Method, apparatus, device, medium and program product for determining search input intent

Abstract

Embodiments of the present disclosure relate to methods, devices, equipment, media, and program products for determining the intent of a search input, and relate to the fields of large model technology and large language model technology. The method includes determining a content description of a set of search results for a target search input and background information for the target search input. The method also includes determining a category of the target search input based on the content description and the background information. The method also includes determining the intent of the target search input based on the category of the search input. Through this method, when performing a search, the intent of the search input can be determined using the background information obtained for the search input and the content description of the search results, thereby improving the efficiency of intent recognition of the search input and the accuracy of intent recognition, and improving the user experience.

Description

Technical Field

[0001] Embodiments of the present disclosure generally relate to the field of information search, and more particularly to methods, devices, equipment, media, and program products for determining search input intent. Background Art

[0002] At present, more and more new search technologies have emerged in the field of information search. These new technologies enable users to improve the efficiency of information search and the accuracy of the information they search for in their daily search. Usually, users use natural language as input query to search for information, and sometimes they also use professional terms to search. For some highly professional search needs, users will also use different search engines to search, and compare the search results in multiple ways to determine the most relevant search results.

[0003] With the update and iteration of technologies related to information search, the search capabilities of search engines have become more and more powerful. The search library of search engines stores more and more searchable content, and the search efficiency and accuracy of search engines are gradually improving. Usually, in a very short time, users are presented with a part of the content that is most relevant to the input query and a simple summary of the content, so that users can choose the content entries that they think are most relevant. However, the current search engines are still unable to fully understand the user's input query, so a lot of optimization operations are still needed for the search engine. Summary of the invention

[0004] Embodiments of the present disclosure provide a method, apparatus, device, medium, and program product for determining search input intent.

[0005] According to a first aspect of the present disclosure, a method for determining the intent of a search input is provided. The method includes determining a content description of a set of search results for a target search input and background information for the target search input. The method also includes determining a category of the target search input based on the content description and the background information. The method also includes determining the intent of the target search input based on the category of the search input.

[0006] In a second aspect of the present disclosure, a device for determining the intent of a search input is provided. The device includes a content description and background information determination module configured to determine a content description of a set of search results for a target search input and background information for the target search input; a category determination module configured to determine the category of the target search input based on the content description and the background information; and an intent determination module configured to determine the intent for the target search input based on the category of the search input.

[0007] In a third aspect of the present disclosure, an electronic device is provided, comprising at least one processor; and a storage device for storing at least one program, wherein when the at least one program is executed by the at least one processor, the at least one processor implements the method according to the first aspect of the present disclosure.

[0008] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method according to the first aspect of the present disclosure is implemented.

[0009] In a fifth aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the method according to the first aspect of the present disclosure when executed by a processor.

[0010] It should be understood that the contents described in this content section are not intended to limit the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present disclosure.

[0012] Figure 1 A schematic diagram illustrating an example environment in which the apparatus and / or methods of some embodiments of the present disclosure may be implemented;

[0013] Figure 2 A schematic diagram illustrating an example method of determining the intent of a search input according to some embodiments of the present disclosure;

[0014] Figure 3 A schematic diagram illustrating an example of a system architecture for determining intent of a search input according to some embodiments of the present disclosure;

[0015] Figure 4 A schematic diagram illustrating an example of a flow chart for determining the intent of a search input according to some embodiments of the present disclosure;

[0016] Figure 5 A schematic diagram illustrating an example of an effect display of determining the intent of a search input according to some embodiments of the present disclosure;

[0017] Figure 6 A schematic diagram illustrating another example of an effect display of determining the intent of a search input according to some embodiments of the present disclosure;

[0018] Figure 7A schematic diagram illustrating yet another example of an effect display of determining the intent of a search input according to some embodiments of the present disclosure;

[0019] Figure 8 A schematic diagram illustrating yet another example of the effect display of determining the intent of a search input according to some embodiments of the present disclosure;

[0020] Fig. 9 A schematic block diagram of an apparatus for determining the intent of a search input according to some embodiments of the present disclosure is illustrated;

[0021] Fig.10 A schematic block diagram of an example device suitable for implementing various embodiments of the present disclosure is illustrated.

[0022] In the various drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0023] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and relevant provisions.

[0024] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0025] For example, when receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested to be performed will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0026] As an optional but non-limiting implementation, in response to receiving an active request from the user, the prompt information may be sent to the user in the form of a pop-up window, in which the prompt information may be presented in text form. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0027] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet the relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0028] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0029] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0030] In the process of identifying search intent, there are still many problems that need to be solved. For example, the user's input query is often in natural language or at least generated based on natural language. The input query based on natural language may be vague and may be ambiguous, for example, the input query has no precise requirements or contains multiple requirements. Usually, when the search engine understands the user's input query, it will first understand the user's query intent corresponding to the input query based on the input query. By understanding the user's query intent, the search engine can better understand the user's real needs, and further conduct more accurate searches to provide users with results that are more in line with the user's expectations, thereby improving the user's search experience. It is also possible to understand the user's query intent and further combine the user's historical search behavior, geographic location and other information to provide search results that meet the user's personalization. In this case, it is more helpful to accurately deliver the scheduled content, thereby improving the click-through rate and conversion rate of the scheduled content, and creating more value.

