A method for recommending entity words, related devices, equipment and storage medium

Abstract

The present application discloses a method for recommending entity words based on artificial intelligence technology, comprising: displaying a first page, wherein the first page provides a page jump control; in response to a first trigger operation of the target user on the page jump control, displaying a second page, and highlighting at least one recommended entity word on the second page, wherein the at least one recommended entity word is determined based on the user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data. The present application also provides related devices, equipment, and storage media. Since the user characteristic data of different users and the set of entity words to be recommended are different and diverse, the present application can achieve a "one-size-fits-all" entity word recommendation effect, thus meeting the demand for personalized recommendation of entity words.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method for recommending entity words, related devices, equipment, and storage media. Background Art

[0002] With the development of internet technology, information is exploding, and users receive a massive amount of information every day. Because each user subscribes to different information sources, the information lists they ultimately see are also completely different. Therefore, in the context of information recommendation, how to process and evaluate this massive amount of information has become a worthy research topic.

[0003] Currently, entity terms are highlighted on the information recommendation page. For example, if you search for the keyword "Shenzhen City" on the Baidu Encyclopedia, the information recommendation page will not only display the entry description but also highlight the entity terms, such as "Huizhou City." Users who click on a highlighted entity term will be redirected to the entity term's detailed information page.

[0004] However, existing entity word recommendations are fixed, resulting in a "same face for thousands of people" effect. That is, for all users, the entity words they see on the information recommendation page are the same, which makes it difficult to meet users' personalized needs, resulting in significant limitations in entity word recommendations. Summary of the Invention

[0005] The embodiments of the present application provide a method for recommending entity words, related devices, equipment and storage media. Since the user feature data of different users and the set of entity words to be recommended are different and diverse, it is possible to achieve a "one-size-fits-all" entity word recommendation effect and meet the needs of personalized entity word recommendation.

[0006] In view of this, the present application provides, on one hand, a method for recommending entity words, comprising:

[0007] Displaying a first page, wherein the first page provides a page jump control;

[0008] In response to a first trigger operation of a page jump control by a target user, displaying a second page, and highlighting at least one recommended entity word on the second page, wherein the at least one recommended entity word is determined based on user feature data of the target user, and the user feature data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data;

[0009] In response to the page display instruction, at least one recommended entity word is highlighted on the second page, wherein the at least one recommended entity word is determined based on user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data.

[0010] Another aspect of the present application provides an entity word recommendation device, comprising:

[0011] A display module, configured to display a first page, wherein the first page provides a page jump control;

[0012] The display module is further configured to, in response to a first trigger operation of the target user on the page jump control, display a second page, and highlight at least one recommended entity word on the second page, wherein the at least one recommended entity word is determined based on user feature data of the target user, and the user feature data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data;

[0013] In one possible design, in another implementation of another aspect of the embodiment of the present application, the entity word recommendation device further includes an acquisition module and a recommendation module;

[0014] An acquisition module is used to obtain user feature data of target users;

[0015] The acquisition module is further configured to acquire a target user's to-be-recommended entity word set, wherein the recommended entity word set includes N entity word sets, each entity word set includes at least one entity word, and N is an integer greater than or equal to 1;

[0016] The acquisition module is further used to obtain the matching score corresponding to each entity word in each entity word set through the fusion recommendation model based on the user feature data and the entity word set to be recommended;

[0017] The recommendation module is used to select at least one recommended entity word from the entity word set to be recommended according to the matching score corresponding to each entity word in each entity word set, wherein at least one recommended entity word is a word highlighted on the page.

[0018] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0019] User behavior data includes at least one of browsing data, click data, copy data, and stay data;

[0020] User profile data includes at least one of age data, gender data, preference data, and region data;

[0021] User relationship data includes at least one of friend data, follow data, like data, and favorite data;

[0022] The user interest data includes at least one interest tag.

[0023] In one possible design, in another implementation of another aspect of the embodiment of the present application, the entity word recommendation device further includes a determination module;

[0024] A determination module, configured to determine a first portrait score corresponding to the target user based on the user behavior data of the target user and the user portrait data of the target user;

[0025] The determination module is further configured to determine an associated user based on the first portrait score, wherein the associated user corresponds to a second portrait score, and an absolute value of a difference between the second portrait score and the first portrait score is less than or equal to a score threshold;

[0026] An acquisition module is specifically used to acquire a target entity word set of a target user and a related entity word set of related users;

[0027] Determining a first similarity between the target user and the associated users based on the target entity word set and the associated entity word set;

[0028] determining a second similarity between the target user and the associated user based on the user relationship data of the target user and the user relationship data of the associated user;

[0029] If the sum of the first similarity and the second similarity is greater than or equal to the similarity threshold, the target entity word set and the associated entity word set are taken as the first entity word set in the entity word set to be recommended.

[0030] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0031] An acquisition module is specifically used to determine the number of intersections according to the target entity word set and the associated entity word set, and to determine the number of unions according to the target entity word set and the associated entity word set;

[0032] The ratio of the number of intersections to the number of unions is used as the first similarity between the target user and the associated users;

[0033] An acquisition module, specifically configured to determine friend similarity based on the target user's friend data and a first weight value;

[0034] Determining attention similarity based on the attention data of the target user, the attention data of the associated users, and the second weight value;

[0035] Determine the likes similarity based on the likes data of the target user, the likes data of the associated users, and the third weight value;

[0036] Determining the collection similarity based on the collection data of the target user, the collection data of the associated users, and the fourth weight value;

[0037] The second similarity between the target user and the associated users is determined based on the friend similarity, the follow similarity, the like similarity, and the favorite similarity.

[0038] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0039] The acquisition module is specifically used to obtain the search keywords of the target user;

[0040] Determine related texts based on search keywords;

[0041] At least one entity word that hits the hot word dictionary is obtained from the associated text, and the at least one entity word is used as the second entity word set in the entity word set to be recommended, wherein the hot word dictionary includes at least one hot word whose popularity value is greater than or equal to a popularity threshold within a preset time period.

[0042] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0043] The acquisition module is specifically used to obtain the search keywords of the target user;

[0044] Determine related texts based on search keywords;

[0045] At least one entity word that hits the uncommon word dictionary is obtained from the associated text, and the at least one entity word is used as a third entity word set in the set of entity words to be recommended, wherein the uncommon word dictionary includes at least one uncommon word whose occurrence frequency within a preset time period is less than or equal to a first frequency threshold.

[0046] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0047] The acquisition module is specifically used to obtain the search keywords of the target user;

[0048] Determine related texts based on search keywords;

[0049] At least one entity word that hits a popular word dictionary is obtained from the associated text, and the at least one entity word is used as a fourth entity word set in the set of entity words to be recommended, wherein the popular word dictionary includes at least one popular word whose appearance frequency within a preset time period is greater than or equal to a second frequency threshold.

[0050] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0051] The acquisition module is specifically used to obtain the search keywords of the target user;

[0052] Determine related texts based on search keywords;

[0053] If the entity word in the associated text hits the audio-visual word dictionary, then obtaining target interaction data of the entity word according to the audio-visual word dictionary, wherein the audio-visual word dictionary includes a mapping relationship between the audio-visual word and the interaction data;

[0054] If there is target interaction data of at least one entity word that meets the recommendation condition, the at least one entity word is used as the fifth entity word set in the set of entity words to be recommended.

[0055] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0056] The acquisition module is specifically used to obtain the search keywords of the target user;

[0057] Determine related texts based on search keywords;

[0058] At least one entity word that hits the alias dictionary is obtained from the associated text, and the at least one entity word and the entity words that have the same pointing object as each entity word are taken as the sixth entity word set in the set of entity words to be recommended, wherein the alias dictionary includes at least one group of entity words that have the same pointing object.

[0059] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0060] An acquisition module, specifically configured to acquire a first set of derived tags having category relevance to the user's interest data, and acquire a second set of derived tags having semantic relevance to the interest tags;

[0061] The user interest data, the first derived tag set, and the second derived tag set are used as the seventh entity word set in the entity word set to be recommended.

[0062] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0063] An acquisition module is specifically used to obtain, for each entity word set in the entity word set to be recommended, a matching score corresponding to each entity word in the entity word set by integrating the recommendation sub-model included in the recommendation model based on the user feature data and the entity word set;

[0064] The recommendation module is specifically configured to determine the comprehensive matching scores of M entity words according to the matching scores corresponding to each entity word in each entity word set, where M is an integer greater than 1;

[0065] The top K entity words with the largest comprehensive matching scores are used as at least one recommended entity word, where K is an integer greater than or equal to 1 and less than or equal to M.

[0066] In one possible design, in another implementation of another aspect of the embodiment of the present application,

[0067] The display module is also used to respond to the target user's second trigger operation on the target recommended entity word, display a third page, and display page information related to the target recommended entity word on the third page, wherein the target recommended entity word is any one of the at least one recommended entity word.

[0068] Another aspect of the present application provides a computer device, comprising: a memory, a processor, and a bus system;

[0069] Wherein, the memory is used to store programs;

[0070] The processor is used to execute the program in the memory, and the processor is used to perform the above-mentioned methods according to the instructions in the program code;

[0071] The bus system is used to connect the memory and the processor so that the memory and the processor can communicate with each other.

[0072] Another aspect of the present application provides a computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium. When the computer-readable storage medium is run on a computer, the computer is enabled to execute the above-mentioned methods.

[0073] Another aspect of the present application provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the above aspects.

[0074] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0075] In an embodiment of the present application, a method for recommending entity words is provided, which first displays a first page, and then generates a page display instruction in response to a selection operation on a page jump control. The page display instruction carries the user identifier of the target user, and thus, in response to the page display instruction, at least one recommended entity word is highlighted on the second page. The at least one recommended entity word is determined based on the user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data. In the above manner, based on the user's own characteristic data, entity words with a higher degree of matching are selected as recommended entity words, and are highlighted on the page. Since the user characteristic data of different users and the set of entity words to be recommended are different and diverse, it is possible to achieve an entity word recommendation effect of "one thousand faces for one thousand people", meeting the demand for personalized recommendation of entity words. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a schematic diagram of an application of recommending entity words based on web page content in an embodiment of the present application;

[0077] Figure 2 This is a schematic diagram of an application of recommending entity words based on video content in an embodiment of the present application;

[0078] Figure 3 This is a schematic diagram of an environment of an entity word recommendation system in an embodiment of the present application;

[0079] Figure 4 This is a schematic diagram of the architecture of the entity word recommendation system in an embodiment of the present application;

[0080] Figure 5 A flowchart of the entity word recommendation method in an embodiment of the present application;

[0081] Figure 6 A schematic diagram of a process for determining associated text based on search keywords in an embodiment of the present application;

[0082] Figure 7 A schematic diagram of a process for constructing a hot word dictionary in an embodiment of the present application;

[0083] Figure 8 This is a schematic diagram of prediction based on the fusion recommendation model in an embodiment of the present application;

[0084] Figure 9 This is a schematic diagram of training based on a fusion recommendation model in an embodiment of the present application;

[0085] Figure 10 A schematic diagram of an interface showing recommended entity words in an embodiment of the present application;

[0086] Figure 11 A schematic diagram of an entity word recommendation device in an embodiment of the present application;

[0087] Figure 12 A schematic diagram of the structure of the server in the embodiment of the present application;

[0088] Figure 13 This is a structural diagram of a terminal device in an embodiment of the present application. DETAILED DESCRIPTION

[0089] The embodiments of the present application provide a method for recommending entity words, related devices, equipment and storage media. Since the user feature data of different users and the set of entity words to be recommended are different and diverse, it is possible to achieve a "one-size-fits-all" entity word recommendation effect and meet the needs of personalized entity word recommendation.

