Content recommendation method and device, equipment and medium

Abstract

The invention belongs to the field of artificial intelligence, and relates to a content recommendation method, which comprises the following steps: acquiring target input content of a user; a language model is adopted to analyze the intention information to obtain intention feature information and content key information; judging whether the user intention is clear or not according to the intention feature information; if the intention is clear, retrieving first target candidate content from a preset content database based on the content key information, and taking the first target candidate content as target recommendation content; if the intention is not clear, obtaining a historical interaction record of the user, retrieving second target candidate content from the content database based on the historical interaction record and the content key information, and taking the second target candidate content as target recommendation content; and recommending the target recommendation content to the user terminal of the user. The invention further provides a device, equipment and a medium. In addition, the invention also relates to a block chain technology, and the target input content can be stored in a block chain. The accuracy of content recommendation can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a content recommendation method, device, equipment and medium. Background Art

[0002] With the rapid development of artificial intelligence technology, human-computer dialogue systems have become a key component in many application scenarios, especially in the field of content recommendation, which can provide users with personalized content recommendations by analyzing user historical behavior, interest preferences and other information. However, existing human-computer dialogue content recommendation methods still face many challenges in practical applications.

[0003] Traditional recommendation algorithms mostly rely on rule-based matching or simple machine learning models. These methods are unable to cope with complex and diverse user needs and content. Specifically, they are obviously inadequate in capturing user potential interests, understanding context, and predicting user future behavior. As a result, the recommended content often deviates from the actual needs of users, and the accuracy of content recommendations is difficult to be satisfactory. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a content recommendation method, device, equipment and medium to solve the problem that the accuracy of content recommendation is not high when using traditional recommendation algorithms to recommend content.

[0005] In order to solve the above technical problems, the present application embodiment provides a content recommendation method, which adopts the following technical solutions:

[0006] Obtain the target input content of the user; use a preset language model to parse the target input content to obtain the user's intention feature information and content key information; determine whether the user's intention is clear based on the intention feature information; if the user's intention is clear, retrieve the corresponding first target candidate content from a preset content database based on the content key information, and use the first target candidate content as the target recommended content; if the user's intention is unclear, obtain the user's historical interaction record, retrieve the corresponding second target candidate content from the content database based on the historical interaction record and content key information, and use the second target candidate content as the target recommended content; recommend the target recommended content to the user's user terminal.

[0007] Furthermore, the step of using a preset language model to parse the target input content to obtain the user's intention feature information and content key information specifically includes:

[0008] The target input content is input into a preset language model for parsing to obtain the parsed input content; the intention recognition algorithm is used to perform intent analysis on the input content to obtain the user's intention feature information; the key information of the input content is extracted to obtain the user's content key information.

[0009] Further, the step of determining whether the user's intention is clear based on the intention feature information specifically includes:

[0010] Obtain an intent feature vector corresponding to the intent feature information; use a pre-trained intent recognition model to analyze the intent feature vector to obtain the user's intent category and confidence score; compare the confidence score with a preset intent clarity threshold; if the confidence score is greater than the intent clarity threshold, determine that the intent corresponding to the intent category is clear; if the confidence score is less than or equal to the intent clarity threshold, determine that the intent corresponding to the intent category is unclear.

[0011] Furthermore, based on the content key information, the step of retrieving the corresponding first target candidate content from a preset content database specifically includes:

[0012] Based on the content key information and the intention feature information, a first search condition is constructed; based on the first search condition, an inverted index algorithm is used to retrieve a first candidate content set from a preset content database; the similarity between each first candidate content in the first candidate content set and the target input content is calculated, and the first candidate content with a similarity higher than a preset threshold is determined as the first target candidate content.

[0013] Furthermore, based on the historical interaction records and the key content information, the step of retrieving the corresponding second target candidate content from the content database specifically includes:

[0014] Based on a preset knowledge graph, semantic expansion is performed on the key information of the content to obtain a key information enhancement set; based on the enhanced information and intention feature information in the key information enhancement set, a second search condition is constructed; a second candidate content set matching the second search condition is retrieved from the content database; according to historical interaction records, each second candidate content in the second candidate content set is sorted to obtain a candidate content sequence; and a second target candidate content is determined from the candidate content sequence.

[0015] Furthermore, the step of sorting the second candidate contents in the second candidate content set according to the historical interaction records to obtain a candidate content sequence specifically includes:

[0016] Perform preference analysis on historical interaction records to obtain historical preference feature information of users; extract content feature information of each second candidate content in the second candidate content set; calculate the similarity between the historical preference feature information and the content feature information, and based on the similarity, sort the second candidate contents in the second candidate content set to obtain a candidate content sequence.

[0017] Furthermore, the step of recommending target recommended content to the user terminal of the user specifically includes:

[0018] The device parameters of the user terminal are obtained, including the device type and screen size; the format of the target recommended content is adjusted according to the device type and screen size to obtain the target recommended content; and the target recommended content is pushed to the user terminal for display.