[0031] Among the above traditional solutions, there are usually several methods for understanding the user's search intent. One of them is a rule-based method, which relies on predefined rules and pattern matching to identify the intent in the text through manually written rules. The advantage of this method is that it is simple and direct, but the disadvantage is that it is difficult to expand and maintain, and has limited processing capabilities for complex language phenomena. Another method is the traditional machine learning method, which uses feature engineering and traditional machine learning algorithms (such as support vector machines, naive Bayes, decision trees, etc.) for text classification. This method requires a lot of manual feature extraction work, and the performance of the model depends on the quality of the features. Another method is the deep learning method. In recent years, deep learning methods have made significant progress in text intent recognition. Commonly used models include convolutional neural networks, recurrent neural networks, and long short-term memory networks. These models can automatically extract features and reduce manual intervention, but their ability to understand and generate natural language is not strong enough. The centralized traditional solutions mentioned above either require a lot of manual intervention or do not have a deep understanding of natural language, which will lead to the inability to accurately determine the user's intent and reduce the user experience.

[0032] At least to solve the above and other potential problems, an embodiment of the present disclosure proposes a method for determining the intent of a search input. In this method, a computing device may first determine a content description of a set of search results for a target search input and background information for the target search input. After determining the content description and background information, the computing device further uses the content description and background information to determine the category of the target search input. Finally, after determining the category of the search input, the computing device finally uses the determined category to determine the intent of the target search input. Through this method, the intent of the search input is determined using the acquired background information for the search input and the content description of the search results, which improves the efficiency of intent recognition of the search input and the accuracy of intent recognition, thereby improving the user experience.

[0033] The embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. Figure 1 An example environment in which the device and / or method of an embodiment of the present disclosure can be implemented is shown. In environment 100, computing device 102 can receive target search input 104, and can determine the intent 114 of target search input 104, so as to better understand the purpose of the user's search. Additionally, the intent can also be used to evaluate search results, thereby further optimizing the search algorithm. Target search input 104 can also be referred to as a search term, where the search term can be understood as the natural language entered by the user in the search box.

[0034] Examples of computing device 102 include, but are not limited to, a personal computer, a server computer, a handheld or laptop device, a mobile device (such as a mobile phone, a personal digital assistant (PDA), a media player, etc.), a multiprocessor system, consumer electronics, a minicomputer, a mainframe computer, a distributed computing environment including any of the above systems or devices, and the like.

[0035] like Figure 1 As shown, first, the computing device 102 can obtain a content description 110 of a group of search results 106 for the target search input 104, and background information 108 for the target search input 104. Among them, the group of search results 106 can be a collection of content searched according to the user input, including but not limited to videos, pictures, texts, etc. In addition, the target search input 104 should be a natural language text with normal logic, and the language can be the language of any country, such as "a song about xxx". If the target search input 104 is unclear, the intention cannot be identified.

[0036] In some implementation examples, the user may perform a search operation in advance in the search engine using a target search input to obtain search results. Then, the obtained search results and the target search input are stored in a storage device. The search process may be performed by a computing device 102 or any other suitable computing device. When determining the intent of the search input, the computing device 102 may obtain these search inputs and corresponding search results from the storage device. In some embodiments, a search operation for the target search input 104 may be performed in the search engine of the computing device 102, thereby obtaining a set of search results 106. The above examples are only used to describe the present disclosure, and are not specific limitations of the present disclosure.

[0037] After obtaining a group of search results 106, the computing device 102 needs to further determine a content description 110 for the group of search results. The computing device 102 can process the search results in a group of search results to obtain description information for the search results, such as a text description for the search results. For different types of search results, different processing methods can be used to obtain a text description for the search results. For example, if the search result is a video, the corresponding text description can be obtained by processing the video; if the search result is audio, the corresponding text description can be obtained by processing the audio; if the search result is text, the text can also be processed to obtain the corresponding text description. Additionally, the text description obtained by the above processing process can be summary information of the search results.

[0038] In addition, the computing device 102 may also obtain background information for the target search input 104. When determining the intent of the target search input 104, in addition to using the search results of the target search input 104, the background information 108 of the target search input 104 may be further determined. The background information 108 may be used to describe the background related to the content in the search input. Additionally, the background information 108 may be generated by using the relevant content obtained from one or more knowledge bases using the target search input 104. Each of the one or more knowledge bases also stores a large amount of information, so the target search input 104 may also be used to obtain relevant retrieval results.