[0090] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the numbers used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "corresponding to" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0091] With the continuous explosion of information on the internet, the depth, breadth, and frequency of information access have greatly increased. The development of cloud technology has not only significantly increased computing power, but also enabled the analysis of large amounts of user data based on big data. Applying cloud computing and big data technologies to analyze individual users creates intelligent recommendation experiences. Cloud computing is the product of the convergence of traditional computer and network technologies, including grid computing, distributed computing, parallel computing, utility computing, network storage technologies, virtualization, and load balancing. Driven by the growth of the internet, real-time data streams, and the diversification of connected devices, as well as the demands of search services, social networks, mobile commerce, and open collaboration, cloud computing has rapidly developed. Unlike previous parallel and distributed computing, the emergence of cloud computing will fundamentally revolutionize the entire internet model and enterprise management model.

[0092] Big data refers to collections of data that cannot be captured, managed, and processed within a specific timeframe using conventional software tools. These massive, rapidly growing, and diverse information assets require new processing models to unlock greater decision-making power, insight discovery, and process optimization. With the advent of the cloud era, big data has attracted increasing attention. Big data requires specialized technologies to efficiently process large amounts of time-sensitive data. Technologies suitable for big data include massively parallel processing databases, data mining, distributed file systems, distributed databases, cloud computing platforms, the internet, and scalable storage systems.

[0093] Based on this, different scenarios are involved in user personalized recommendations. The following will introduce the entity word recommendation method provided by this application in combination with two specific scenarios.

[0094] Scenario 1: Entity word recommendation based on web pages;

[0095] For ease of understanding, please refer to Figure 1 , Figure 1 This is a schematic diagram of an application that recommends entity words based on webpage content in an embodiment of the present application. As shown in the figure, the entity words "information overload" and "recommendation system" are highlighted on the webpage. If the user selects the entity word "recommendation system", an introduction to terms related to "recommendation system" can be displayed, or the user can jump to a page related to "recommendation system".

[0096] Scenario 2: Entity word recommendation based on video;

[0097] For ease of understanding, please refer to Figure 2 , Figure 2 This is a schematic diagram of an application that recommends entity words based on video content in an embodiment of this application. As shown in the figure, the entity words "ice cream," "pasta," and "motorcycle" are highlighted in the video application. If the user selects the entity word "ice cream," other videos or entry descriptions related to "ice cream" are displayed.

[0098] In order to achieve personalized recommendation of entity words in the above scenario, this application proposes an entity word recommendation system, which is applied to Figure 3 The entity word recommendation system shown in the figure includes a server and a terminal device, and the client is deployed on the terminal device, wherein the client can be run on the terminal device in the form of a browser, or can be run on the terminal device in the form of an independent application (APP), etc. The specific presentation form of the client is not limited here. The server involved in this application can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms. The terminal device can be a smart phone, tablet computer, laptop computer, PDA, personal computer, smart TV, smart watch, car-mounted device, wearable device, etc., but is not limited to this. The terminal device and the server can be directly or indirectly connected by wired or wireless communication, and this application does not limit this. The number of servers and terminal devices is also not limited. The solution provided in this application can be completed independently by the terminal device, can be completed independently by the server, or can be completed by the terminal device and the server in cooperation. This application does not make any specific limitations on this.

[0099] As disclosed in the present application, in the entity word recommendation or device, multiple servers can be combined into a blockchain, and the servers are nodes on the blockchain.

[0100] by Figure 3For example, different users may use different terminal devices, and the server provides entity word recommendation services for different users through the terminal devices. For example, user A uses terminal device A, and user B uses terminal device B. The server pushes several entity words (for example, milk tea, ice cream, and milk) to user A through terminal device A, and pushes several entity words (for example, playground and electronic products) to user B through terminal device B. As a result, several entity words provided for user A are displayed on the page displayed by terminal device A, and several entity words provided for user B are displayed on the page displayed by terminal device B. Similarly, the server can also push personalized entity words to user C and user D through terminal device C and terminal device D, respectively.

[0101] Based on the above introduction, please refer to Figure 4 , Figure 4 This is an architectural diagram of the entity word recommendation system in the embodiment of the present application. As shown in the figure, the server can process the information flow in real time, that is, perform word segmentation, named entity recognition, entity filtering, entity extraction and entity classification on the information flow. An entity dictionary can be constructed based on the real-time processed information flow, wherein the entity dictionary includes but is not limited to a hot word dictionary, a rare word dictionary, a popular word dictionary, an audio-visual word dictionary and an alias dictionary. Before recommending entity words to a user, it is also necessary to extract the user's user behavior data, user portrait data, user relationship data and user interest data. Combined with the user's feature data and the entity dictionary, the entity word set related to the user is screened from different dimensions (collaborative filtering, current hot spots, user interests and knowledge recommendations, etc.). Finally, the user's feature data and the entity word set screened out under each dimension are input into the fusion recommendation model together, and the matching score is output. Thus, the top K entity words with the largest matching score are recommended to the user.

[0102] Understandably, identifying information flows primarily involves natural language processing (NLP) technology, a key area of ​​research in computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. Natural language processing (NLP) integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language we use in everyday life—and is closely linked to the study of linguistics. NLP technologies typically include text processing, semantic understanding, machine translation, robotic question-answering, and knowledge graphs.

[0103] Furthermore, information extraction is the process of extracting unstructured information embedded in text and converting it into structured data. It extracts the relationship between named entities from the corpus composed of natural language and is a deeper level of research based on named entity recognition.

[0104] The training and prediction of fusion recommendation models primarily involve machine learning (ML) technology. ML is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0105] Both natural language processing and machine learning fall under the umbrella of artificial intelligence (AI). AI refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive field of computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI is the study of the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making.

[0106] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0107] Combined with the above introduction, the following will introduce the recommended method for entity words in this application. Please refer to Figure 5 In one embodiment of the entity word recommendation method of the present application, the following steps are included:

[0108] 110. Displaying a first page, wherein the first page provides a page jump control;

[0109] In one or more embodiments, the entity word recommendation device displays a first page. The first page can be understood as a first-level page provided by an application (e.g., a browser application or a video application, etc.). Taking a browser application as an example, a user can enter the first page (e.g., a search engine page) after entering the web address on the navigation bar. Based on this, in one implementation, the user can enter a search term on the first page and click on the page jump control. In another implementation, the user can select a certain entry, picture, or title on the first page, thereby triggering the page jump control.

[0110] It should be noted that the entity word recommendation device can be deployed on a terminal device, or on a server, or on an entity word recommendation system composed of a server and a terminal device, which is not limited here.

[0111] 120. In response to a first trigger operation of a page jump control by a target user, a second page is displayed, and at least one recommended entity word is highlighted on the second page, wherein the at least one recommended entity word is determined based on user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data.

[0112] In one or more embodiments, the entity word recommendation device responds to a first trigger operation for a page jump control, displays a second page, and highlights at least one recommended entity word on the second page. It is understandable that a page display instruction can be generated by the first trigger operation, wherein the page display instruction carries the user identifier of the target user. Taking a browser application as an example, in one implementation, after the target user logs in to the browser application, it can be associated with the user identifier of the target user (for example, a user account, a user email address, or a user mobile phone number, etc.).

[0113] In response to a first trigger operation, the entity word recommendation device extracts user feature data of the target user based on the target user's user identifier, where the target user's user feature data includes at least one of user behavior data, user profile data, user relationship data, and user interest data. Based on this, at least one recommended entity word for the target user is determined based on the target user's user feature data, thereby achieving the purpose of personalized entity word recommendation.

[0114] It can be understood that the second page can be understood as a secondary page, that is, after clicking a link in the primary page, you can enter the secondary page.

[0115] In an embodiment of the present application, a method for recommending entity words is provided. Based on the user's own characteristic data, entity words with a high degree of matching are selected as recommended entity words, and these entity words are highlighted on the page. Because different users have different characteristic data and a set of entity words to be recommended, this method can achieve a personalized entity word recommendation effect, meeting the demand for personalized entity word recommendation.

[0116] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, another optional embodiment provided by the embodiment of the present application may further include:

[0117] Obtain user characteristic data of target users;

[0118] Obtain a target user's to-be-recommended entity word set, wherein the recommended entity word set includes N entity word sets, each entity word set includes at least one entity word, and N is an integer greater than or equal to 1;

[0119] Based on user feature data and the set of entity words to be recommended, the matching score corresponding to each entity word in each entity word set is obtained through the fusion recommendation model;

[0120] At least one recommended entity word is selected from the set of entity words to be recommended according to the matching score corresponding to each entity word in each set of entity words.

[0121] In one or more embodiments, the entity word recommendation device obtains user feature data of the target user. Specifically, each user corresponds to a user ID. Therefore, for the target user, the user feature data of the target user can be found through the user ID of the target user. The user ID includes but is not limited to the account number, mobile phone number, email address and registration number. The entity word recommendation device obtains the target user's entity word set to be recommended. The recommended entity word set includes N entity word sets, each entity word set belongs to a type. For example, one entity word set belongs to the "hot spot" type, and the other entity word set belongs to the "buzzword" type.

[0122] The entity word recommendation device can perform feature processing on user feature data and the set of entity words to be recommended based on feature engineering, and the user feature data after feature processing and the set of entity words to be recommended are input into the fusion recommendation model, and the fusion recommendation model outputs the matching score corresponding to each entity word in each entity word set. Assume that the fusion recommendation model includes a first recommendation sub-model and a second recommendation sub-model, and assume that the set of entity words to be recommended includes two groups of entity word sets, namely entity word set A and entity word set B, wherein entity word set A includes 5 entity words and entity word set B includes 8 entity words. Based on this, the entity word set A and the user feature data are input into the first recommendation sub-model, thereby outputting the matching scores corresponding to the 5 entity words in the entity word set A. Similarly, the entity word set B and the user feature data are input into the second recommendation sub-model, thereby outputting the matching scores corresponding to the 8 entity words in the entity word set B.

[0123] The entity word recommendation device sorts all entity words according to the matching score corresponding to each entity word in each entity word set, in descending order of matching scores, and finally selects the K entity words as recommended entity words. Wherein, K is an integer greater than or equal to 1, so at least one recommended entity word can be obtained. Based on this, the recommended entity words can be highlighted on the page, for example, displayed in a highlighted form, or displayed in a bold form, or displayed in an italic form, etc., which are not limited here.