[0019] In order to solve the above technical problems, the present application also provides a content recommendation device, which adopts the following technical solutions:

[0020] An acquisition module is used to acquire the target input content of the user;

[0021] The parsing module is used to parse the target input content using a preset language model to obtain the user's intention feature information and content key information;

[0022] A judgment module is used to judge whether the user's intention is clear based on the intention feature information;

[0023] A first retrieval module is used to retrieve corresponding first target candidate content from a preset content database based on content key information if the user's intention is clear, and use the first target candidate content as target recommended content;

[0024] A second retrieval module is used to obtain the user's historical interaction records if the user's intention is unclear, retrieve the corresponding second target candidate content from the content database based on the historical interaction records and content key information, and use the second target candidate content as the target recommended content;

[0025] The recommendation module is used to recommend target recommended content to the user terminal of the user.

[0026] In order to solve the above technical problems, an embodiment of the present application further provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the above content recommendation method are implemented.

[0027] In order to solve the above technical problems, an embodiment of the present application also provides a computer-readable storage medium, which stores computer-readable instructions. The computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above content recommendation method.

[0028] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects: by introducing a preset language model, the user's target input content is deeply analyzed, and the user's intention feature information and content key information are accurately extracted. Based on this, it is possible to intelligently determine whether the user's intention is clear, and adopt different strategies for content recommendation. When the user's intention is clear, the first target candidate content is directly retrieved from the content database based on the content key information to ensure that the recommended content is highly matched with the user's needs; when the user's intention is unclear, a comprehensive analysis is performed in combination with the user's historical interaction records, and the second target candidate content that is more in line with the user's potential interests and preferences is screened from the content database as a recommendation. This innovative method not only significantly improves the accuracy and personalization of content recommendations, but also effectively solves the shortcomings of traditional recommendation algorithms in capturing user potential interests and understanding contextual contexts, bringing users a more accurate and intelligent content recommendation experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0030] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0031] Figure 2 It is a flowchart of a content recommendation method provided by this application;

[0032] Figure 3 It is a structural diagram of a content recommendation device provided by the present application;

[0033] Figure 4 It is a structural schematic diagram of a computer device provided by this application. DETAILED DESCRIPTION

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0035] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0036] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0037] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0038] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0039] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to a laptop computer 1011, a tablet computer 1012 or a mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0040] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0041] It should be noted that the content recommendation method provided in the embodiment of the present application is generally executed by a server, and accordingly, the content recommendation device is generally arranged in the server.

[0042] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0043] Continue to refer Figure 2 , shows a flow chart of an embodiment of a content recommendation method according to the present application. The content recommendation method comprises the following steps:

[0044] Step S201, obtaining the target input content of the user.

[0045] In this embodiment, the content recommendation method is executed on the electronic device (eg Figure 1 The server shown in the figure) can obtain the target input content through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0046] Among them, users refer to the objects for which content is recommended by the human-computer dialogue system, and have specific needs and interest preferences. For example, a consumer who is using an online shopping platform to find a specific product.

[0047] The target input content refers to the text information input by the user through the human-computer dialogue system to express his needs or inquiries. For example, the user inputs "I want to find a pair of sports shoes suitable for summer".

[0048] Step S202: parse the target input content using a preset language model to obtain user's intention feature information and content key information.

[0049] The language model refers to an algorithm model used to process and understand natural language, which can parse and generate input text. The natural language model can be used to parse the target input content and extract intent and key information.

[0050] Among them, the intention feature information refers to the intention description of the target input content input by the user obtained through language model analysis, which represents the user's real needs and purposes. For example, if the user inputs "I want to watch a movie", the intention feature information is "entertainment" and "watch a movie".

[0051] Among them, content key information refers to the information in the user input content that is crucial for content recommendation, such as item name, type, price, etc. For example, if the user inputs "I want to buy a smartphone with a price lower than 500 yuan", the content key information may include "smartphone" and "price lower than 500 yuan".

[0052] Step S203: determine whether the user's intention is clear based on the intention feature information.

[0053] In one embodiment, the intention feature information can be used to determine whether the intention of the user input content is clear and specific, and whether it can be directly used for content recommendation. For example, the user inputs "I want to find something to watch", the intention is not clear, while "I want to watch science fiction movie A" is clear.

[0054] Step S204: if the user's intention is clear, then based on the content key information, the corresponding first target candidate content is retrieved from the preset content database, and the first target candidate content is used as the target recommended content.

[0055] Among them, a content database refers to a database that stores a large amount of content data and is used to retrieve and recommend content based on user needs and intentions. For example, a database containing movies, books, music, etc.

[0056] The first target candidate content refers to the content that is retrieved from the content database and highly matches the user input content when the user's intention is clear. For example, if the user wants to watch science fiction movie A, the first target candidate content is science fiction movie A.

[0057] The target recommended content refers to the content that is ultimately recommended to the user, which may be the first target candidate content or the second target candidate content, for example, a movie or a book recommended based on user input.

[0058] Step S205: if the user's intention is unclear, the user's historical interaction records are obtained, and based on the historical interaction records and content key information, the corresponding second target candidate content is retrieved from the content database, and the second target candidate content is used as the target recommended content.

[0059] The historical interaction records refer to the user's previous interaction data with the system, including user click, browse, purchase and other behavior records. For example, a list of products that a user has browsed before.