[0039] After determining the content description 110 and the background information 108, the computing device 102 may further use the content description 110 and the background information 108 to determine the category 112 of the target search input 104. For example, a set of categories for the intent of the search input may be pre-set to classify the search input, such as general entertainment, information knowledge, etc. Then, the category 112 corresponding to the target search input 104 is determined from the set of categories by using the content description 110 and the background information 108. Finally, after determining the category 112 of the search input, the computing device finally uses the determined category 112 to determine the intent 114 for the target search input 104.

[0040] Through this method, when searching, the background information obtained for the search input and the content description of the search results can be used to determine the intention of the search input, thereby improving the efficiency of intent recognition of the search input and the accuracy of intent recognition, and improving the user experience.

[0041] Combined with the above Figure 1 A schematic diagram of an example environment in which the devices and / or methods of some embodiments of the present disclosure may be implemented is described below. Figure 2 A schematic diagram depicting an example method 200 for determining intent of a search input according to some embodiments of the present disclosure. Figure 2 The example method in can be represented by Figure 1 The computing device 102 and any suitable device in the system may be executed.

[0042] like Figure 2 As shown, in the example method 200, at box 202, the computing device 102 determines the content description 110 of a set of search results 106 for the target search input 104 and the context information 108 for the target search input 104. In order to better determine the intent of the target search input 104, the content description generated by the search results of the target search input 104 and the context information 108 corresponding to the target search input 104 are needed.

[0043] When selecting a target search input for which the intent is to be determined, the selected target search input 104 should be a natural language text with normal logic, such as "information about the most popular movies". In addition, the language in which the search input is executed can be a common language of any country. If the search input is a meaningless content, the intent cannot be identified, for example, if it is a string of meaningless symbols "&!!%!$%@#$!", the intent cannot be identified for the search input.

[0044] In some embodiments, the background information of the target search input 104 may be pre-generated, and when determining the intent of the target search input, the computing device 102 obtains the background information 108 from the computing device or storage device storing the background information. In some embodiments, the background information 108 is obtained by the computing device 102. For example, the computing device 102 obtains a set of search results associated with the target search input 104 from a target knowledge base storing a large amount of information. For the convenience of description and to distinguish from the search results described above, the set of search results may be referred to as a set of reference search results. Due to different data sources, the set of reference search results may be different from the set of search results 106. Alternatively, the set of reference search results and the set of search results 106 may have partially repeated content. Then, the computing device 102 extracts the background information of the set of reference search results to obtain the background information 108 corresponding to the target search input.

[0045] In some embodiments, a machine learning model may be used in the process of the computing device 102 processing a set of reference search results to obtain background information. The machine learning model is a language model, such as a large language model. For ease of description, the language model may also be referred to as a first language model. In the process of obtaining background information 108, the background information 108 of the target search input 104 is determined by applying a set of reference search results and the target search input 104 to the first language model. In some embodiments, the mapping relationship between the search results and the search input and the background information can be predetermined. Then, the background information corresponding to the set of reference search results and the target search input 104 is obtained based on the mapping relationship.

[0046] In addition, the computing device 102 also needs to obtain a content description 110. In this process, the computing device 102 first determines a group of search results 106 corresponding to the target search input 104, and then determines a text description corresponding to the search results in the group of search results 106. In addition, when the search results in the group of search results 106 contain video or image content, an optical character recognition (OCR) method can be used to recognize the text in the image or video frame to generate a corresponding text description. Alternatively, when the search results contain video or image content, a machine learning model can also be used to process the video or image to convert non-text content such as the video or image into a text description. When the search results contain audio content, an automatic speech recognition (ASR) method can be used to convert the voice information into a text description. Alternatively, the machine learning model can also be used to process the audio information to convert it into a text description. After a search result is converted into a text description, a content description 110 for a group of search results 106 can be further determined based on these text descriptions. For example, the content description 110 is a combination of the text descriptions of each search result in a group of search results. Additionally, each text description may be processed in a predetermined manner to generate the content description, for example, each text description may be organized into data in a predetermined format and then the texts may be spliced ​​together.

[0047] Additionally, the set of search results obtained by the computing device 102 is selected from the plurality of search results. For example, the computing device 102 may obtain a plurality of search results for the target search input 104 through a search engine, and the number of these results is relatively large, for example, 50 results are obtained. Then, the computing device 102 further selects a set of search results 106 from the plurality of search results according to a set of parameters associated with each search result.

[0048] For example, the computing device 102 can calculate the score of each search result based on the click-through rate, like rate, or average browsing time of each search result. The click-through rate refers to the ratio of the number of times a certain content on a website page is clicked to the number of times it is displayed. For example, if a content is displayed 1,000 times and clicked 50 times, then the click-through rate is 50 / 1,000×100%=5%. The like rate refers to the ratio of the number of likes received by the content to the exposure or browsing volume of the content. For example, if a social media article has 10,000 views and 1,000 likes, the like rate is 1,000 / 10,000×100%=10%. Average browsing time refers to the average time spent by each user when browsing specific content (such as a web page, article, video, etc.). The calculation method is to divide the total time that all users browse the content by the total number of users who browse the content. For example, if a video is watched by 100 users and the total viewing time of these users is 500 minutes, then the average browsing time per person is 500 / 100 = 5 minutes.