[0124] Secondly, in an embodiment of the present application, a method for determining recommended entity words is provided. Through the above-described method, based on the user's own characteristic data, a matching score between the user and each entity word in the set of entity words to be recommended is determined. Thus, entity words with higher matching scores are selected as recommended entity words, and are highlighted on the page. Because different users have different and diverse user characteristic data and the set of entity words to be recommended, this method can achieve a personalized recommendation effect, meeting the demand for personalized entity word recommendations.

[0125] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, in another optional embodiment provided by the embodiment of the present application, the user behavior data includes at least one of browsing data, click data, copy data, and stay data;

[0126] User profile data includes at least one of age data, gender data, preference data, and region data;

[0127] User relationship data includes at least one of friend data, follow data, like data, and favorite data;

[0128] The user interest data includes at least one interest tag.

[0129] In one or more embodiments, a method for obtaining user feature data is described. As can be seen from the aforementioned embodiments, user feature data can be extracted based on big data technology. User feature data includes, but is not limited to, user behavior data, user profile data, user relationship data, and user interest data. Each type of data is described in detail below.

[0130] 1. User behavior data;

[0131] Specifically, user behavior data includes but is not limited to browsing data, click data, copy data, and stay data.

[0132] For example, the target user's browsing data represents the target user's browsing history of pages within a preset time. For example, if the target user browses web pages related to "food" 90 times within 1 month, the target user's browsing data includes "food 90 times".

[0133] For example, the target user's click data represents the target user's history of clicking pages within a preset time. For example, if the target user clicks on a web page related to "food" 10 times in one month, the target user's click data includes "food 10 times".

[0134] For example, the target user's copy data represents the target user's history of copying information within a preset time. For example, if the user copies text related to "food" 20 times in one month, the target user's copy data includes "food 20 times".

[0135] For example, the target user's stay data represents the target user's history of staying on a page within a preset time. For example, if the target user stayed on a page related to "food" 15 times in one month, where "stay" means that the duration of stay on the same page exceeds a threshold (for example, 10 minutes), then the target user's stay data includes "food 15 times."

[0136] 2. User portrait data;

[0137] Specifically, user portrait data includes but is not limited to age data, gender data, preference data, and regional data.

[0138] For example, the target user's age data indicates the target user's age. For example, if the target user is 23 years old, the target user's age data is "23".

[0139] Exemplarily, the gender data of the target user indicates the gender of the target user. For example, if the target user is male, the gender data of the target user is “male”.

[0140] For example, the target user's preference data represents the target user's preferences. For example, if the target user likes comments and articles about "food", the target user's preference data is "food".

[0141] For example, the region data of the target user represents the region where the target user is located. For example, if the region where the target user is located is "Shenzhen", the region data of the target user is "Shenzhen".

[0142] 3. User relationship data:

[0143] Specifically, user relationship data includes but is not limited to friend data, follow data, like data, and favorite data.

[0144] For example, the target user's friend data represents other users who have a friend relationship with the target user. For example, the target user's friends are "User A", "User D" and "User E", that is, the target user's friend data includes "User A", "User D" and "User E".

[0145] For example, the target user's attention data represents other users that the target user has followed. For example, the target user follows "User A", "User D" and "User E", that is, the target user's attention data includes "User A", "User D" and "User E".

[0146] For example, the target user's likes data represents other users whose comments, images, or videos the target user has liked. For example, if the target user has liked posts by "User A," "User D," and "User E," the target user's likes data includes "User A," "User D," and "User E."

[0147] For example, the target user's favorite data represents other users whose images, texts, or videos have been favorited by the target user. For example, if the target user has favorited images, texts, or videos for "User A," "User D," and "User E," the target user's favorite data includes "User A," "User D," and "User E."

[0148] 4. User interest data;

[0149] Specifically, the user interest data includes at least one interest tag.

[0150] For example, interest tags include, but are not limited to, food, entertainment, sports, film and television, finance, technology, cars, games, and the National Basketball Association (NBA). The target user can select tags of interest when registering an account, or at least one interest tag specific to the target user can be summarized based on the content the target user searches for.

[0151] Secondly, in an embodiment of the present application, a method for obtaining user feature data is provided. Through the above method, user behavior data, user portrait data, user relationship data and user interest data can be extracted, and these data also include various feature data related to the user, thereby increasing the diversity and richness of user feature data, thereby more accurately describing the user's characteristics.

[0152] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, another optional embodiment provided by the embodiment of the present application may further include:

[0153] Determine a first profile score corresponding to the target user based on the user behavior data and the user profile data of the target user;

[0154] Determining an associated user based on the first portrait score, wherein the associated user corresponds to a second portrait score, and an absolute value of a difference between the second portrait score and the first portrait score is less than or equal to a score threshold;

[0155] Get the target user's entity word set to be recommended, which can include:

[0156] Obtain the target entity word set of the target user and the associated entity word set of the associated users;

[0157] Determining a first similarity between the target user and the associated users based on the target entity word set and the associated entity word set;

[0158] determining a second similarity between the target user and the associated user based on the user relationship data of the target user and the user relationship data of the associated user;

[0159] If the sum of the first similarity and the second similarity is greater than or equal to the similarity threshold, the target entity word set and the associated entity word set are taken as the first entity word set in the entity word set to be recommended.

[0160] In one or more embodiments, a method for obtaining a first entity word set based on user profile data is introduced. As can be seen from the above embodiments, before calculating the similarity between two users, it is necessary to select at least one associated user related to the target user based on the profile score.

[0161] Specifically, the user's portrait score can be calculated using the following formula:

[0162]

[0163] Among them, profile(user) represents the user's profile score. Tag represents the user profile data, S() represents the function, and weight represents the weight value. Actioni Indicates the i-th behavior in the user behavior data (for example, at least one of the i-th browsing data, the i-th click data, the i-th copy data, and the i-th stay data). j Represents the jth tag in the user portrait data (e.g., age tag, gender tag, preference tag, and region tag). W1 and W2 represent weight values ​​respectively. i Indicates the time when the i-th behavior occurs.

[0164] Based on formula (1), the portrait scores of different users can be calculated. Taking the target user as an example, the first portrait score is calculated. Similarly, the corresponding portrait scores of other users are calculated. Assume that the first portrait score of the target user is 88, and the preset score threshold is 2. At this time, if the portrait score of user B is 89, the absolute value of the difference between the two portrait scores is 1. Therefore, user B is the associated user of the target user, that is, the portrait score of 89 is the second portrait score.

[0165] After determining the target user and associated users, the target entity word set for the target user and the associated entity word set for the associated users are obtained. Then, based on the degree of overlap between the target entity word set and the associated entity word set, the first similarity between the target user and the associated users is calculated. Furthermore, it is necessary to consider whether there are other connections between the target user and the associated users. For example, based on user relationship data, it is determined whether the users are friends, whether they like each other's posts, and whether they collect or forward each other's posts. The more frequent the interactions between users, the more likely their preferences or interests are similar. Therefore, based on these interactions, the second similarity between the target user and the associated users can be calculated.

[0166] The first similarity and the second similarity are added to obtain a comprehensive similarity. Then, it can be further determined whether the comprehensive similarity is greater than or equal to a similarity threshold. If so, the target entity word set of the target user and the associated entity word set of the associated users are taken together as the first entity word set of the target user.

[0167] Again, in an embodiment of the present application, a method for obtaining a first entity word set based on user portrait data is provided. Through the above method, based on the user behavior data of the target user, associated users with a high portrait similarity to the target user can be found, and then the similarities are calculated separately according to the target entity word set of the target user and the associated entity word set of the associated users, and the user relationship data of the target user and the user relationship data of the associated users. The two similarities are obtained by comprehensive calculation, which can more accurately evaluate the similarity between the target user and the associated users, thereby obtaining the first entity word set suitable for the target user.

[0168] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, in another optional embodiment provided by the embodiments of the present application, determining the first similarity between the target user and the associated users based on the target entity word set and the associated entity word set may specifically include:

[0169] Determine the number of intersections based on the target entity word set and the associated entity word set, and determine the number of unions based on the target entity word set and the associated entity word set;

[0170] The ratio of the number of intersections to the number of unions is used as the first similarity between the target user and the associated users;

[0171] Determining the second similarity between the target user and the associated user based on the user relationship data of the target user and the user relationship data of the associated user may specifically include:

[0172] Determining friend similarity based on the target user's friend data and the first weight value;

[0173] Determining attention similarity based on the attention data of the target user, the attention data of the associated users, and the second weight value;

[0174] Determine the likes similarity based on the likes data of the target user, the likes data of the associated users, and the third weight value;

[0175] Determining the collection similarity based on the collection data of the target user, the collection data of the associated users, and the fourth weight value;

[0176] The second similarity between the target user and the associated users is determined based on the friend similarity, the follow similarity, the like similarity, and the favorite similarity.

[0177] In one or more embodiments, a method for calculating a first similarity and a second similarity is introduced. As can be seen from the aforementioned embodiments, after selecting at least one associated user related to a target user based on a profile score, it is necessary to calculate the first similarity and the second similarity between the two users (i.e., the target user and the associated user).

[0178] Specifically, the similarity between users can be measured based on collaborative filtering. Collaborative filtering refers to leveraging the preferences of a group of people with similar interests and shared experiences to recommend information of interest to users. Individuals collaborate to provide a certain level of response (e.g., rating) to the information and record this response to achieve filtering purposes, thereby helping others filter information. Responses are not necessarily limited to information of particular interest; records of information of particular disinterest may also be included in the calculation.

[0179] The following will explain them separately with examples.

[0180] 1. A method for calculating the first similarity;

[0181] For example, assume that the target user's target entity word set has 50 entity words, and the associated user's associated entity word set has 100 entity words, where the target entity word set and the associated entity word set have 20 identical entity words. Based on this, the number of intersections determined by the target entity word set and the associated entity word set is 20. The number of unions determined by the target entity word set and the associated entity word set is 130. Therefore, the first similarity between the target user and the associated user can be calculated using the following formula:

[0182]

[0183] Among them, S uv represents the similarity (i.e., the first similarity) between user u (i.e., the target user) and user v (i.e., the associated user). N(u) represents the entity word set of user u (i.e., the target entity word set). N(v) represents the entity word set of user v (i.e., the associated entity word set). |N(u)∩N(v)| represents the number of intersections. |N(u)∪N(v)| represents the number of unions.

[0184] 2. A method for calculating the second similarity;

[0185] For example, assuming that the target user's friend data includes the associated user's account, the friend similarity between the target user and the associated user is 1*W1, where W1 represents a first weight value. For example, W1 can be set to 0.5 or 1, and is not limited here. Assuming that the target user's friend data does not include the associated user's account, the friend similarity between the target user and the associated user is 0.

