriting model can be obtained through Figure 5 as shown. Please refer to Figure 5 , Figure 5 is a flowchart of obtaining the rewriting model provided by an embodiment of the present application. In this embodiment, the rewriting model can be obtained through Figure 5 the steps S101 to S103 shown.

[0067] Step S101: Obtain a plurality of sample titles.

[0068] Step S102: Process each of the sample titles to obtain at least one sample word segmentation corresponding to each of the sample titles.

[0069] Step S103: Train the rewriting model based on each of the sample titles and at least one sample word segmentation corresponding to each of the sample titles.

[0070] In this embodiment, since the input during the application of the rewriting model is the query length, the data construction in the model training stage also needs to correspond to the query length (short text). During the data construction process, open news corpora, etc. can be used to collect coherent news titles, etc. as the sample titles. Then, process each of the collected sample titles to obtain at least one sample word segmentation corresponding to each of the sample titles.

[0071] For example, for a sample title, first remove the stop words in the sample title. Optionally, the string in the sample title can be matched with a preset string, where the preset string is a pre-set stop word, such as meaningless words like "de", "a", "o". After that, the substrings in the sample title that match the preset string can be deleted. After removing the stop words, word segmentation can be performed. Then, for the obtained word segmentations, remove words with low-information part-of-speech such as auxiliary words, conjunctions, and prepositions, and use the finally obtained word segmentations as at least one sample word segmentation corresponding to the sample title.

[0072] In the case of obtaining each sample title and at least one sample word segmentation corresponding to each sample title, the at least one sample word segmentation corresponding to a sample title can be used as the input, and the sample title can be used as the label to train a preset network to obtain the rewriting model.

[0073] Optionally, the rewriting model can include a seq2seq model, a transformer model, etc. The output of the rewriting model can include at least one rewritten title and the probability score corresponding to each rewritten title, etc. Among them, the range of the probability score can be [0 - 1].

[0074] In the case of obtaining the target query text, the rewriting model can be used to obtain at least one rewritten result corresponding to each target query text based on each target query text. Among them, the rewritten result includes candidate titles.

[0075] Please refer to Figure 6 , Figure 6 For Figure 4 the flowchart of the sub-steps included in sub-step S131. In this embodiment, sub-step S131 can include sub-steps S1311 to S1313.

[0076] Sub-step S1311: For each of the target query texts, preprocess the target query text to obtain a preprocessing result.

[0077] Sub-step S1312: Perform word segmentation on the preprocessing result to obtain at least one word segment corresponding to the preprocessing result.

[0078] Among them, the preprocessing may include removing stop words, and this process can be the same as the method in the process of constructing data when training the model, which will not be elaborated here. A target query text after preprocessing can be used as a preprocessing result. Then, for this preprocessing result, perform word segmentation to obtain word groups. Next, for the word groups corresponding to this preprocessing result, remove the words with low information content such as auxiliary words, conjunctions, and prepositions in the word group, so as to obtain at least one word segment corresponding to the preprocessing result.

[0079] Sub-step S1313: Input at least one word segment corresponding to the preprocessing result into the rewriting model to obtain at least one rewriting result corresponding to the target query text.

[0080] Input at least one word segment corresponding to each preprocessing result into the rewriting model in sequence to obtain at least one rewriting result corresponding to each target query text. For example, a total of 5 target query texts A, B, C, D, and E are obtained. After the above processing, at least one analysis of each of the target query texts A, B, C, D, and E can be obtained. The at least one word segment corresponding to the target query text A can be input into the rewriting model to obtain at least one rewriting result corresponding to the target query text A; similarly, at least one rewriting result corresponding to each of the target query texts B, C, D, and E can be obtained in sequence.

[0081] Sub-step S132: Select the target title from the obtained candidate titles.

[0082] Optionally, one candidate title in one of the rewriting results corresponding to each target query text can be selected as the target title. The specific selection method can be set according to actual needs. For example, select the candidate title with the longest or shortest length among the obtained candidate titles as the target title.

[0083] As a possible implementation manner, the rewriting result further includes the score of the candidate title, and the candidate title serving as the target title can be determined according to the scores of the candidate titles. For example, the candidate title with the highest score can be selected as the target title, or any candidate title with a score greater than the preset score can be selected as the target title, etc.