[0049] The computing device 102 can normalize a set of parameters (click-through rate, like rate or average browsing time per person) of each search result to obtain the sum of these parameters as the score for the search result. Additionally, different weights can be assigned to different parameters, and then this set of parameters can be weighted to determine the score of the search result. The search results can then be sorted using the scores for the search results to select a set of search results 106 from multiple search results. For example, the computing device 102 selects the search results with the top 10 scores. Additionally, the results searched by the search engine can also be converted into the same target language, such as English or Chinese. This application is not limited here.

[0050] Subsequently, at box 204, the computing device 102 determines the category 112 of the target search input 104 based on the content description 110 and the background information 108. Since the category of the target search input has a greater impact on the intent, it is necessary to determine the category 112 corresponding to the target search input 104. As mentioned above, the classification of the search input is mainly based on the intent of the search input, which can be divided into a group of categories such as pan-entertainment, information knowledge, etc. For the convenience of description, this group of categories can also be referred to as the first group of categories.

[0051] In some embodiments, when computing device 102 determines category 112, context information 108 and content description 110 may be input into a second language model to determine category 112, for example, the second language model is a large language model. In some embodiments, a mapping relationship between context and content and category may be predetermined, and then category 112 corresponding to context information 108 and content description 110 may be determined based on the mapping relationship.

[0052] Finally, at block 206, based on the category 112 of the search input, the computing device 102 determines the intent 114 for the target search input 104. After obtaining the category 112 for the target search input 104, the computing device 102 may use the category to determine the intent 114. For ease of description, this category may be referred to as a first category.

[0053] In some embodiments, computing device 102 may determine category 112 for target search input 104 as intent 114 for target search input 104. In some embodiments, computing device 102 may use other information to determine intent 114 in addition to category 112. For example, computing device 102 may also determine a result category for each search result in the search results based on content description 110, where the result category comes from a second group of categories determined by classifying the content of the search results. The second group of categories classifies the content of the search results, for example, the second group of categories includes categories such as film and television, novels, and games.

[0054] Then, the computing device 102 further determines a second category for a group of search results 106 according to the result category of each search result and the weight for the result category. For example, the group of search results includes 10 search results, of which 5 search results are classified as movies and 5 results are classified as novels, but the weight of the movie category is 0.6 and the weight of the novel category is 0.2. At this time, the score of the movie category is 3 (5 0.6), and the score for fiction was 1 (5 0.2). Therefore, the category of the group of search results 106 is the film and television category, that is, the second category is the film and television category. The computing device 102 finally uses the determined first category and second category to determine the intent 114 for the target search input 104. In one example, the computing device 102 determines the type of each search result in a group of search results by inputting the content description 110 into a large language model.

[0055] In some embodiments, in addition to using the first category and the second category to determine the intent 114 for the target search input 104, the computing device 102 further uses the precision category of the intent 114 to determine the intent. At this time, the computing device 102 determines the precision category of the target search input based on the background information 108 and the content description 110, and the precision category is used to indicate whether the search input is for a predetermined resource. For example, the precision category can indicate whether the target search input is precise or imprecise. If it is imprecise, it indicates that the search input is a general search or a search for a certain entity. If it is precise, it can be further determined that it is for a specific resource, such as a movie. Additionally, the computing device 102 obtains the precision category by inputting the background information 108 and the content description 110 into a large language model.

[0056] In some embodiments, the computing device 102 may also obtain and determine a timeliness category for the target search input 104. Timeliness categories include timeliness and non-timeliness. For example, the timeliness category is determined using background information 108 and content description 110. The timeliness category is used to indicate the degree of association between the search input and the hot content. For example, the computing device 102 inputs the background information and content description into a large language model to obtain the timeliness category. Then, the computing device 102 uses the first category, the second category, the accuracy category, and the timeliness category to determine the intention for the target search input.

[0057] In some embodiments, the computing device 102 may also use the background information and content description to determine the validity of the search input. The validity includes valid and invalid. The process mainly determines whether the search meets the requirements of laws and regulations or whether the search language input is the local mainstream language. Then, the computing device 102 determines the intention 114 for the target search input 104 based on the aforementioned first category, second category, accuracy category, timeliness category and validity.

[0058] Through this method, when searching, the background information obtained for the search input and the content description of the search results can be used to determine the intention of the search input, thereby improving the efficiency of intent recognition of the search input and the accuracy of intent recognition, and improving the user experience.