[0060] The second target candidate content refers to the content retrieved from the content database in combination with the historical interaction records and the key content information when the user intention is unclear.

[0061] Step S206: recommending target recommended content to the user terminal of the user.

[0062] The user terminal refers to a device used by the user to receive and display recommended content, such as a smart phone, tablet computer, computer, etc. For example, the user receives recommended content through a shopping application on a smart phone.

[0063] The embodiment of the present application can introduce a preset language model to deeply analyze the user's target input content and accurately extract the user's intention feature information and content key information. Based on this, it is possible to intelligently determine whether the user's intention is clear, and adopt different strategies for content recommendation. When the user's intention is clear, the first target candidate content is directly retrieved from the content database based on the content key information to ensure that the recommended content is highly matched with the user's needs; when the user's intention is unclear, a comprehensive analysis is performed in combination with the user's historical interaction records, and the second target candidate content that is more in line with the user's potential interests and preferences is screened from the content database as a recommendation. This innovative method not only significantly improves the accuracy and personalization of content recommendations, but also effectively solves the shortcomings of traditional recommendation algorithms in capturing user potential interests and understanding contextual contexts, bringing users a more accurate and intelligent content recommendation experience.

[0064] In some optional implementations of this embodiment, step 202, using a preset language model to parse the target input content to obtain the user's intention feature information and content key information, specifically includes the following steps:

[0065] The target input content is input into a preset language model for parsing to obtain the parsed input content; the intention recognition algorithm is used to perform intent analysis on the input content to obtain the user's intention feature information; the key information of the input content is extracted to obtain the user's content key information.

[0066] The parsed input content refers to the text information obtained after the language model parses the target input content, which is easier to understand and analyze. For example, the spoken sentence input by the user is parsed into standardized text.

[0067] Among them, the intent recognition algorithm is used to analyze the parsed input content and recognize the user's intent. For example, the intent recognition model based on deep learning.

[0068] Among them, key information refers to the information fragments extracted from the input content that are crucial for content recommendation, such as the name or brand of the item in the user input.

[0069] In one example, the target input content may be text information about investment needs input by the user through the human-computer dialogue system. For example, the user inputs "I want to find a financial product with low risk and stable income". First, the system can input the target input content into a preset language model for parsing. The language model has been trained with a large number of financial field texts and can accurately understand the financial terms and expressions input by the user, thereby obtaining the parsed input content. Then, the system uses the intention recognition algorithm to perform intent analysis on the parsed input content. In this example, the intention recognition algorithm can recognize that the user's intention is to find a financial product that meets his risk tolerance and income expectations. Through algorithm analysis, the user's intention feature information is obtained, that is, the user's preference for the risk level and income stability of financial products. Finally, the key information of the input content is extracted to obtain the user's content key information. In this example, the key information may include two keywords "low risk" and "stable income".

[0070] In one example, the target input content can be text information about financial information needs input by the user through the human-computer dialogue system, such as "I want to know the latest stock market trends". The system can use the preset language model to parse the target input content to obtain the parsed input content. Then, the intention recognition algorithm is used to analyze the input content and identify the user's intention feature information about stock market trends. Then, the system extracts key information of the input content, such as keywords such as "stock market trends", to obtain key content information.

[0071] The embodiment of the present application can parse and process the target input content by inputting it into a preset language model, which can make full use of the language model's ability to understand and generate natural language, accurately capture the semantics and contextual information of the user's input, and thus obtain more accurate and comprehensive parsed input content. Furthermore, the intent recognition algorithm is used to perform intent analysis on the parsed input content, which can deeply explore the user's potential needs and purposes, extract the user's intent feature information, and provide a more accurate basis for subsequent recommendations. At the same time, extracting the key information of the input content can highlight the core points of the user's input and ignore irrelevant and redundant information, thereby obtaining more refined and valuable content key information.

[0072] In some optional implementations of this embodiment, step S203, judging whether the user's intention is clear according to the intention feature information, specifically includes the following steps:

[0073] Obtain an intent feature vector corresponding to the intent feature information; use a pre-trained intent recognition model to analyze the intent feature vector to obtain the user's intent category and confidence score; compare the confidence score with a preset intent clarity threshold; if the confidence score is greater than the intent clarity threshold, determine that the intent corresponding to the intent category is clear; if the confidence score is less than or equal to the intent clarity threshold, determine that the intent corresponding to the intent category is unclear.

[0074] The intention feature vector refers to the representation of the user intention in the form of a vector, which is convenient for subsequent analysis and processing. For example, the user intention "watch a movie" is represented as a vector containing multiple feature dimensions.

[0075] The intent recognition model refers to a model that has been trained to recognize user intent and can be based on a deep learning algorithm, for example, a neural network model that can recognize the category of user input intent.

[0076] Among them, the intent category refers to the classification result of the user's intention, such as shopping, entertainment, learning, etc. For example, if the user enters "I want to learn programming", its intent category may be "learning".

[0077] The confidence score is a score that indicates the credibility or accuracy of the intent recognition model's judgment of the user's intent. For example, the confidence score of the intent recognition model for the user's intent "shopping" is 0.9.