[0084] For example, five target query texts A, B, C, D, and E are obtained. Based on the target query text A, using a rewriting model, three candidate titles A1, A2, A3 and their respective scores are obtained. Similarly, based on the target query text B, using the rewriting model, three candidate titles B1, B2, B3 and their respective scores are obtained. And so on, the candidate titles and scores corresponding to the target query texts C, D, and E are also obtained. The candidate title with the highest score can be selected from the candidate titles A1, A2, A3, B1, B2, B3... as the target title.

[0085] To execute the corresponding steps in the above embodiments and each possible manner, an implementation manner of a title generation device 200 is given below. Optionally, the title generation device 200 may adopt the device structure of the electronic device 100 shown above. Further, please refer to Figure 1 The device structure of the electronic device 100 shown. Further, please refer to Figure 7 , Figure 7 FIG. 211 is one of the block diagrams of the title generation device 200 provided in the embodiment of the present application. It should be noted that the basic principle and the technical effects generated by the title generation device 200 provided in this embodiment are the same as those in the above embodiments. For the sake of brief description, for the parts not mentioned in this embodiment, reference may be made to the corresponding content in the above embodiments. The title generation device 200 may include: an information acquisition module 210, an analysis module 220, and a generation module 230.

[0086] The information acquisition module 210 is configured to acquire the set information of the target set. Wherein, the target set includes at least one object, and the set information includes the query click logs of the at least one object respectively, and the query click logs include query texts.

[0087] The analysis module 220 is configured to perform frequency statistical analysis on the query texts in the query click logs of the at least one object, and select at least one target query text from the query texts in the query click logs of the at least one object in the order of decreasing frequency.

[0088] The generation module 230 is configured to generate the target title of the target set according to the at least one target query text.

[0089] Optionally, in this embodiment, the generation module 230 is specifically configured to: using a preset rewriting model, based on each of the target query texts, obtain at least one rewriting result corresponding to each of the target query texts, where the rewriting result includes candidate titles; select the target title from the obtained candidate titles.

[0090] Optionally, in this embodiment, the generation module 230 is specifically used to: preprocess each target query text to obtain a preprocessing result; perform word segmentation on the preprocessing result to obtain at least one word segmentation corresponding to the preprocessing result; input the at least one word segmentation corresponding to the preprocessing result into the rewriting model to obtain at least one rewriting result corresponding to the target query text.

[0091] Optionally, in this embodiment, the rewriting result further includes scores of the candidate titles, and the generating module 230 is specifically configured to determine a candidate title as the target title based on the scores of the candidate titles.

[0092] Optionally, in this embodiment, the analysis module 220 is specifically used to: for each of the objects, based on the query click log of the object, obtain the first occurrence frequency of each query text corresponding to the object, and select the initial target query text from the query texts corresponding to the object in descending order of the first occurrence frequency; based on the initial target query text corresponding to each object and the first occurrence frequency of the initial target query text, obtain the second occurrence frequency of each initial target query text in the query click log corresponding to the target set; based on the second occurrence frequency of each initial target query text, select the target query text from the initial target query texts in descending order of the second occurrence frequency.

[0093] Optionally, in this embodiment, the target set includes multiple audios belonging to the same content category, the audio quality of each audio meets a preset requirement, and multiple target query texts are obtained.

[0094] Please refer to Figure 8 , Figure 8 This is a second block diagram of a title generation device 200 provided in an embodiment of the present application. Optionally, in this embodiment, the title generation device 200 may further include a training module 201. The training module 201 is specifically configured to: obtain a plurality of sample titles; process each of the sample titles to obtain at least one sample segmentation corresponding to each sample title; and train the rewriting model based on each of the sample titles and the at least one sample segmentation corresponding to each sample title.

[0095] Optionally, the above modules can be stored in the form of software or firmware. Figure 1 The memory 110 shown in FIG. 110 or the operating system (OS) of the electronic device 100 may be fixed and may be used by Figure 1 Meanwhile, the data, program codes, etc. required to execute the above modules may be stored in the memory 110.

[0096] The embodiment of the present application also provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the title generation method described above is implemented.

[0097] In summary, the embodiment of the present application provides a title generation method, device, electronic device and readable storage medium. First, obtain the set information of the target set, where the target set includes at least one object, and the set information includes the query click logs of the above at least one object respectively, and the query click log includes a query text; based on the set information, perform frequency statistical analysis on the query text, and select at least one target query text from the query texts of the query click logs of the above at least one object in the order of frequency from large to small; finally, generate a target title for the target set according to the obtained at least one target query text. In this way, a title that is convenient for users to retrieve the target set can be generated based on the target query text with a higher frequency of use when searching and clicking on the objects in the target set, and this method can improve the accuracy of the title and the recommendation efficiency based on the title.