[0059] Combined with the above Figure 2 A schematic diagram of an example method 200 for determining the intent of a search input according to some disclosed embodiments is described below. Figure 3 A diagram depicting an example of a system architecture for determining intent of a search input according to some embodiments of the present disclosure.

[0060] like Figure 3As shown, in Example 300, a multi-agent architecture is adopted, with a total of 6 agents, which respectively perform intent recognition from 6 dimensions by combining knowledge base information and search recall content. An agent refers to an entity that can perceive its environment and take actions to achieve a certain goal. In fields such as computer science, artificial intelligence, and robotics, agents are usually designed as software programs or hardware devices that can make decisions and perform tasks autonomously. An agent in the present disclosure can be understood as a large language model that performs a specified intent recognition task based on a preset set of instructions. The instruction sets between different agents are different, and the models may also be different. Additionally, in this architecture, all non-English content can be uniformly converted to English for processing, and most of the corpus pre-trained for the large language model is in English, which makes it easier for the large language model to understand.

[0061] The computing device 102 obtains the search recall content 302 for the search term, which is a set of search results. The search term can also be called the search input, which is the natural language text entered by the user in the search box. The search recall content can be understood as a set of content retrieved according to the user input, including but not limited to videos, pictures, text, etc. For example, by searching for the target search input using a search engine, 50 results can be obtained. The search engine first determines the top 10 results out of the 50 results as the search recall content based on the correlation between each result and the search input. For example, the click-through rate, like rate, and average browsing time of the retrieved content are weighted and sorted to obtain the top ten.

[0062] Afterwards, at box 304, feature extraction is performed on the search recall content (converted into text information). Feature extraction is to extract text descriptions that can describe the current content from the search recall content, such as general text about the theme of the video. After feature extraction, the computing device 102 obtains a recall content summary 312 for the search recall content 302, which is a content description for the recall content. The computing device 102 also obtains background information from the knowledge base 310. At box 308, the computing device 102 uses an intelligent agent to understand the background of the search term, such as using a large language model 306 to determine background information 314 for the search input. For example, relevant information for this search is retrieved from the knowledge base, and then the relevant information is input into the large model 306 for processing to supplement the background information. For example, when a word has multiple meanings, it can help the large language model determine which layer of meaning to take, such as "rock-hard" is both a common adjective and a movie name.

[0063] Then, at box 316, an agent is used to classify the search terms using the recalled content summary and background information. In this process, the large language model 326 is used to process the recalled content summary and background information to determine the search term type 336. According to the recalled content summary and the search term background information, it is determined which category the search term belongs to, such as pan-entertainment, information knowledge, etc.

[0064] Then, at box 318, an agent is used to classify the recalled content by using a large language model 328 to classify the recalled content, and then calculating according to the weight of the corresponding content to determine the category with the largest weight, thereby determining the recalled content category 338 for the search recalled content. For example, the top ten contents of the search recall are classified separately, and then the category with the largest weight is calculated according to the weight of the corresponding content.

[0065] At box 320, the agent is also used to classify the search input for the degree of accuracy of the requirements. For example, the large language model 330 processes the recalled content summary and background information to determine the accuracy category 340 of the search term. It can determine whether the search is looking for a specific resource. For example, searching for "movie" 》original version", the search intent is clearly to find the complete version of 》movie, rather than clips, behind-the-scenes footage, recent news, etc., which are only related to the movie. At box 322, an agent is used to process the recalled content summary and background information for timeliness classification. For example, the recalled content summary and background information are processed by a large language model 332 to determine the timeliness category 342 of the search term. It is used to determine whether the current search is looking for hot information, such as searching for athletes during a sports meeting, which is a typical example of searching for hot information.

[0066] At box 324, the agent is used to process the recalled content summary and background information to determine the validity of the search term. For example, the recalled content summary and background information are processed by the large language model 334 to determine whether the search is an invalid search at box 344. It is mainly used to determine whether the search term violates the law, whether it is the same as the local mainstream language, etc. The computing device 102 can use one or more of the above classification results to determine the intention for the search term or search input.

[0067] Combined with the above Figure 3 A schematic diagram of an example of a system architecture for determining search input intent according to some disclosed embodiments is described. Figure 4 A schematic diagram describing an example of a flowchart for determining search input intent according to some embodiments of the present disclosure. Figure 4 The process can be found in Figure 1The system may be executed on the computing device 102 as shown, or on any suitable device.

[0068] like Figure 4 As shown, in example 400, first, the user can obtain 414 the search terms and the recalled content from the search data 416 through the search service platform 402. The search data 416 is the search data obtained by the user after searching the cloud records in box 426 and filtering 424 the search results obtained by conditions. The filtering conditions can be time and place, etc. For example, data in a specific area within a predetermined period of time is obtained.

[0069] The search service platform 402 sends 412 an intent recognition request to the effect evaluation platform 404. The intent recognition request includes a search term and a corresponding recall result. The effect evaluation platform 404 generates an intent recognition task 418 according to the intent recognition request.