[0186] For example, assume that the target user's follow data contains 50 accounts (personal accounts and public accounts, etc.), and the associated user's follow data contains 100 accounts (personal accounts and public accounts, etc.). Among them, the target user's follow data and the associated user's follow data have 20 common accounts. Based on this, according to the target user's follow data and the associated user's follow data, the number of account follow intersections is determined to be 20, and the number of account follow unions is determined to be 130. Therefore, the following formula can be used to calculate the follow similarity between the target user and the associated user:

[0187]

[0188] Among them, S1 uvrepresents the attention similarity between user u (i.e., the target user) and user v (i.e., the associated user). W2 represents the second weight value, which can be set to 0.5, 1, etc., and is not limited here. N1(u) represents the attention data of user u. N1(v) represents the attention data of user v. |N1(u)∩N1(v)| represents the intersection number of account attention sets. |N1(u)∪N1(v)| represents the union number of account attention sets.

[0189] For example, suppose the target user's likes data contains 50 likes for graphic content, and the associated user's likes data contains 100 likes for graphic content. The target user's likes data and the associated user's likes data have 20 identical graphic content. Based on this, the target user's likes data and the associated user's likes data determine that the number of likes intersections is 20, and the number of likes unions is 130. Therefore, the likes similarity between the target user and the associated user can be calculated using the following formula:

[0190]

[0191] Among them, S2 uv represents the likes similarity between user u (i.e., the target user) and user v (i.e., the associated user). W3 represents the third weight value, which can be set to 0.5, 1, etc., and is not limited here. N2(u) represents the likes data of user u. N2(v) represents the likes data of user v. |N2(u)∩N2(v)| represents the number of likes intersections. |N2(u)∪N2(v)| represents the number of likes unions.

[0192] For example, assume that the target user's favorites include 50 favorited images and texts, and the associated user's favorites include 100 favorited images and texts. The target user's favorites and the associated user's favorites share 20 identical images and texts. Based on this, the target user's favorites and the associated user's favorites determine that the number of favorite intersections is 20, and the number of favorite unions is 130. Therefore, the following formula can be used to calculate the similarity between the target user and the associated user's favorites:

[0193]

[0194] Among them, S3 uv represents the similarity of favorites between user u (i.e., the target user) and user v (i.e., the associated user). W4 represents the fourth weight value, which can be set to 0.5, 1, etc., and is not limited here. N3(u) represents the favorites of user u. N3(v) represents the favorites of user v. |N3(u)∩N3(v)| represents the number of favorite intersections. |N3(u)∪N3(v)| represents the number of favorite unions.

[0195] Add the friend similarity, follow similarity, like similarity, and favorite similarity, or after weighted processing, the second similarity between the target user and the associated user can be obtained.

[0196] Furthermore, in the embodiments of the present application, a method for calculating the first similarity and the second similarity is provided. Through the above method, the similarity between users is measured using the idea of collaborative filtering. Thus, on the one hand, the similarity between users can be calculated without constructing a complex model, and on the other hand, it can combine the commonalities between users to objectively measure the possible common preferences and interests between users, thereby improving the feasibility of the solution.

[0197] Optionally, based on the corresponding respective embodiments above, in another optional embodiment provided by the embodiments of the present application, obtaining the set of entity words to be recommended for the target user may specifically include: Figure 5 Obtain the search keywords of the target user;

[0198] Determine the associated text according to the search keywords;

[0199] Obtain at least one entity word that hits the hot word dictionary from the associated text, and use the at least one entity word as the second set of entity words in the set of entity words to be recommended, where the hot word dictionary includes at least one hot word whose heat value is greater than or equal to the heat threshold within a preset time period.

[0200] In one or more embodiments, a method for obtaining the second set of entity words from the hot word dictionary is introduced. As can be seen from the foregoing embodiments, a second set of entity words related to the target user can also be obtained, where the second set of entity words includes at least one entity word belonging to the hot word. Hereinafter, an example of a search keyword queried by the target user within a period of time will be used for introduction. It can be understood that in practical applications, corresponding processing can be performed on each search keyword queried by the target user, so details are not described here.

[0201] Specifically, taking the search keyword "Tianwen" input by the user as an example, please refer to

[0202] [[ID=二十五]]Specifically, taking the search keyword "Tianwen" input by the user as an example, please refer to Figure 6 , Figure 6This is a flow chart of determining associated texts based on search keywords in an embodiment of the present application. As shown in the figure, based on the inverted index principle, the associated texts (i.e., news titles and news texts) containing the search keywords "Tianwen" can be found according to the search keyword "Tianwen". Here, the word document matrix model can be used, wherein the word document matrix is ​​a conceptual model that expresses a kind of inclusion relationship between the two. Through the word document matrix, it is very convenient to know which keywords a certain document contains and which documents contain a certain keyword. For example, the associated texts containing the search keyword "Tianwen" include "Tianwen-1 successfully landed on Mars" and "What are the characteristics of the Mars rover".

[0203] As you can understand, an inverted index is generally represented by a keyword, followed by its frequency (i.e., the number of times it appears), and location (i.e., the article or webpage in which it appears, along with related information such as the date and author). It is equivalent to indexing thousands of web pages on the Internet, much like the table of contents or tags in a book. Readers who want to read a chapter on a certain topic can directly find the relevant pages based on the table of contents, without having to search from the first page to the last page.

[0204] Based on this, at least one entity word that hits the hot word dictionary is found from these associated texts, that is, the second entity word set is composed of at least one entity word that hits the hot word dictionary. Among them, the hot word dictionary includes at least one hot word whose popularity value is greater than or equal to the popularity threshold within a preset time period. Taking the preset time period of nearly 1 month as an example, assuming that the popularity threshold is 500, when the number of searches for a certain entity word is greater than or equal to 600, it means that the popularity value of the entity word is greater than or equal to the popularity threshold, and therefore, the entity word is regarded as a hot word.

[0205] The following will be combined Figure 7 Introducing the method of constructing a hot word dictionary. For easier understanding, please refer to Figure 7 , Figure 7 A flowchart of constructing a hot word dictionary in an embodiment of the present application is shown in the figure. Specifically:

[0206] In step S1, hot news is sorted in real time. The goal is to select hot words to be added to the hot word dictionary from the news titles and texts ranked in the top ten of "hot searches".

[0207] In step S2, the news title and news body are segmented. For Chinese, word segmentation can be performed based on a rule-based method (e.g., based on a vocabulary) or a statistical method. Word segmentation is the primary and basic work of text processing.

[0208] In step S3, each word after the word segmentation process is tagged with a part of speech. Part of speech tagging refers to the process of marking a correct part of speech for each word in the word segmentation result, that is, the process of determining whether each word is a noun, verb, adjective or other part of speech.

[0209] In step S4, named entity recognition (NER) is performed on the words after part-of-speech tagging. Named entity recognition, also known as "proper name recognition", refers to the identification of entities with specific meanings in the text, mainly including names of people, places, organizations, and proper nouns. Generally, the task of named entity recognition is to identify three major categories (entity class, time class, and number class) and seven subcategories (names of people, organizations, places, time, date, currency, and percentage) of named entities in the text.

[0210] In step S5, the weight value of each entity word is set. For example, the entity word appearing in the news title is given a weight value of 2.73, and the entity word appearing in the news text is given a weight value of 1. Based on this, the heat value of the entity word can be calculated separately. For example, the entity word "Tianwen" appears 100 times in the news title and 300 times in the news text, that is, the heat value of the entity word "Centennial Celebration" is (100*2.73+300*1), that is, the heat value is "573". Assuming that the heat threshold is "500", the heat value of the entity word "Tianwen" is greater than the heat threshold. Therefore, the entity word "Tianwen" belongs to the hot word, so the hot word "Tianwen" is added to the hot word dictionary.

[0211] In step S6, an inverted index is constructed for each hot word. That is, the hot word dictionary can also be stored in the form of an inverted index. The inverted index originates from the need to search for records based on attribute values ​​in practical applications. Each entry in this index table includes an attribute value and the address of each record with that attribute value. This is because the attribute value is not determined by the record, but rather by the record's location.

[0212] Secondly, in an embodiment of the present application, a method for obtaining a second entity word set from a hot word dictionary is provided. Through the above method, based on the associated text of the target user, at least one entity word can be obtained from the hot word dictionary as the second entity word set. Because the second entity word set includes hot entity words in the recent period, the entity word set to be recommended has entity words with richer content and types, thereby making the recommended entity words more diversified, thereby providing more in-depth and broad information on the page displaying the recommended entity words.

[0213] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, in another optional embodiment provided by the embodiments of the present application, obtaining a set of entity words to be recommended to a target user may specifically include:

[0214] Obtain the search keywords of the target user;

[0215] Determine the associated text according to the search keywords;

[0216] Obtain at least one entity word that hits the rare word dictionary from the associated text, and use the at least one entity word as the third entity word set in the set of entity words to be recommended, where the rare word dictionary includes at least one rare word whose occurrence frequency is less than or equal to the first frequency threshold within a preset time period.

[0217] In one or more embodiments, a method for obtaining the third entity word set from the rare word dictionary is introduced. As can be seen from the foregoing embodiments, it is also possible to obtain the third entity word set related to the target user, where the third entity word set includes at least one entity word belonging to a rare word. Hereinafter, an example of a search keyword queried by the target user within a period of time will be used for introduction. It can be understood that in practical applications, corresponding processing can be performed on each search keyword queried by the target user, so it will not be elaborated here.

[0218] Specifically, taking the search keyword "Tianwen" input by the user as an example, please refer to again Figure 6 , based on the inverted index principle, the information titles and information texts containing the search keyword "Tianwen" can be found according to the search keyword "Tianwen". It should be noted that the method for finding the associated text based on the inverted index principle has been described in the foregoing embodiments, so it will not be limited here.

[0219] Based on this, at least one entity word that hits the rare word dictionary is found from these associated texts, that is, the third entity word set is composed of at least one entity word that hits the rare word dictionary. Among them, the rare word dictionary includes at least one rare word whose occurrence frequency is less than or equal to the first frequency threshold within a preset time period. Taking the preset time period as the recent 1 month as an example, assuming that the first frequency threshold is 0.01, when the occurrence frequency of a certain entity word is less than or equal to 0.01, the entity word can be used as a rare word.

[0220] The method for constructing the rare word dictionary will be introduced below. First, a large amount of text (for example, one million articles) is crawled from the Internet, and then these texts are subjected to word segmentation, part-of-speech tagging, named entity recognition and other processing to obtain a number of entity words. For each entity word, the number of times it appears in the article is counted respectively. For example, the entity word "Yaoling" appears 10 times in one million articles. Based on this, the occurrence frequency of the entity word "Yaoling" is 10 / 1000000 = 0.00001. Assuming that the first frequency threshold is 0.01, the occurrence frequency of the entity word "Yaoling" is less than the first frequency threshold. Therefore, the entity word "Yaoling" belongs to a rare word, and the rare word "Yaoling" is added to the rare word dictionary.

[0221] In addition, the uncommon word dictionary can also be maintained manually, for example, uncommon words can be added to the uncommon word dictionary regularly, or uncommon words can be deleted from the uncommon word dictionary.