[0098] In several embodiments provided by the present application, it should be understood that the disclosed device and method can also be implemented in other ways. The device embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the drawings show the possible architectures, functions and operations of the device, method and computer program product according to multiple embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment or a part of code, and the module, program segment or part of code includes one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0099] In addition, each functional module in various embodiments of the present application may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0100] When the above-mentioned functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0101] The above are only optional embodiments of this application and are not used to limit this application. For those skilled in the art, this application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A title generation method, characterized in that, The method includes: Obtaining set information of a target set, where the target set includes at least one object, and the set information includes query click logs of each of the at least one object. The query click logs include query texts, and the query texts are determined based on query description texts input by a user during a search before clicking on the object corresponding to the query text. The object corresponding to the query text is a recommended object determined based on the query text and the title of the object; Performing frequency statistical analysis on the query texts in the query click logs of the at least one object, and selecting at least one target query text from the query texts in the query click logs of the at least one object in descending order of the frequency; Generating a target title of the target set according to the at least one target query text; Wherein, generating the target title of the target set according to the at least one target query text includes: using a preset rewriting model to obtain at least one rewritten result corresponding to each of the target query texts based on each of the target query texts, where the rewritten result includes candidate titles; and selecting the target title from the obtained candidate titles.

2. The method according to claim 1, characterized in that, The using a preset rewriting model to obtain at least one rewritten result corresponding to each of the target query texts based on each of the target query texts includes: For each of the target query texts, performing preprocessing on the target query text to obtain a preprocessing result; Performing word segmentation on the preprocessing result to obtain at least one word segment corresponding to the preprocessing result; Inputting the at least one word segment corresponding to the preprocessing result into the rewriting model to obtain at least one rewritten result corresponding to the target query text.

3. The method according to claim 1, wherein The rewritten result further includes a score of the candidate title, and the selecting the target title from the obtained candidate titles includes: Determining a candidate title as the target title according to the scores of the candidate titles.

4. The method according to claim 1, characterized in that The rewriting model is obtained through the following manner: Obtaining a plurality of sample titles; Processing each of the sample titles to obtain at least one sample word segment corresponding to each of the sample titles; Training the rewriting model according to each of the sample titles and the at least one sample word segment corresponding to each of the sample titles.

5. The method according to claim 1, characterized in that The performing frequency statistical analysis on the query texts in the query click logs of the at least one object, and selecting at least one target query text from the query texts in the query click logs of the at least one object in descending order of the frequency includes: For each of the objects, according to the query click log of the object, statistically obtaining a first occurrence frequency of each query text corresponding to the object, and selecting an initial target query text from the query texts corresponding to the object in descending order of the first occurrence frequency; Obtaining a second occurrence frequency of each initial target query text in the query click log corresponding to the target set according to the initial target query texts corresponding to each object and the first occurrence frequency of the initial target query text; Select the target query text from the initial target query texts according to the second occurrence frequency of each initial target query text, in descending order of the second occurrence frequency.

6. The method according to any one of claims 1-5, characterized in that, The target set includes multiple audios with the same content category, and the audio quality of each audio meets the preset requirements. The obtained target query texts are multiple.

7. A title generation device, characterized in that, The device includes: An information acquisition module, configured to acquire the set information of a target set, where the target set includes at least one object, the set information includes the query click logs of the at least one object respectively, the query click logs include query texts, and the query texts are determined based on the query description texts input by the user during a search before clicking on the object corresponding to the query text, and the object corresponding to the query text is a recommended object determined based on the query text and the title of the object; An analysis module, configured to perform frequency statistical analysis on the query texts in the query click logs of the at least one object, and select at least one target query text from the query texts in the query click logs of the at least one object in descending order of the frequency; A generation module, configured to generate a target title for the target set according to the at least one target query text; Among them, the generation module is specifically configured to: use a preset rewriting model to obtain at least one rewriting result corresponding to each target query text based on each target query text, where the rewriting result includes a candidate title; select the target title from the obtained candidate titles.

8. An electronic device, characterized in that, It includes a processor and a memory. The memory stores machine-executable instructions that can be executed by the processor, and the processor can execute the machine-executable instructions to implement the title generation method according to any one of claims 1-6.

9. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the title generation method according to any one of claims 1-6.

Images