[0070] The automatic scheduling 428 controls the execution process of the intent recognition task. For example, the automatic scheduling 428 performs scheduling according to the time window. For example, there are many pending tasks in the queue, which are sorted and scheduled in chronological order. When the automatic scheduling 428 starts, all recognition tasks within 3 days will be executed one by one according to time. If the executable tasks do not meet the maximum number of executable tasks, the earlier recognition tasks will be added to the executing tasks to maximize the use of computing resources. The above examples are only used to describe the present disclosure, and are not specific limitations of the present disclosure.

[0071] The automated scheduling 428 puts the task into the computing engine 430. The computing engine 430 represents the task execution logic and records the execution of the task. When executing the intent recognition task, in order to avoid excessive number of requests and thus causing the language model cluster to be overloaded, a distributed current limiter 420 is used to limit the number of requests per minute.

[0072] Subsequently, the intent of the search term is identified through multi-agent intent recognition 422, for example, the system architecture in Example 300 can be used for intent recognition. After the recognition task is completed, in box 408, the identified intent is integrated. During the result integration process, the machine-processed information is mapped to business text based on field mapping. Then, at box 406, the final model result is persisted, for example, the result is stored in a storage device. In this way, the evaluation results can be stored to avoid the need to re-identify the intent the next time it is used, thereby improving the efficiency of intent recognition. Then, the intent recognition result is returned 410 to the effect evaluation platform 404.

[0073] Finally, the effect evaluation platform 404 returns 410 the intent recognition result to the search service platform 402, and the search service platform performs subsequent processing according to the recognized intent.

[0074] Combined with the above Figure 4 A schematic diagram of an example of a flowchart for determining search input intent according to some disclosed embodiments is described. Figure 5 A schematic diagram describing an example of displaying the effect of determining search input intent according to some embodiments of the present disclosure.

[0075] like Figure 5 As shown, in example 500, the search input in search 502 is " Object", and region 504 is displayed as "TH", it means that the user searched in Chinese in Thailand, and you can see that the search time span is 2024-05-05-2024-06-04.

[0076] After the intent recognition task for the search input, the intent recognition result is displayed in box 506. It includes query meaning, query background, possible demands, drop, intensity of demand, query demand type, explicit demand, and event related to timelines.

[0077] It can be seen that in this example 500, for " In the recognition result of the intent of "object", background information, effectiveness, accuracy category, timeliness, etc. are analyzed and displayed in box 506. The recognition of intent can be comprehensively considered based on these categories and the final search results can be displayed to the user.

[0078] It can be seen that the user inputs Chinese in Thailand, and the recognition result for the intent is output in English, so in this case, the recognition result of the search input intent is converted to English. Depending on the region and the settings of the search engine used by the user, the converted target language can be any language, and this application does not limit it here.

[0079] Combined with the above Figure 5 A schematic diagram of an example of displaying the effect of determining the search input intention according to some disclosed embodiments is described. Figure 6 A schematic diagram depicting another example of displaying the effect of determining search input intent according to some embodiments of the present disclosure.

[0080] like Figure 6 As shown, in example 600, the search input in search 602 is "xxx concert", and region 604 is displayed as "CN", indicating that the user is searching in Chinese in China, and it can be seen that the dates of the search time span are 2024-09-01-2024-10-01.

[0081] Box 606 shows that the user's search intention is to find information about xxx's concert, which is entertainment-oriented, and there is a lot of information related to the concert, such as the number of shows, ticket prices, concert dates, etc., so the intention is not accurate. In addition, the concert is time-sensitive and the search input logic is clear and does not contain any sensitive content, so the intention is also valid.

[0082] Combined with the above Figure 6 A schematic diagram of another example of determining the effect display of search input intent according to some disclosed embodiments is described. Figure 7 A schematic diagram describing yet another example of displaying the effect of determining search input intent according to some embodiments of the present disclosure.

[0083] like Figure 7 As shown, in example 700, the search input in search 702 is "full version of "xxx", and region 704 is displayed as "CN", which indicates that the user is searching in Chinese in China, and it can be seen that the dates of the search time span are 2024-09-01-2024-10-01.

[0084] In box 706, it is shown that the user's search intention is to find the full version of the film and television resource "xxx", and is entertainment-oriented. The user's search input indicates that the user is looking for the full version of the film and television resource "xxx", so it can be considered that the accuracy category is accurate.

[0085] In addition, the search for film and television resources is not time-sensitive and the film and television resources that the search input is looking for are generally subject to copyright restrictions, so the intention is invalid.

[0086] Combined with the above Figure 7 A schematic diagram of another example of displaying the effect of determining the search input intention according to some disclosed embodiments is described. Figure 8 A schematic diagram describing yet another example of displaying the effect of determining search input intent according to some embodiments of the present disclosure.