[0078] The intent clarity threshold is a score threshold used to determine whether the user's intent is clear. For example, the intent clarity threshold is set to 0.7, and when the confidence score is greater than 0.7, the user's intent is considered to be clear.

[0079] In one example, first, the intent feature information can be converted into an intent feature vector using specific technical means, such as word embedding or TF-IDF. This conversion can process and analyze user intent more efficiently. Then, the intent feature vector can be analyzed using a pre-trained intent recognition model. The intent recognition model has been trained with a large amount of data in the field of financial management, and can accurately identify the user's financial management intention category, such as "seeking high-yield products" and "pursuing stable investment", and give a corresponding confidence score. The confidence score reflects the accuracy of the model's judgment of the user's intention. In order to ensure the reliability of the recommendation, the confidence score can be compared with the preset intent clarity threshold. In the example, the intent clarity threshold can be set to 0.8 (this value can be adjusted according to the actual application scenario). If the confidence score is greater than 0.8, it is determined that the user's intention is clear. Take a user input "I want to find an investment product with an annualized yield of more than 5% and low risk" as an example. First, "I want to find an investment product with an annualized yield of more than 5% and low risk" is converted into intent feature information, and further converted into an intent feature vector. After analysis by the pre-trained intent recognition model, the user's intent category is identified as "pursuing a stable and high-yield investment" and a high confidence score (such as 0.9) is given. Therefore, it is determined that the user's intent is clear.

[0080] The embodiment of the present application can obtain the intent feature vector corresponding to the intent feature information and analyze it using the pre-trained intent recognition model, so as to more deeply understand the user's intent, accurately identify the user's intent category, and give a corresponding confidence score. This score reflects the reliability of the system's judgment of the user's intent. Furthermore, by comparing the confidence score with the preset intent clarity threshold, the clarity of the user's intent can be automatically determined. When the confidence score is higher than the threshold, the user's intent can be confirmed, thereby providing more accurate services and responses.

[0081] In some optional implementations of this embodiment, step S204, based on the content key information, retrieves the corresponding first target candidate content from a preset content database, specifically including the following steps:

[0082] Based on the content key information and the intention feature information, a first search condition is constructed; based on the first search condition, an inverted index algorithm is used to retrieve a first candidate content set from a preset content database; the similarity between each first candidate content in the first candidate content set and the target input content is calculated, and the first candidate content with a similarity higher than a preset threshold is determined as the first target candidate content.

[0083] The first search condition refers to a search condition constructed based on user intention feature information and content key information, and is used to retrieve the first target candidate content from the content database. For example, a search condition is constructed based on "science fiction movie A" and a price range input by the user.

[0084] Among them, the inverted index algorithm refers to an efficient text retrieval algorithm that quickly locates text that meets the retrieval conditions by building an inverted index table. For example, quickly find a list of movies that meet user needs in a content database.

[0085] The first candidate content set refers to a content set retrieved from the content database according to the first search condition, for example, a series of movies retrieved according to the first search condition.

[0086] In one example, the key content information mainly refers to the specific topics, keywords, etc. involved in the user input, while the intention feature information reflects the user's real intention and needs. Based on the extracted key content information and intention feature information, a first search condition is constructed. This search condition not only includes the specific content input by the user, but also incorporates the user's real intention, thereby more comprehensively reflecting the user's needs. Subsequently, an inverted index algorithm is used to retrieve a first candidate content set from a preset content database according to the constructed first search condition. The inverted index algorithm indexes the content in the content database in advance, making the retrieval process more efficient and faster. In order to further improve the accuracy of the recommendation, the similarity between each first candidate content in the first candidate content set and the target content input by the user is also calculated. The calculation of the similarity can be based on the cosine similarity algorithm, which can quantify the degree of similarity between two contents. Finally, the first candidate content with a similarity higher than a preset threshold is determined as the first target candidate content and presented to the user.

[0087] The embodiment of the present application can construct the first search condition by integrating content key information and intention feature information, which significantly improves the pertinence and accuracy of content retrieval. By using the inverted index algorithm, a quick search is performed in the preset content database, which effectively reduces the scope of candidate content and improves retrieval efficiency. Furthermore, by calculating the similarity between the first candidate content and the target input content, and setting a reasonable preset threshold for screening, it is ensured that the final first target candidate content is not only highly relevant to the user input, but also can accurately meet the user's actual needs.

[0088] In some optional implementations of this embodiment, step S205, based on the historical interaction records and the content key information, retrieves the corresponding second target candidate content from the content database, specifically including the following steps:

[0089] Based on a preset knowledge graph, semantic expansion is performed on the key information of the content to obtain a key information enhancement set; based on the enhanced information and intention feature information in the key information enhancement set, a second search condition is constructed; a second candidate content set matching the second search condition is retrieved from the content database; according to historical interaction records, each second candidate content in the second candidate content set is sorted to obtain a candidate content sequence; and a second target candidate content is determined from the candidate content sequence.

[0090] Among them, knowledge graph refers to a structured knowledge representation method that can represent information such as entities, attributes, relationships, etc. For example, a graph containing information about movies, actors, directors, etc.