[0222] It is understandable that the regular expression algorithm for determining whether an entity word is an uncommon word is as follows:

[0223] NSPredicate*predicate=[NSPredicate predicateWithFormat:@"SELFMATCHES%@",@"[\u4e00-\u9fa5]"];

[0224] / *Regular expression to determine whether it is Chinese* /

[0225] if([predicate evaluateWithObject:name]){

[0226] / / It is Chinese

[0227] }else{

[0228] / / Not Chinese

[0229] Secondly, in an embodiment of the present application, a method for obtaining a third entity word set from a dictionary of uncommon words is provided. Through the above method, based on the associated text of the target user, at least one entity word can be obtained from the dictionary of uncommon words as the third entity word set. Because the third entity word set includes some infrequently used entity words (i.e., uncommon words), the entity word set to be recommended has entity words with richer content and types, thereby making the recommended entity words more diversified, thereby providing more in-depth and broad information on the page displaying the recommended entity words.

[0230] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, in another optional embodiment provided by the embodiments of the present application, obtaining a set of entity words to be recommended to a target user may specifically include:

[0231] Get the target user's search keywords;

[0232] Determine related texts based on search keywords;

[0233] At least one entity word that hits a popular word dictionary is obtained from the associated text, and the at least one entity word is used as a fourth entity word set in the set of entity words to be recommended, wherein the popular word dictionary includes at least one popular word whose appearance frequency within a preset time period is greater than or equal to a second frequency threshold.

[0234] In one or more embodiments, a method for obtaining a fourth set of entity words from a popular word dictionary is introduced. As can be seen from the foregoing embodiments, a fourth set of entity words related to the target user can also be obtained, where the fourth set of entity words includes at least one entity word belonging to popular words. Hereinafter, an example of a search keyword queried by the target user within a period of time will be used for introduction. It can be understood that in practical applications, corresponding processing can be performed on each search keyword queried by the target user, so details will not be elaborated herein.

[0235] Specifically, taking the search keyword "Tianwen" input by the user as an example, please refer to again Figure 6 , based on the principle of inverted index, the information titles and information texts containing the search keyword "Tianwen" can be found according to the search keyword "Tianwen". It should be noted that the method for finding associated texts based on the principle of inverted index has been described in the foregoing embodiments, so no limitation will be made herein.

[0236] Based on this, at least one entity word that hits the popular word dictionary is found from these associated texts, that is, the fourth set of entity words is composed of at least one entity word that hits the popular word dictionary. Among them, the popular word dictionary includes at least one popular word whose appearance frequency is greater than or equal to the second frequency threshold within a preset time period. Taking the preset time period as the recent 1 month as an example, assuming that the second frequency threshold is 0.1, if the appearance frequency of a certain entity word is greater than or equal to 0.1, this entity word can be used as a popular word.

[0237] Hereinafter, the method for constructing the popular word dictionary will be introduced. First, a large amount of text (for example, 10,000 articles) is crawled from the Internet, and then these texts are subjected to word segmentation, part-of-speech tagging, named entity recognition and other processing to obtain a number of entity words. For each entity word, the number of times it appears in the article is respectively counted. For example, the entity word "foodie" appears 2000 times in 10,000 articles. Based on this, the appearance frequency of the entity word "foodie" is 2000 / 10000 = 0.2. Assuming that the second frequency threshold is 0.2, the appearance frequency of the entity word "foodie" is greater than the second frequency threshold. Therefore, the entity word "foodie" belongs to the popular words, and the popular word "foodie" is added to the popular word dictionary.

[0238] In addition, the popular word dictionary can also be maintained manually. For example, popular words are regularly added to the popular word dictionary, or popular words are deleted from the popular word dictionary.

[0239] For ease of understanding, please refer to Table 1, which is a schematic data storage structure of popular words.

[0240] Table 1 <00005​​​​Among them, "heat" is related to the occurrence frequency. For example, if the occurrence frequency of a certain popular word is greater than or equal to the second frequency threshold and greater than or equal to the third frequency threshold, then the heat of this popular word is considered "high". Among them, the third frequency threshold is greater than the second frequency threshold.

[0243] Secondly, in the embodiments of the present application, a method for obtaining a fourth set of entity words from a popular word dictionary is provided. Through the above method, based on the associated text of the target user, at least one entity word can be obtained from the popular word dictionary as the fourth set of entity words. Since the fourth set of entity words includes entity words that are popular on the network, the set of entity words to be recommended has entity words with richer content and types, thereby making the recommended entity words more diversified, and thus providing more in-depth and extensive information on the page displaying the recommended entity words.

[0244] Optionally, on the basis of the above Figure 5 corresponding embodiments, in another optional embodiment provided by the embodiments of the present application, obtaining the set of entity words to be recommended for the target user may specifically include:

[0245] Obtain the search keywords of the target user;

[0246] Determine the associated text according to the search keywords;

[0247] If the entity word in the associated text hits the audio-visual word dictionary, obtain the target interaction data of the entity word according to the audio-visual word dictionary, where the audio-visual word dictionary includes the mapping relationship between the audio-visual word and the interaction data;

[0248] If the target interaction data of at least one entity word meets the recommendation condition, then use at least one entity word as the fifth set of entity words in the set of entity words to be recommended.

[0249] In one or more embodiments, a method for obtaining a fifth set of entity words from an audio-visual word dictionary is introduced. It can be seen from the foregoing embodiments that a fifth set of entity words related to the target user can also be obtained, where the fifth set of entity words includes at least one entity word belonging to the audio-visual word. Hereinafter, an example of a search keyword queried by the target user within a period of time will be used for introduction. It can be understood that in actual applications, corresponding processing can be performed on each search keyword queried by the target user, so it will not be elaborated here.

[0250] Specifically, taking the search keyword "Tianwen" input by the user as an example, please refer to Figure 6 again. Based on the principle of the inverted index, the information titles and information texts containing the search keyword "Tianwen" can be found according to the search keyword "Tianwen". It should be noted that the method of finding the associated text based on the principle of the inverted index has been described in the foregoing embodiments, so it will not be limited here.

[0251] Based on this, at least one entity word that hits the audio-visual word dictionary is found from these associated texts, that is, the fifth entity word set is composed of at least one entity word that hits the audio-visual word dictionary. Among them, the audio-visual word dictionary includes the mapping relationship between audio-visual words and interaction data. The method of constructing the audio-visual word dictionary will be introduced below. For each audio-visual word, its interaction data is counted respectively. Among them, the interaction data includes but is not limited to the view count, evaluation count, share count, and click count. For ease of understanding, please refer to Table 2, which is an illustration of the mapping relationship between audio-visual words and interaction data.

[0252] Table 2

[0253]

[0254] Taking Table 2 as an example, assuming that the associated text includes the entity word "Detective Conan", the target interaction data is the interaction data corresponding to the entity word "Detective Conan". The audio-visual word dictionary can also be maintained manually. For example, audio-visual words are added to the audio-visual word dictionary regularly, or audio-visual words are deleted from the audio-visual word dictionary.

[0255] Exemplarily, if the view count, evaluation count, share count, or click count included in the target interaction data is greater than or equal to the quantity threshold, it is determined that the target interaction data meets the recommendation condition.

[0256] Exemplarily, if the sum of the view count, evaluation count, share count, and click count included in the target interaction data is greater than or equal to the quantity threshold, it is determined that the target interaction data meets the recommendation condition.

[0257] Secondly, in the embodiments of the present application, a method for obtaining the fifth entity word set from the audio-visual word dictionary is provided. Through the above method, based on the associated text of the target user, at least one entity word's target interaction data can be obtained from the audio-visual word dictionary. The target interaction data can represent the popularity and popularity of the audio-visual word. If the target interaction data of at least one entity word meets the recommendation condition, it means that these entity words can be used as the fifth entity word set of the target user. Since the fifth entity word set includes audio-visual words with higher popularity, the待推荐实体词集合 has entity words with richer content and types, thereby making the recommended entity words more diversified, and thus providing more in-depth and extensive information on the page displaying the recommended entity words.

[0258] Optionally, based on the corresponding embodiments above Figure 5 In another optional embodiment provided by the embodiments of the present application, obtaining the待推荐实体词集合 of the target user specifically may include:

[0259] Obtain the search keywords of the target user;

[0260] Determine associated text according to search keywords;

[0261] Obtain at least one entity word in the alias dictionary that matches the associated text, and use the at least one entity word and the entity words that have the same pointed object as each entity word as the sixth entity word set in the set of entity words to be recommended, where the alias dictionary includes at least one group of entity words that have the same pointed object.

[0262] In one or more embodiments, a method for obtaining the sixth entity word set from the alias dictionary is introduced. As can be seen from the foregoing embodiments, it is also possible to obtain the sixth entity word set related to the target user, where the sixth entity word set includes at least two entity words that have the same pointed object. Hereinafter, an example of a search keyword queried by the target user within a period of time will be used for introduction. It can be understood that in practical applications, corresponding processing can be performed on each search keyword queried by the target user, so it will not be elaborated here.

[0263] Specifically, taking the search keyword "Tianwen" input by the user as an example, please refer again to Figure 6 , based on the inverted index principle, the information titles and information texts containing the search keyword "Tianwen" can be found according to the search keyword "Tianwen". It should be noted that the method for finding the associated text based on the inverted index principle has been described in the foregoing embodiments, so it will not be limited here.

[0264] Based on this, find at least one entity word in the alias dictionary from these associated texts, that is, the sixth entity word set is composed of at least one entity word that matches the alias dictionary. Among them, the alias dictionary includes at least one group of entity words that have the same pointed object. Hereinafter, the method for constructing the alias dictionary will be introduced. For the convenience of understanding, please refer to Table 3, and Table 3 is a schematic diagram of multiple groups of entity words that have the same pointed object.

[0265] Table 3

[0266] Original name Aliases Abbreviation English name Former name Zhang Xiaohua Story King AAA TIM - Heart-saving pills Baoxinwan - Quickactingheartpill -

[0267] Taking Table 3 as an example, assuming that the entity word "heart pill" is included in the associated text, then through the alias dictionary, it can be seen that the entity word "heart pill" also has an alias and an English name (that is, the same pointed object is "heart pill"), so the alias and the English name are also added to the sixth entity word set. The alias dictionary can also be manually maintained. For example, relevant content (such as the original name and alias, etc.) can be added to the alias dictionary regularly, or some words can be deleted from the alias dictionary.

[0268] Secondly, in an embodiment of the present application, a method for obtaining a sixth entity word set from an alias dictionary is provided. Through the above method, based on the associated text of the target user, at least one entity word can be obtained from the alias dictionary as the sixth entity word set. Because the sixth entity word set includes some entity words that have the same reference object but different literal descriptions, the recommended entity word set has entity words with richer and more comprehensive content and types, thereby making the recommended entity words more diversified, thereby providing more in-depth and broad information on the page displaying the recommended entity words.

[0269] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, in another optional embodiment provided by the embodiments of the present application, obtaining a set of entity words to be recommended to a target user may specifically include:

[0270] Acquire a first set of derived tags that are category-relevant to the user's interest data, and acquire a second set of derived tags that are semantically relevant to the interest tags;

[0271] The user interest data, the first derived tag set, and the second derived tag set are used as the seventh entity word set in the entity word set to be recommended.