[0087] like Figure 8 As shown, in example 800, the search input in search 802 is " Games", and region 804 is displayed as "US", it means that the user is searching in a foreign country in Chinese, and the search time span is 2024-09-01-2024-10-01

[0088] According to the content in box 806, it can be judged that the user's search intention is to find game-related content, and the content is entertainment-oriented. The user's search input indicates that the user is looking for game-related content, and there are many related events, so the intention is inaccurate.

[0089] In addition, the search for games was conducted in 2024 and is therefore timely, and the search input factually identified games that were in the game category and therefore the intent was valid.

[0090] Combined with the above Figure 8 A schematic diagram of another example of displaying the effect of determining the search input intention according to some disclosed embodiments is described. Fig. 9 A schematic diagram depicting yet another example of displaying the effect of determining search input intent according to some embodiments of the present disclosure.

[0091] like Fig. 9 As shown, the device 900 includes a content description and background information determination module 902, which is configured to determine the content description of a set of search results for a target search input and the background information for the target search input; a category determination module 904, which is configured to determine the category of the target search input based on the content description and the background information; and an intention determination module 906, which is configured to determine the intention for the target search input based on the category of the search input.

[0092] In some embodiments, the content description and background information determination module 902 includes: a reference search result determination module, configured to obtain a set of reference search results associated with the target search input from a target knowledge base based on the target search input; and a background information determination module, configured to determine the background information of the target search input based on the set of reference search results.

[0093] In some embodiments, the context information determination module includes: a first language model application module configured to determine context information of a target search input by applying a set of reference search results and the target search input to a first language model.

[0094] In some embodiments, the content description and background information determination module 902 also includes: a group of search result determination modules, configured to determine a group of search results corresponding to the target search input; a text description determination module, configured to determine a text description corresponding to the search results in the group of search results; and a content description determination module, configured to determine a content description of a group of search results based on the text description.

[0095] In some embodiments, a set of search result determination modules includes: multiple search result acquisition modules, configured to acquire multiple search results corresponding to the target search input through a search engine; and a set of search result selection modules, configured to select a set of search results from the multiple search results based on a set of parameters associated with each search result in the multiple search results.

[0096] In some embodiments, one of the set of parameters includes at least one of the following: click-through rate, like rate, or average browsing time of search results.

[0097] In some embodiments, the content description and context information determination module 902 further includes: a target language conversion module configured to convert a set of search results and a set of reference search results into a target language.

[0098] In some embodiments, the category determination module 904 includes: a second language model application module configured to determine the category of the target search input by applying the content description and the context information to the second language model.

[0099] In some embodiments, the category is a first category, the first category comes from a first group of categories determined by classifying the intent of the search input, and the intent determination module 906 includes: a result category determination module, configured to determine a result category for each search result in a set of search results based on the content description, the result category comes from a second group of categories determined by classifying the content of the search results; a second category determination module, configured to determine a second category for a set of search results based on the result category and the weight corresponding to the result category, the second category comes from the second group of categories; and an intent determination module, configured to determine the intent for the target search input based on the first category and the second category.

[0100] In some embodiments, the intent determination module includes: a precision category determination module, configured to determine the precision category of the intent based on background information and content description, the precision category being used to indicate whether the search input is for a predetermined resource; and an intent determination module, configured to determine the intent for the target search input based on the first category, the second category and the precision category.

[0101] In some embodiments, the accuracy level categories include an accurate category and an inaccurate category.

[0102] In some embodiments, the intent determination module includes: a timeliness category determination module, configured to determine the timeliness category of the target search input based on background information and content description, the timeliness category being used to indicate the degree of association between the search input and the hot content; and an intent determination module, configured to determine the intent for the target search input based on the first category, the second category, the precision category and the timeliness category.

[0103] In some embodiments, the timeliness categories include time-sensitive and non-time-sensitive.

[0104] In some embodiments, the intent determination module includes: a validity determination module configured to determine the validity of the intent based on background information and content description; and an intent determination module configured to determine the intent for the target search input based on the first category, the second category, the accuracy category, the timeliness category and the validity.

[0105] In some embodiments, validity includes valid and invalid.

[0106] Fig.10 A schematic block diagram of an example device 1000 that may be used to implement embodiments of the present disclosure is shown. Figure 1 The computing device 102 in the example device 1000 can be implemented using the example device 1000. As shown in the figure, the example device 1000 includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 1002 or computer program instructions loaded from a storage unit 1008 to a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for device operation can also be stored. The CPU 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0107] Multiple components in the example device 1000 are connected to the I / O interface 1005, including: an input unit 1006, such as a keyboard, a mouse, etc.; an output unit 1007, such as various types of displays, speakers, etc.; a storage unit 1008, such as a disk, an optical disk, etc.; and a communication unit 1009, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1009 allows the example device 1000 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0108] The various processes and processing described above, such as the example method 200 and examples 400 to 800, may be performed by the processing unit 1001. For example, in some embodiments, the example method 200 and examples 400 to 800 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as the storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on the example device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into the RAM 1003 and executed by the CPU 1001, one or more actions of the example method 200 and examples 400 to 800 described above may be performed.