[0091] Among them, the key information enhanced set refers to the set obtained by expanding the key information of the content based on the knowledge graph, which contains more information related to user needs. For example, the user input "science fiction movie" is expanded to a set containing information such as director and actor.

[0092] The second search condition refers to a search condition constructed based on the enhanced information and the intention feature information in the key information enhancement set, and is used to retrieve the second target candidate content from the content database.

[0093] The second candidate content set refers to a content set retrieved from the content database according to the second search condition, for example, a series of movies retrieved according to the expanded search condition (the second search condition).

[0094] The candidate content sequence refers to a sequence obtained by sorting the second candidate contents in the second candidate content set, and the sorting basis may be similarity with the user's historical preferences. For example, the movies most similar to the user's historical preferences are sorted at the front.

[0095] In one example, for the key information of the user's content, such as "science fiction movies", the preset knowledge graph can be used to semantically expand these key information. The knowledge graph is a graph structure database containing a large number of entities, attributes and relationships, which can reflect the association and hierarchical relationship between entities. By querying the knowledge graph, enhanced information related to "science fiction movies" can be obtained, such as "adaptation of science fiction novels", "high-grossing science fiction movies", etc., to form a key information enhancement set. After obtaining the key information enhancement set, the second search condition is constructed in combination with the user's intention feature information (such as "find a high-rated movie"). This condition not only includes the key information of the content directly input by the user, but also incorporates the enhanced information obtained through the knowledge graph expansion, thereby more comprehensively reflecting the user's needs. According to the constructed second search condition, a matching second candidate content set is retrieved from the preset content database. The content database can be a movie database, an article database, etc., and is selected according to the actual application scenario. In order to further improve the accuracy of the recommendation, the user's historical interaction records can also be used to sort the second candidate contents in the second candidate content set. The historical interaction records include the user's past clicks, ratings, comments and other behavioral data. By analyzing these data, the user's preference for the candidate content is evaluated, thereby generating a candidate content sequence. Finally, the top-ranked candidate content is selected from the candidate content sequence as the second target candidate content and presented to the user. These target candidate contents are not only highly relevant to the key content information and intention feature information input by the user, but are also optimized and sorted based on the user's historical interaction records, so they are more likely to meet the user's actual needs.

[0096] The embodiment of the present application can significantly enhance the comprehensiveness and accuracy of information retrieval by introducing a preset knowledge graph to perform deep semantic expansion of content key information. The second search condition constructed based on the expanded key information enhancement set and the intent feature information more accurately reflects user needs, and the second candidate content set retrieved from the content database is therefore highly consistent with user expectations. Furthermore, the use of historical interaction records to intelligently sort candidate content not only improves the personalization of recommended content, but also ensures the consistency of the second target candidate content with user preferences.

[0097] In some optional implementations of this embodiment, the step of “sorting the second candidate contents in the second candidate content set according to the historical interaction records to obtain a candidate content sequence” specifically includes the following steps:

[0098] Perform preference analysis on historical interaction records to obtain historical preference feature information of users; extract content feature information of each second candidate content in the second candidate content set; calculate the similarity between the historical preference feature information and the content feature information, and based on the similarity, sort the second candidate contents in the second candidate content set to obtain a candidate content sequence.

[0099] The historical preference feature information refers to feature information extracted from the user's historical interaction records that can represent the user's preferences, for example, the type or brand of goods that the user has purchased before.

[0100] The content feature information refers to feature information describing the attributes of the second candidate content, such as the type, director, and actors of the movie.

[0101] In one example, historical interaction records of users are obtained, including but not limited to the user's actions such as clicking, browsing, rating, and commenting on the content. Then, natural language processing technology and machine learning algorithms are used to perform in-depth analysis on these records to extract the user's historical preference feature information. These feature information may include the content type, theme, style, etc. preferred by the user, as well as the user's emotional tendency and interest changes in the content. After obtaining the second candidate content set, content feature extraction is performed on each second candidate content. This includes analyzing the text information, image information, audio information, etc. of the content to extract feature information such as the theme, keywords, emotional color, style, etc. of the content. These feature information will be used for subsequent similarity calculation with the user's historical preference feature information. Using a similarity calculation algorithm (such as cosine similarity), the similarity between the user's historical preference feature information and the content feature information of each second candidate content is calculated. Then, based on the similarity, the second candidate content in the second candidate content set is sorted to obtain a candidate content sequence. The higher the similarity, the higher the ranking, indicating that it is more in line with the user's historical preferences and current needs.

[0102] The embodiment of the present application can accurately extract the user's historical preference feature information by deeply analyzing the user's historical interaction records, laying a solid foundation for personalized content recommendation. Furthermore, the content feature information of each candidate content in the second candidate content set is carefully extracted to ensure the accurate match between the recommended content and the user's needs. By calculating the similarity between the historical preference feature information and the content feature information, the intelligent sorting of the second candidate content is achieved, so that the candidate content sequence is more in line with the user's personalized preferences.

[0103] In some optional implementations of this embodiment, step S206, recommending target recommended content to the user terminal of the user, may specifically include the following steps:

[0104] The device parameters of the user terminal are obtained, including the device type and screen size; the format of the target recommended content is adjusted according to the device type and screen size to obtain the target recommended content; and the target recommended content is pushed to the user terminal for display.