[0272] In one or more embodiments, a method for obtaining a seventh entity word set through category expansion and semantic expansion is described. As can be seen from the aforementioned embodiments, user interest data of the target user can also be obtained. Based on this user interest data, a first set of derived tags and a second set of derived tags can be extracted, respectively. These are described below.

[0273] 1. Extracting a first derived tag set;

[0274] Specifically, taking any interest tag included in the user's interest data as an example, a first derived tag is obtained based on the category relevance with the interest tag. The category relevance can be set manually. For ease of understanding, please refer to Table 4, which is an example of the category relevance between interest tags.

[0275] Table 4

[0276] Interest tags First derivative label gourmet food Sports and entertainment entertainment Finance, sports, film and television, food physical education Food and entertainment Finance Entertainment, Technology science and technology Film and television, finance Film and Television Entertainment, technology, finance

[0277] As shown in Table 4, if the target user's interest data includes the interest tag "Entertainment," the first derived tag set includes "Finance," "Sports," "Film and Television," and "Food." This means that not only entity words related to "Entertainment" can be recommended to the target user, but also entity words related to "Finance," "Sports," "Film and Television," and "Food." Category expansion can prevent content recommendations from being too narrow.

[0278] 2. extracting a second derived tag set;

[0279] Specifically, taking any interest tag included in the user's interest data as an example, a second derivative tag is obtained based on the semantic relevance with the interest tag. For example, if the interest tag is "UEFA Champions League (i.e., UEFA Champions League)", then the second derivative tag set may include "Premier League (i.e., English Premier League)" and "La Liga (i.e., Spanish La Liga)". In other words, not only entity words related to "UEFA Champions League" can be recommended to the target user, but also entity words related to "La Liga" and "Premier League". Category expansion can avoid content recommendations being too single.

[0280] Secondly, in an embodiment of the present application, a method for obtaining the seventh entity word set through category expansion and semantic expansion is provided. Through the above method, user interest data can be collected to obtain some derivative labels as entity words in the seventh entity word set. Whether based on category expansion or semantic expansion, it can avoid the problem of overly single recommended content to a certain extent, thereby improving the diversity and richness of entity word recommendations.

[0281] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, another optional embodiment provided by the embodiment of the present application, based on user feature data and the set of entity words to be recommended, obtaining the matching score corresponding to each entity word in each entity word set through a fusion recommendation model, specifically may include:

[0282] For each entity word set in the entity word set to be recommended, based on the user feature data and the entity word set, the matching score corresponding to each entity word in the entity word set is obtained by integrating the recommendation sub-model included in the recommendation model;

[0283] Selecting at least one recommended entity word from the set of entity words to be recommended based on the matching score corresponding to each entity word in each entity word set may specifically include:

[0284] Determine the comprehensive matching scores of the M entity words according to the matching scores corresponding to each entity word in each entity word set, where M is an integer greater than 1;

[0285] The top K entity words with the largest comprehensive matching scores are used as at least one recommended entity word, where K is an integer greater than or equal to 1 and less than or equal to M.

[0286] In one or more embodiments, a method for selecting at least one recommended entity word based on a matching score is described. As can be seen from the aforementioned embodiments, the fused recommendation model includes at least one recommendation sub-model, where each recommendation sub-model is continuously self-learned based on recommendation and user behavior models. In this application, each recommendation sub-model inputs a set of entity words and extracted user feature data.

[0287] Specifically, for easier understanding, see Figure 8 , Figure 8 This is a schematic diagram of prediction based on the fusion recommendation model in an embodiment of the present application. As shown in the figure, the entity word set to be recommended includes N entity word sets. After the user feature data and the entity word set are feature processed, they are input into the corresponding recommendation sub-model as a whole, and the recommendation sub-model outputs the matching score of each entity word. Assume that the entity word set to be recommended includes two entity word sets, namely entity word set 1 and entity word set 2, wherein entity word set 1 includes the entity words "Shenzhen" and "Tianwen", and entity word set 2 includes the entity words "Shenzhen" and "Aspirin". After the entity word set 1 and the user feature data are input into the recommendation sub-model 1, the matching score of the output entity word "Shenzhen" is 8, and the matching score of the entity word "Tianwen" is 5. After the entity word set 2 and the user feature data are input into the recommendation sub-model 2, the matching score of the output entity word "Shenzhen" is 6, and the matching score of the entity word "Aspirin" is 3. Based on this, the matching scores output by each recommendation sub-model are summed or weighted averaged to obtain the comprehensive matching scores of the M entity words.

[0288] Taking the summation process as an example, the comprehensive matching score of the entity word "Shenzhen" is 14, the comprehensive matching score of the entity word "Tianwen" is 5, and the comprehensive matching score of the entity word "aspirin" is 3. The top K entity words with the largest comprehensive matching scores are used as at least one recommended entity word. Assuming K is 2, the recommended entity words are "Shenzhen" and "Tianwen".

[0289] When training the fusion recommendation model, user behavior records (for example, clicks, stays, and recommendations) are required as the true matching scores of entity words. Figure 9 , Figure 9 This is a schematic diagram of training based on the fusion recommendation model in an embodiment of the present application. As shown in the figure, the training of the fusion recommendation model requires training each recommendation sub-model separately, and each recommendation sub-model can output the predicted matching score of the entity word. Based on the predicted matching score and the actual matching score of the entity word, the model parameters of the recommendation sub-model are updated using the loss function.

[0290] Secondly, in an embodiment of the present application, a method for selecting at least one recommended entity word based on a matching score is provided. Through this method, a single entity word set can be input into the fusion recommendation model, and the recommendation sub-model in the fusion recommendation model performs predictions. Finally, the prediction results of each recommendation sub-model are combined, sorted according to user preferences, and the top K entity words are used as the K recommended entity words recommended to the user. In this way, entity words that are more in line with user preferences can be predicted, thereby improving the effect of personalized recommendations.

[0291] Optionally, in the above Figure 5 On the basis of the corresponding embodiments, another optional embodiment provided by the embodiment of the present application may further include:

[0292] In response to the target user's second trigger operation on the target recommended entity word, a third page is displayed, and page information related to the target recommended entity word is displayed on the third page, wherein the target recommended entity word is any one of the at least one recommended entity word.

[0293] In one or more embodiments, a method for displaying jump links on a terminal device is introduced. As can be seen from the above embodiments, entity words can be personalized recommended for each user. The personalized recommendation method will be described below with reference to diagrams.

[0294] Specifically, see Figure 10 , Figure 10 This is a schematic diagram of an interface showing recommended entity words in an embodiment of the present application. Figure 10 The figure (A) shows the page recommended for user A. Figure 10 (B) in the figure shows the page recommended for user B. For user A, the recommended entity words highlighted on the page are "information overload" and "recommendation system". If user A clicks "information overload", the second trigger operation for the target recommended entity word is triggered, wherein the target recommended entity word is "information overload". Based on this, in response to the selection instruction, the terminal device displays the third page, and the third page displays information related to "information overload". For user B, the recommended entity words highlighted on the page are "user behavior data" and "recommendation system". If user B clicks "user behavior data", the second trigger operation for the target recommended entity word is triggered, and a page link request is generated, wherein the target recommended entity word is "user behavior data". Based on this, in response to the selection instruction, the terminal device displays the third page, and the third page displays information related to "user behavior data".

[0295] It is understandable that for recommended entity words, if the recommendation times threshold is exceeded, no further recommendation will be made.

[0296] Secondly, in an embodiment of the present application, a method for displaying jump links on a terminal device is provided. Through this method, on the one hand, it is possible to provide users with entity words that are of greater interest to them, that is, to comprehensively recommend entities based on factors such as user interests, demographic characteristics, social relationships, user behavior, current hot topics, uncommon words, and sentiment analysis, thereby achieving personalized entity recommendations. On the other hand, it can also achieve content diversion, through which information can be directed to related news and videos.

[0297] The entity word recommendation device in this application is described in detail below. Figure 11 , Figure 11 This is a schematic diagram of an embodiment of an entity word recommendation device in an embodiment of the present application. The entity word recommendation device 20 includes:

[0298] The display module 210 is configured to display a first page, wherein the first page provides a page jump control;

[0299] The display module 210 is also used to respond to the target user's first trigger operation on the page jump control, display the second page, and highlight at least one recommended entity word on the second page, wherein the at least one recommended entity word is determined based on the user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data, and user interest data.

[0300] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application, the entity word recommendation device 20 further includes an acquisition module 220 and a recommendation module 230;

[0301] An acquisition module 220 is used to acquire user characteristic data of a target user;

[0302] The acquisition module 220 is further configured to acquire a set of entity words to be recommended for a target user, wherein the recommended entity word set includes N entity word sets, each entity word set includes at least one entity word, and N is an integer greater than or equal to 1;

[0303] The acquisition module 220 is further configured to obtain a matching score corresponding to each entity word in each entity word set through a fusion recommendation model based on the user feature data and the entity word set to be recommended;

[0304] The recommendation module 230 is configured to select at least one recommended entity word from the set of entity words to be recommended according to the matching score corresponding to each entity word in each set of entity words.

[0305] Optionally, in the above Figure 11On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application, the entity word recommendation device 20 further includes a determination module 240;

[0306] A determination module 240 is configured to determine a first portrait score corresponding to the target user based on the user behavior data and the user portrait data of the target user;

[0307] The determination module 240 is further configured to determine an associated user based on the first portrait score, wherein the associated user corresponds to a second portrait score, and an absolute value of a difference between the second portrait score and the first portrait score is less than or equal to a score threshold;

[0308] An acquisition module 210 is specifically configured to acquire a target entity word set of a target user and a related entity word set of related users;

[0309] Determining a first similarity between the target user and the associated users based on the target entity word set and the associated entity word set;

[0310] determining a second similarity between the target user and the associated user based on the user relationship data of the target user and the user relationship data of the associated user;

[0311] If the sum of the first similarity and the second similarity is greater than or equal to the similarity threshold, the target entity word set and the associated entity word set are taken as the first entity word set in the entity word set to be recommended.

[0312] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0313] The acquisition module 220 is specifically configured to determine the number of intersections according to the target entity word set and the associated entity word set, and to determine the number of unions according to the target entity word set and the associated entity word set;

[0314] The ratio of the number of intersections to the number of unions is used as the first similarity between the target user and the associated users;

[0315] The acquisition module 220 is specifically configured to determine the friend similarity based on the friend data of the target user and the first weight value;

[0316] Determining attention similarity based on the attention data of the target user, the attention data of the associated users, and the second weight value;

[0317] Determine the likes similarity based on the likes data of the target user, the likes data of the associated users, and the third weight value;

[0318] Determining the collection similarity based on the collection data of the target user, the collection data of the associated users, and the fourth weight value;

[0319] The second similarity between the target user and the associated users is determined based on the friend similarity, the follow similarity, the like similarity, and the favorite similarity.