[0109] The present disclosure may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present disclosure.

[0110] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive lists) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer-readable storage medium used herein is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (e.g., a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0111] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0112] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0113] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0114] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0115] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0116] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0117] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for determining the intent of a search input, comprising: Determine a content description of a set of search results for a target search input and background information for the target search input, wherein the background information is determined based on a set of reference search results obtained by searching a knowledge base using the target search input; Determining a category of the target search input based on the content description and the background information; as well as Determining the intent for the target search input based on the category of the search input; Wherein the category is a first category, the first category is from a first group of categories determined by classifying the intent of the search input, and based on the category of the search input, determining the intent for the target search input comprises: Based on the content description, determining a result category for each search result in the set of search results, the result category being from a second set of categories determined by classifying the content of the search results; determining a score for each result category based on the number of search results included in each result category in the second group of categories and a weight corresponding to the result category; selecting a result category having a maximum score from the second set of categories as a second category of the set of search results, the second category being from the second set of categories; and Based on the first category and the second category, the intent for the targeted search input is determined.

2. The method of claim 1, wherein determining a content description of a set of search results for a target search input and context information for the target search input comprises: Based on the target search input, obtaining a set of reference search results associated with the target search input from a target knowledge base; as well as Based on the set of reference search results, context information of the target search input is determined.

3. The method of claim 2, wherein determining the context information of the target search input based on the set of reference search results comprises: The context information of the target search input is determined by applying the set of reference search results and the target search input to a first language model.

4. The method according to claim 2, wherein determining a content description of a set of search results for a target search input and background information for the target search input further comprises: determining a set of search results corresponding to the target search input; determining a text description corresponding to a search result in the set of search results; as well as Based on the text description, a content description of the set of search results is determined.

5. The method of claim 4, wherein determining a set of search results corresponding to the target search input comprises: Acquire multiple search results corresponding to the target search input through a search engine; as well as The set of search results is selected from the plurality of search results based on a set of parameters associated with each search result in the plurality of search results.

6. The method of claim 5, wherein the set of parameters comprises at least one of the following: The click-through rate, like rate or average browsing time of the search results.

7. The method according to claim 2, further comprising: The set of search results and the set of reference search results are converted into a target language.

8. The method according to claim 1, wherein determining the category of the target search input based on the content description and the background information comprises: The category of the target search input is determined by applying the content description and the context information to a second language model.

9. The method of claim 1, wherein based on the first category and the second category, determining the intent for the target search input comprises: Determining a precision category of the intent based on the context information and the content description, the precision category being used to indicate whether the search input is for a predetermined resource; as well as The intent for the target search input is determined based on the first category, the second category, and the precision category.

10. The method according to claim 9, wherein the accuracy level categories include an accurate category and an inaccurate category.

11. The method of claim 9, wherein determining the intent for the target search input based on the first category, the second category, and the precision category comprises: Determining a timeliness category for the target search input based on the background information and the content description, wherein the timeliness category is used to indicate a degree of association between the search input and hot content; as well as The intent for the target search input is determined based on the first category, the second category, the precision category, and the timeliness category. The method according to claim 11 , wherein the timeliness category includes time-sensitive and non-time-sensitive.

13. The method of claim 11, wherein determining the intent for the target search input based on the first category, the second category, the precision category, and the timeliness category comprises: Determining validity of the search input based on the background information and the content description; as well as The intent for the target search input is determined based on the first category, the second category, the precision category, the timeliness category, and the effectiveness. The method according to claim 13 , wherein the validity includes valid and invalid.

15. A device for determining search input intention, comprising: a content description and background information determination module, configured to determine a content description of a set of search results for a target search input and background information for the target search input, wherein the background information is determined based on a set of reference search results obtained by searching a knowledge base using the target search input; a category determination module, configured to determine the category of the target search input based on the content description and the background information; as well as an intention determination module, configured to determine the intention for the target search input based on the category of the search input; The category is a first category, the first category is from a first group of categories determined by classifying the intent of the search input, and the intent determination module includes: A result category determination module configured to determine a result category for each search result in the set of search results based on the content description, wherein the result category comes from a second set of categories determined by classifying the content of the search results; a score determination module configured to determine a score for each result category based on the number of search results contained in each result category in the second group of categories and a weight corresponding to the result category; A second category determination module configured to select a result category with a maximum score from the second group of categories as a second category of the set of search results, the second category being from the second group of categories; and An intent module is configured to determine the intent for the target search input based on the first category and the second category.

16. An electronic device, comprising: at least one processor; as well as A storage device, used to store at least one program, when the at least one program is executed by the at least one processor, so that the at least one processor implements the method according to any one of claims 1-14.

17. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 14 when executed by a processor.

18. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 14.

Images