[0105] Among them, device parameters refer to parameters that describe the performance and characteristics of user terminal devices, such as device type, screen size, etc. For example, the model and screen size of the smartphone used by the user.

[0106] Among them, device type refers to the category of user terminal, such as smartphones, tablet computers, etc.

[0107] The screen size refers to the size of the user terminal screen, which is used to adjust the display format of the recommended content. For example, the screen size of the smartphone used by the user is 6.5 inches.

[0108] In one example, the device parameters of the user terminal are obtained, including but not limited to the device type (such as a smart phone, tablet computer, smart TV, etc.) and screen size (such as screen size, resolution, etc.). These device parameters are an important basis for subsequent content format adjustment. The format of the target recommended content is adjusted according to these parameters. For example, for smart phones with smaller screen sizes, the pictures and texts of the recommended content will be appropriately reduced and streamlined to adapt to the small screen display. For smart TVs with larger screen sizes and higher resolutions, richer and higher-definition pictures and video content will be provided to enhance the user's viewing experience. After the format is adjusted, the target recommended content is pushed to the user terminal for display.

[0109] The embodiment of the present application can achieve intelligent adjustment of the target recommended content format by accurately obtaining the device parameters of the user terminal, including the device type and screen size. In view of the differences in different device types and screen sizes, it can automatically adapt to the most suitable content display method, such as adjusting the image size, text layout, etc., to ensure that the target recommended content can present the best display effect on the user terminal. This innovative method not only improves the user experience of content recommendation, makes the recommended content more in line with the characteristics of the user's device, but also enhances the readability and attractiveness of the content. At the same time, by intelligently adjusting the recommended content format, display problems caused by device incompatibility are effectively avoided, and the success rate of content push and user satisfaction are improved.

[0110] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target input content, intent feature information and content key information, the above-mentioned target input content, intent feature information and content key information can also be stored in a blockchain node.

[0111] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0112] The embodiments of the present application can build and optimize related models and networks based on artificial intelligence technology, such as language models. Among them, artificial intelligence (AI) models are the crystallization of theory and practice that simulate the decision-making process of human intelligence through algorithms and data analysis to solve complex problems, predict future trends or realize automation tasks. These models use a large amount of historical data and real-time information, and are trained and optimized through a specific algorithm framework to achieve efficient, accurate and reliable performance.

[0113] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0114] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0115] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a content recommendation device, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0116] like Figure 3 As shown, the content recommendation device 400 of this embodiment includes: an acquisition module 401, a parsing module 402, a judgment module 403, a first search module 404, a second search module 405 and a recommendation module 406. Among them:

[0117] The acquisition module 401 is used to acquire the target input content of the user;

[0118] The parsing module 402 is used to parse the target input content using a preset language model to obtain the user's intention feature information and content key information;

[0119] The judgment module 403 is used to judge whether the user's intention is clear according to the intention feature information;

[0120] The first search module 404 is used to search for corresponding first target candidate content from a preset content database based on the content key information if the user's intention is clear, and use the first target candidate content as the target recommended content;

[0121] The second retrieval module 405 is used to obtain the user's historical interaction records if the user's intention is unclear, retrieve the corresponding second target candidate content from the content database based on the historical interaction records and content key information, and use the second target candidate content as the target recommended content;

[0122] The recommendation module 406 is used to recommend target recommended content to the user terminal of the user.

[0123] The embodiment of the present application can introduce a preset language model to deeply analyze the user's target input content and accurately extract the user's intention feature information and content key information. Based on this, it is possible to intelligently determine whether the user's intention is clear, and adopt different strategies for content recommendation. When the user's intention is clear, the first target candidate content is directly retrieved from the content database based on the content key information to ensure that the recommended content is highly matched with the user's needs; when the user's intention is unclear, a comprehensive analysis is performed in combination with the user's historical interaction records, and the second target candidate content that is more in line with the user's potential interests and preferences is screened from the content database as a recommendation. This innovative method not only significantly improves the accuracy and personalization of content recommendations, but also effectively solves the shortcomings of traditional recommendation algorithms in capturing user potential interests and understanding contextual contexts, bringing users a more accurate and intelligent content recommendation experience.

[0124] In one embodiment, the parsing module 402 includes:

[0125] The parsing submodule is used to input the target input content into a preset language model for parsing and processing to obtain the parsed input content;

[0126] The intention analysis submodule is used to use the intention recognition algorithm to perform intention analysis on the input content and obtain the user's intention feature information;

[0127] The extraction submodule is used to extract key information of the input content and obtain the user's content key information.

[0128] The embodiment of the present application can parse and process the target input content by inputting it into a preset language model, which can make full use of the language model's ability to understand and generate natural language, accurately capture the semantics and contextual information of the user's input, and thus obtain more accurate and comprehensive parsed input content. Furthermore, the intent recognition algorithm is used to perform intent analysis on the parsed input content, which can deeply explore the user's potential needs and purposes, extract the user's intent feature information, and provide a more accurate basis for subsequent recommendations. At the same time, extracting the key information of the input content can highlight the core points of the user's input and ignore irrelevant and redundant information, thereby obtaining more refined and valuable content key information.