[0320] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0321] Acquisition module 220, specifically used to obtain the search keywords of the target user;

[0322] Determine related text based on search keywords;

[0323] At least one entity word that hits the hot word dictionary is obtained from the associated text, and the at least one entity word is used as the second entity word set in the entity word set to be recommended, wherein the hot word dictionary includes at least one hot word whose popularity value is greater than or equal to a popularity threshold within a preset time period.

[0324] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0325] Acquisition module 220, specifically used to obtain the search keywords of the target user;

[0326] Determine related text based on search keywords;

[0327] At least one entity word that hits the uncommon word dictionary is obtained from the associated text, and the at least one entity word is used as a third entity word set in the set of entity words to be recommended, wherein the uncommon word dictionary includes at least one uncommon word whose occurrence frequency within a preset time period is less than or equal to a first frequency threshold.

[0328] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0329] Acquisition module 220, specifically used to obtain the search keywords of the target user;

[0330] Determine related text based on search keywords;

[0331] At least one entity word that hits a popular word dictionary is obtained from the associated text, and the at least one entity word is used as a fourth entity word set in the set of entity words to be recommended, wherein the popular word dictionary includes at least one popular word whose appearance frequency within a preset time period is greater than or equal to a second frequency threshold.

[0332] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0333] Acquisition module 220, specifically used to obtain the search keywords of the target user;

[0334] Determine related text based on search keywords;

[0335] If the entity word in the associated text hits the audio-visual word dictionary, then obtaining target interaction data of the entity word according to the audio-visual word dictionary, wherein the audio-visual word dictionary includes a mapping relationship between the audio-visual word and the interaction data;

[0336] If there is target interaction data of at least one entity word that meets the recommendation condition, the at least one entity word is used as the fifth entity word set in the set of entity words to be recommended.

[0337] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0338] Acquisition module 220, specifically used to obtain the search keywords of the target user;

[0339] Determine related text based on search keywords;

[0340] At least one entity word that hits the alias dictionary is obtained from the associated text, and the at least one entity word and the entity words that have the same pointing object as each entity word are taken as the sixth entity word set in the set of entity words to be recommended, wherein the alias dictionary includes at least one group of entity words that have the same pointing object.

[0341] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0342] The acquisition module 220 is specifically configured to acquire a first set of derived tags that are category-relevant to the user's interest data, and acquire a second set of derived tags that are semantically relevant to the interest tags;

[0343] The user interest data, the first derived tag set, and the second derived tag set are used as the seventh entity word set in the entity word set to be recommended.

[0344] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0345] The acquisition module 220 is specifically configured to obtain, for each entity word set in the entity word set to be recommended, a matching score corresponding to each entity word in the entity word set by integrating the recommendation sub-model included in the recommendation model based on the user feature data and the entity word set;

[0346] The recommendation module 230 is specifically configured to determine the comprehensive matching scores of the M entity words according to the matching scores corresponding to each entity word in each entity word set, where M is an integer greater than 1;

[0347] The top K entity words with the largest comprehensive matching scores are used as at least one recommended entity word, where K is an integer greater than or equal to 1 and less than or equal to M.

[0348] Optionally, in the above Figure 11 On the basis of the corresponding embodiment, in another embodiment of the entity word recommendation device 20 provided in the embodiment of the present application,

[0349] The display module 210 is also used to respond to the target user's second trigger operation on the target recommended entity word, display a third page, and display page information related to the target recommended entity word on the third page, wherein the target recommended entity word is any one of the at least one recommended entity word.

[0350] The entity word recommendation device provided in this application can be applied to the server. Figure 12 , Figure 12 This is a schematic diagram of a server structure provided in an embodiment of the present application. The server 300 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 322 (for example, one or more processors) and memory 332, and one or more storage media 330 (for example, one or more mass storage devices) for storing application programs 342 or data 344. Among them, the memory 332 and the storage medium 330 can be short-term storage or persistent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 322 can be configured to communicate with the storage medium 330 to execute a series of instruction operations in the storage medium 330 on the server 300.

[0351] The server 300 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input and output interfaces 358, and / or one or more operating systems 341, such as Windows Server 2003 or Windows Server 2003R. TM , Mac OS XTM , Unix TM ,Linux TM , FreeBSD TM etc.

[0352] The steps performed by the server in the above embodiment can be based on the Figure 12 The server structure shown.

[0353] The entity word recommendation device provided by this application can be applied to terminal devices, such as Figure 13 For ease of explanation, only the parts related to the embodiments of the present application are shown. For specific technical details not disclosed, please refer to the method part of the embodiments of the present application. The terminal device can be any terminal device including a mobile phone, tablet computer, personal digital assistant (PDA), point of sales (POS), car computer, etc., taking the mobile phone as an example:

[0354] Figure 13 The block diagram shows a partial structure of a mobile phone related to the terminal device provided in the embodiment of the present application. Figure 13 The mobile phone includes components such as a radio frequency (RF) circuit 410, a memory 420, an input unit 430, a display unit 440, a sensor 450, an audio circuit 460, a wireless fidelity (WiFi) module 470, a processor 480, and a power supply 490. Those skilled in the art will understand that Figure 13 The mobile phone structure shown in the figure does not constitute a limitation to the mobile phone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0355] The following combination Figure 13 A detailed introduction to the various components of a mobile phone:

[0356] The RF circuit 410 can be used to receive and send signals during information transmission or calls. In particular, after receiving the downlink information from the base station, it is sent to the processor 480 for processing; in addition, the designed uplink data is sent to the base station. Generally, the RF circuit 410 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, etc. In addition, the RF circuit 410 can also communicate with the network and other devices through wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0357] Memory 420 can be used to store software programs and modules. Processor 480 executes the various functional applications and data processing of the mobile phone by running the software programs and modules stored in memory 420. Memory 420 may mainly include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, memory 420 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0358] The input unit 430 can be used to receive input digital or character information, and to generate key signal input related to the user settings and function control of the mobile phone. Specifically, the input unit 430 may include a touch panel 431 and other input devices 432. The touch panel 431, also known as a touch screen, can collect user touch operations on or near it (such as operations performed by the user using any suitable object or accessory such as a finger, stylus, etc. on or near the touch panel 431) and drive the corresponding connection device according to a pre-set program. Optionally, the touch panel 431 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch direction, detects the signal caused by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device and converts it into touch point coordinates, which are then sent to the processor 480, and can receive commands sent by the processor 480 and execute them. In addition, the touch panel 431 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch panel 431, the input unit 430 may further include other input devices 432. Specifically, the other input devices 432 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power keys, etc.), a trackball, a mouse, and a joystick.

[0359] The display unit 440 can be used to display information input by the user or information provided to the user and various menus of the mobile phone. The display unit 440 may include a display panel 441. Optionally, the display panel 441 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), etc. Further, the touch panel 431 may cover the display panel 441. When the touch panel 431 detects a touch operation on or near it, it is transmitted to the processor 480 to determine the type of touch event. Subsequently, the processor 480 provides corresponding visual output on the display panel 441 according to the type of touch event. Although in Figure 13 In the embodiment, the touch panel 431 and the display panel 441 are used as two independent components to realize the input and output functions of the mobile phone, but in some embodiments, the touch panel 431 and the display panel 441 can be integrated to realize the input and output functions of the mobile phone.

[0360] The mobile phone may also include at least one sensor 450, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 441 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 441 and / or the backlight when the mobile phone is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that identify the posture of the mobile phone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that the mobile phone can also be configured with, such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., they will not be repeated here.

[0361] Audio circuit 460, speaker 461, and microphone 462 provide an audio interface between the user and the phone. Audio circuit 460 converts received audio data into electrical signals and transmits them to speaker 461, which then converts them into sound signals for output. Microphone 462, on the other hand, converts collected sound signals into electrical signals, which are then received by audio circuit 460 and converted into audio data. The audio data is then processed by processor 480 and transmitted to, for example, another phone via RF circuit 410, or stored in memory 420 for further processing.

[0362] WiFi is a short-range wireless transmission technology. The mobile phone can help users send and receive emails, browse the web and access streaming media through the WiFi module 470. It provides users with wireless broadband Internet access. Figure 13 A WiFi module 470 is shown, but it is understandable that it is not an essential component of the mobile phone and can be omitted as needed without changing the essence of the invention.

[0363] Processor 480 is the control center of the mobile phone. It connects all parts of the mobile phone using various interfaces and circuits. It executes software programs and / or modules stored in memory 420 and accesses data stored in memory 420 to perform various functions and process data. Processor 480 may optionally include one or more processing units. Alternatively, processor 480 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and application programs, while the modem processor primarily handles wireless communications. It is understood that the modem processor may not be integrated into processor 480.

[0364] The mobile phone also includes a power supply 490 (such as a battery) for supplying power to various components. Optionally, the power supply can be logically connected to the processor 480 through a power management system, thereby managing functions such as charging, discharging, and power consumption through the power management system.

[0365] Although not shown, the mobile phone may also include a camera, a Bluetooth module, etc., which will not be described in detail here.

[0366] The steps performed by the terminal device in the above embodiment can be based on the Figure 13 The terminal device structure shown.

[0367] A computer-readable storage medium is also provided in an embodiment of the present application. The computer-readable storage medium stores a computer program, which, when executed on a computer, enables the computer to execute the methods described in the aforementioned embodiments.

[0368] An embodiment of the present application also provides a computer program product including a program, which, when executed on a computer, enables the computer to execute the methods described in the aforementioned embodiments.

[0369] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0370] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0371] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0372] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0373] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0374] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for recommending entity words, characterized in that: include: Displaying a first page, wherein the first page provides a page jump control; In response to the target user's first trigger operation on the page jump control, a second page is displayed, and at least one recommended entity word on the relevant content in the second page is highlighted, wherein the at least one recommended entity word is determined based on the user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data and user interest data, and the relevant content includes title and / or text.

2. The recommendation method according to claim 1, characterized in that The method further comprises: Acquiring user characteristic data of the target user; Obtaining a set of entity words to be recommended for the target user, wherein the set of entity words to be recommended includes N entity word sets, each entity word set includes at least one entity word, and N is an integer greater than or equal to 1; Based on the user feature data and the set of entity words to be recommended, obtaining a matching score corresponding to each entity word in each set of entity words through a fusion recommendation model; At least one recommended entity word is selected from the set of entity words to be recommended according to the matching score corresponding to each entity word in each set of entity words.

3. The recommendation method according to claim 2, characterized in that: The method further comprises: Determining a first portrait score corresponding to the target user based on the user behavior data of the target user and the user portrait data of the target user; Determining an associated user according to the first portrait score, wherein the associated user corresponds to a second portrait score, and an absolute value of a difference between the second portrait score and the first portrait score is less than or equal to a score threshold; The step of obtaining the target user's entity word set to be recommended includes: Obtaining a target entity word set of the target user and a related entity word set of the related user; Determining a first similarity between the target user and the associated user based on the target entity word set and the associated entity word set; determining a second similarity between the target user and the associated user based on the user relationship data of the target user and the user relationship data of the associated user; If the sum of the first similarity and the second similarity is greater than or equal to a similarity threshold, the target entity word set and the associated entity word set are used as the first entity word set in the entity word set to be recommended.