[0129] In one embodiment, the determination module 403 includes:

[0130] The first acquisition submodule is used to acquire the intention feature vector corresponding to the intention feature information;

[0131] The analysis submodule is used to analyze the intent feature vector using a pre-trained intent recognition model to obtain the user's intent category and confidence score;

[0132] A comparison submodule, for comparing the confidence score with a preset intent clarity threshold;

[0133] A first determination submodule, configured to determine that the intent corresponding to the intent category is clear if the confidence score is greater than the intent clarity threshold;

[0134] The second determination submodule is configured to determine that the intent corresponding to the intent category is unclear if the confidence score is less than or equal to the intent clarity threshold.

[0135] The embodiment of the present application can obtain the intent feature vector corresponding to the intent feature information and analyze it using the pre-trained intent recognition model, so as to more deeply understand the user's intent, accurately identify the user's intent category, and give a corresponding confidence score. This score reflects the reliability of the system's judgment of the user's intent. Furthermore, by comparing the confidence score with the preset intent clarity threshold, the clarity of the user's intent can be automatically determined. When the confidence score is higher than the threshold, the user's intent can be confirmed, thereby providing more accurate services and responses.

[0136] In one embodiment, the first search module 404 includes:

[0137] A first construction submodule is used to construct a first search condition according to content key information and intention feature information;

[0138] A first search submodule is used to retrieve a first candidate content set from a preset content database using an inverted index algorithm according to a first search condition;

[0139] The calculation submodule is used to calculate the similarity between each first candidate content in the first candidate content set and the target input content, and determine the first candidate content with a similarity higher than a preset threshold as the first target candidate content.

[0140] The embodiment of the present application can construct the first search condition by integrating content key information and intention feature information, which significantly improves the pertinence and accuracy of content retrieval. By using the inverted index algorithm, a quick search is performed in the preset content database, which effectively reduces the scope of candidate content and improves retrieval efficiency. Furthermore, by calculating the similarity between the first candidate content and the target input content, and setting a reasonable preset threshold for screening, it is ensured that the final first target candidate content is not only highly relevant to the user input, but also can accurately meet the user's actual needs.

[0141] In one embodiment, the second search module 405 includes:

[0142] The extension submodule is used to semantically expand the key information of the content based on the preset knowledge graph to obtain an enhanced set of key information;

[0143] A second construction submodule is used to construct a second search condition based on the enhanced information and the intention feature information in the key information enhancement set;

[0144] A second search submodule, used to retrieve a second candidate content set matching a second search condition from the content database;

[0145] A sorting submodule, used to sort the second candidate contents in the second candidate content set according to the historical interaction records to obtain a candidate content sequence;

[0146] The determination submodule is used to determine the second target candidate content from the candidate content sequence.

[0147] The embodiment of the present application can significantly enhance the comprehensiveness and accuracy of information retrieval by introducing a preset knowledge graph to perform deep semantic expansion of content key information. The second search condition constructed based on the expanded key information enhancement set and the intent feature information more accurately reflects user needs, and the second candidate content set retrieved from the content database is therefore highly consistent with user expectations. Furthermore, the use of historical interaction records to intelligently sort candidate content not only improves the personalization of recommended content, but also ensures the consistency of the second target candidate content with user preferences.

[0148] In one embodiment, the sorting submodule is also used to perform preference analysis on historical interaction records to obtain historical preference feature information of users; extract content feature information of each second candidate content in the second candidate content set; calculate the similarity between the historical preference feature information and the content feature information, and based on the similarity, sort the second candidate contents in the second candidate content set to obtain a candidate content sequence.

[0149] The embodiment of the present application can accurately extract the user's historical preference feature information by deeply analyzing the user's historical interaction records, laying a solid foundation for personalized content recommendation. Furthermore, the content feature information of each candidate content in the second candidate content set is carefully extracted to ensure the accurate match between the recommended content and the user's needs. By calculating the similarity between the historical preference feature information and the content feature information, the intelligent sorting of the second candidate content is achieved, so that the candidate content sequence is more in line with the user's personalized preferences.

[0150] In one embodiment, the recommendation module 406 includes:

[0151] The second acquisition submodule is used to acquire device parameters of the user terminal, where the device parameters include device type and screen size;

[0152] An adjustment submodule, used to adjust the format of the target recommended content according to the device type and screen size to obtain the target recommended content;

[0153] The push submodule is used to push the target recommended content to the user terminal for display.

[0154] The embodiment of the present application can achieve intelligent adjustment of the target recommended content format by accurately obtaining the device parameters of the user terminal, including the device type and screen size. In view of the differences in different device types and screen sizes, it can automatically adapt to the most suitable content display method, such as adjusting the image size, text layout, etc., to ensure that the target recommended content can present the best display effect on the user terminal. This innovative method not only improves the user experience of content recommendation, makes the recommended content more in line with the characteristics of the user's device, but also enhances the readability and attractiveness of the content. At the same time, by intelligently adjusting the recommended content format, display problems caused by device incompatibility are effectively avoided, and the success rate of content push and user satisfaction are improved.