4. The recommendation method according to claim 3, characterized in that: The determining, based on the target entity word set and the associated entity word set, a first similarity between the target user and the associated user includes: Determining the number of intersections based on the target entity word set and the associated entity word set, and determining the number of unions based on the target entity word set and the associated entity word set; Using the ratio of the number of intersections to the number of unions as the first similarity between the target user and the associated user; The determining, based on the user relationship data of the target user and the user relationship data of the associated user, a second similarity between the target user and the associated user includes: Determining friend similarity based on the friend data of the target user and the first weight value; Determining attention similarity according to the attention data of the target user, the attention data of the associated users, and a second weight value; Determining a like similarity based on the like data of the target user, the like data of the associated users, and a third weight value; determining a collection similarity based on the collection data of the target user, the collection data of the associated user, and a fourth weight value; The second similarity between the target user and the associated user is determined according to the friend similarity, the follow similarity, the like similarity, and the favorite similarity.

5. The recommendation method according to claim 2, characterized in that: The step of obtaining the target user's entity word set to be recommended includes: Obtaining the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits the hot word dictionary is obtained from the associated text, and the at least one entity word is used as the second entity word set in the entity word set to be recommended, wherein the hot word dictionary includes at least one hot word whose popularity value is greater than or equal to a popularity threshold within a preset time period.

6. The recommendation method according to claim 2, characterized in that: The step of obtaining the target user's entity word set to be recommended includes: Obtain the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits the uncommon word dictionary is obtained from the associated text, and the at least one entity word is used as the third entity word set in the set of entity words to be recommended, wherein the uncommon word dictionary includes at least one uncommon word whose occurrence frequency within a preset time period is less than or equal to a first frequency threshold.

7. The recommendation method according to claim 2, characterized in that: The step of obtaining the target user's entity word set to be recommended includes: Obtain the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits a buzzword dictionary is obtained from the associated text, and the at least one entity word is used as a fourth entity word set in the set of entity words to be recommended, wherein the buzzword dictionary includes at least one buzzword whose appearance frequency within a preset time period is greater than or equal to a second frequency threshold.

8. The recommendation method according to claim 2, characterized in that: The step of obtaining the target user's entity word set to be recommended includes: Obtain the target user's search keywords; Determining related texts according to the search keywords; If the entity word in the associated text hits the audio-visual word dictionary, obtaining target interaction data of the entity word according to the audio-visual word dictionary, wherein the audio-visual word dictionary includes a mapping relationship between audio-visual words and interaction data; If there is target interaction data of at least one entity word that meets the recommendation condition, the at least one entity word is used as the fifth entity word set in the set of entity words to be recommended.

9. The recommendation method according to claim 2, characterized in that: The step of obtaining the target user's entity word set to be recommended includes: Obtaining the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits the alias dictionary is obtained from the associated text, and the at least one entity word and the entity words that have the same pointing object as each entity word are used as the sixth entity word set in the set of entity words to be recommended, wherein the alias dictionary includes at least one group of entity words that have the same pointing object.

10. The recommendation method according to claim 2, characterized in that: The step of obtaining the target user's entity word set to be recommended includes: Acquire a first set of derived tags that are category-relevant to the user interest data, and acquire a second set of derived tags that are semantically relevant to the interest tags; The user interest data, the first derived tag set, and the second derived tag set are used as a seventh entity word set in the to-be-recommended entity word set.

11. The recommendation method according to claim 2, characterized in that: The obtaining, based on the user feature data and the set of entity words to be recommended, a matching score corresponding to each entity word in each set of entity words by a fusion recommendation model includes: For each entity word set in the set of entity words to be recommended, based on the user feature data and the entity word set, obtaining a matching score corresponding to each entity word in the set of entity words by using a recommendation sub-model included in the fusion recommendation model; Selecting at least one recommended entity word from the set of entity words to be recommended according to the matching score corresponding to each entity word in each set of entity words includes: Determining a comprehensive matching score of M entity words according to the matching score corresponding to each entity word in each entity word set, wherein M is an integer greater than 1; The top K entity words with the largest comprehensive matching scores are used as the at least one recommended entity word, where K is an integer greater than or equal to 1 and less than or equal to M.

12. The recommendation method according to any one of claims 1 to 11, characterized in that: The method further comprises: In response to the target user's second trigger operation on the target recommended entity word, a third page is displayed, and page information related to the target recommended entity word is displayed on the third page, wherein the target recommended entity word is any one of the at least one recommended entity word.

13. An entity word recommendation device, characterized in that: include: A display module, configured to display a first page, wherein the first page provides a page jump control; A display module is used to respond to the target user's first trigger operation on the page jump control, display the second page, and highlight at least one recommended entity word on the relevant content in the second page, wherein the at least one recommended entity word is determined based on the user characteristic data of the target user, and the user characteristic data of the target user includes at least one of user behavior data, user portrait data, user relationship data and user interest data, and the relevant content includes a title and / or text.

14. The device according to claim 13, characterized in that The device also includes an acquisition module and a recommendation module; The acquisition module is configured to acquire user feature data of the target user; acquire a set of entity words to be recommended for the target user, wherein the set of entity words to be recommended includes N entity word sets, each entity word set includes at least one entity word, and N is an integer greater than or equal to 1; and based on the user feature data and the set of entity words to be recommended, acquire a matching score corresponding to each entity word in each entity word set through a fusion recommendation model; The recommendation module is configured to select the at least one recommended entity word from the set of entity words to be recommended according to the matching score corresponding to each entity word in each set of entity words.

15. The device according to claim 14, characterized in that The apparatus further includes a determination module; The determination module is configured to determine a first portrait score corresponding to the target user based on the user behavior data of the target user and the user portrait data of the target user; Determining an associated user according to the first portrait score, wherein the associated user corresponds to a second portrait score, and an absolute value of a difference between the second portrait score and the first portrait score is less than or equal to a score threshold; The acquisition module is specifically used to: Obtaining a target entity word set of the target user and a related entity word set of the related user; Determining a first similarity between the target user and the associated user based on the target entity word set and the associated entity word set; determining a second similarity between the target user and the associated user based on the user relationship data of the target user and the user relationship data of the associated user; If the sum of the first similarity and the second similarity is greater than or equal to a similarity threshold, the target entity word set and the associated entity word set are used as the first entity word set in the entity word set to be recommended.

16. The device according to claim 15, characterized in that The acquisition module is specifically used to: Determining the number of intersections based on the target entity word set and the associated entity word set, and determining the number of unions based on the target entity word set and the associated entity word set; Using the ratio of the number of intersections to the number of unions as the first similarity between the target user and the associated user; The acquisition module is specifically used to: Determining friend similarity based on the friend data of the target user and the first weight value; Determining attention similarity according to the attention data of the target user, the attention data of the associated users, and a second weight value; Determining a like similarity based on the like data of the target user, the like data of the associated users, and a third weight value; determining a collection similarity based on the collection data of the target user, the collection data of the associated user, and a fourth weight value; The second similarity between the target user and the associated user is determined according to the friend similarity, the follow similarity, the like similarity, and the favorite similarity.

17. The device according to claim 14, characterized in that The acquisition module is specifically used to: Obtain the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits the hot word dictionary is obtained from the associated text, and the at least one entity word is used as the second entity word set in the entity word set to be recommended, wherein the hot word dictionary includes at least one hot word whose popularity value is greater than or equal to a popularity threshold within a preset time period.

18. The device according to claim 14, characterized in that The acquisition module is specifically used to: Obtaining the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits the uncommon word dictionary is obtained from the associated text, and the at least one entity word is used as the third entity word set in the set of entity words to be recommended, wherein the uncommon word dictionary includes at least one uncommon word whose occurrence frequency within a preset time period is less than or equal to a first frequency threshold.

19. The device according to claim 14, characterized in that The acquisition module is specifically used to: Obtaining the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits a buzzword dictionary is obtained from the associated text, and the at least one entity word is used as a fourth entity word set in the set of entity words to be recommended, wherein the buzzword dictionary includes at least one buzzword whose appearance frequency within a preset time period is greater than or equal to a second frequency threshold.

20. The device according to claim 14, characterized in that The acquisition module is specifically used to: Obtaining the target user's search keywords; Determining related texts according to the search keywords; If the entity word in the associated text hits the audio-visual word dictionary, obtaining target interaction data of the entity word according to the audio-visual word dictionary, wherein the audio-visual word dictionary includes a mapping relationship between audio-visual words and interaction data; If there is target interaction data of at least one entity word that meets the recommendation condition, the at least one entity word is used as the fifth entity word set in the set of entity words to be recommended.

21. The device according to claim 14, characterized in that The acquisition module is specifically used to: Obtaining the target user's search keywords; Determining related texts according to the search keywords; At least one entity word that hits the alias dictionary is obtained from the associated text, and the at least one entity word and the entity words that have the same pointing object as each entity word are used as the sixth entity word set in the set of entity words to be recommended, wherein the alias dictionary includes at least one group of entity words that have the same pointing object.

22. The device according to claim 14, characterized in that The acquisition module is specifically used to: Acquire a first set of derived tags that are category-relevant to the user interest data, and acquire a second set of derived tags that are semantically relevant to the interest tags; The user interest data, the first derived tag set, and the second derived tag set are used as a seventh entity word set in the to-be-recommended entity word set.

23. The device according to claim 14, characterized in that The acquisition module is specifically used to: For each entity word set in the set of entity words to be recommended, based on the user feature data and the entity word set, obtaining a matching score corresponding to each entity word in the set of entity words by using a recommendation sub-model included in the fusion recommendation model; Selecting at least one recommended entity word from the set of entity words to be recommended according to the matching score corresponding to each entity word in each set of entity words includes: Determining a comprehensive matching score of M entity words according to the matching score corresponding to each entity word in each entity word set, wherein M is an integer greater than 1; The top K entity words with the largest comprehensive matching scores are used as the at least one recommended entity word, where K is an integer greater than or equal to 1 and less than or equal to M.

24. The device according to any one of claims 13 to 23, characterized in that The display module is also used to respond to the target user's second trigger operation on the target recommended entity word, display a third page, and display page information related to the target recommended entity word on the third page, wherein the target recommended entity word is any one of the at least one recommended entity word.

25. A computer device, characterized in that: include: Memory, processor, and bus system; Wherein, the memory is used to store programs; The processor is configured to execute the program in the memory, and the processor is configured to execute the recommendation method according to any one of claims 1 to 12 according to instructions in the program code; The bus system is used to connect the memory and the processor so that the memory and the processor can communicate with each other. 26 . A computer-readable storage medium comprising instructions, which, when executed on a computer, causes the computer to perform the recommendation method according to claim 1 .

27. A computer program product, characterized in that The computer program product includes computer instructions. A processor of a computer device executes the computer instructions, so that the computer device performs the recommendation method according to any one of claims 1 to 12.

Images