[0155] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0156] The computer device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 6 having a memory 61, a processor 62, and a network interface 63, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASIC), programmable gate arrays (FPGA), digital signal processors (DSP), embedded devices, etc.

[0157] Computer devices can be computing devices such as desktop computers, notebooks, PDAs, and cloud servers. Computer devices can interact with users through keyboards, mice, remote controls, touch pads, or voice control devices.

[0158] The memory 61 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 61 can be an internal storage unit of the computer device 6, such as a hard disk or memory of the computer device 6. In other embodiments, the memory 61 can also be an external storage device of the computer device 6, such as a plug-in hard disk equipped on the computer device 6, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 61 can also include both the internal storage unit of the computer device 6 and its external storage device. In this embodiment, the memory 61 is generally used to store the operating system and various application software installed on the computer device 6, such as computer-readable instructions of the content recommendation method, etc. In addition, the memory 61 can also be used to temporarily store various types of data that have been output or are to be output.

[0159] In some embodiments, the processor 62 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 62 is generally used to control the overall operation of the computer device 6. In this embodiment, the processor 62 is used to run computer-readable instructions stored in the memory 61 or process data, such as computer-readable instructions for running the content recommendation method.

[0160] The network interface 63 may include a wireless network interface or a wired network interface, and the network interface 63 is generally used to establish a communication connection between the computer device 6 and other electronic devices.

[0161] The embodiment of the present application can introduce a preset language model to deeply analyze the user's target input content and accurately extract the user's intention feature information and content key information. Based on this, it is possible to intelligently determine whether the user's intention is clear, and adopt different strategies for content recommendation. When the user's intention is clear, the first target candidate content is directly retrieved from the content database based on the content key information to ensure that the recommended content is highly matched with the user's needs; when the user's intention is unclear, a comprehensive analysis is performed in combination with the user's historical interaction records, and the second target candidate content that is more in line with the user's potential interests and preferences is screened from the content database as a recommendation. This innovative method not only significantly improves the accuracy and personalization of content recommendations, but also effectively solves the shortcomings of traditional recommendation algorithms in capturing user potential interests and understanding contextual contexts, bringing users a more accurate and intelligent content recommendation experience.

[0162] The present application also provides another implementation, namely, providing a computer-readable storage medium, which stores computer-readable instructions. The computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the content recommendation method as described above.

[0163] The embodiment of the present application can introduce a preset language model to deeply analyze the user's target input content and accurately extract the user's intention feature information and content key information. Based on this, it is possible to intelligently determine whether the user's intention is clear, and adopt different strategies for content recommendation. When the user's intention is clear, the first target candidate content is directly retrieved from the content database based on the content key information to ensure that the recommended content is highly matched with the user's needs; when the user's intention is unclear, a comprehensive analysis is performed in combination with the user's historical interaction records, and the second target candidate content that is more in line with the user's potential interests and preferences is screened from the content database as a recommendation. This innovative method not only significantly improves the accuracy and personalization of content recommendations, but also effectively solves the shortcomings of traditional recommendation algorithms in capturing user potential interests and understanding contextual contexts, bringing users a more accurate and intelligent content recommendation experience.

[0164] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0165] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A content recommendation method, characterized in that: The steps include: Get the user's target input content; Using a preset language model, the target input content is parsed to obtain the user's intention feature information and content key information; Determining whether the user's intention is clear according to the intention feature information; If the user's intention is clear, then based on the content key information, the corresponding first target candidate content is retrieved from a preset content database, and the first target candidate content is used as the target recommended content; If the user's intention is unclear, obtaining the user's historical interaction record, retrieving corresponding second target candidate content from the content database based on the historical interaction record and the content key information, and using the second target candidate content as the target recommended content; The target recommended content is recommended to a user terminal of the user.

2. The method according to claim 1, characterized in that The step of using a preset language model to parse the target input content to obtain the user's intention feature information and content key information specifically includes: Inputting the target input content into a preset language model for parsing to obtain parsed input content; Using an intention recognition algorithm to perform intention analysis on the input content to obtain intention feature information of the user; Extract key information of the input content to obtain key content information of the user.

3. The method according to claim 1, characterized in that The step of judging whether the user's intention is clear according to the intention feature information specifically includes: Obtaining an intention feature vector corresponding to the intention feature information; Using a pre-trained intent recognition model, the intent feature vector is analyzed to obtain the user's intent category and confidence score; comparing the confidence score to a preset intent clarity threshold; If the confidence score is greater than the intent clarity threshold, it is determined that the intent corresponding to the intent category is clear; If the confidence score is less than or equal to the intent clarity threshold, it is determined that the intent corresponding to the intent category is unclear.

4. The method according to claim 1, characterized in that: The step of retrieving the corresponding first target candidate content from a preset content database based on the content key information specifically includes: Constructing a first search condition according to the content key information and the intention feature information; According to the first search condition, a first candidate content set is retrieved from a preset content database using an inverted index algorithm; The similarity between each first candidate content in the first candidate content set and the target input content is calculated, and the first candidate content with the similarity higher than a preset threshold is determined as the first target candidate content.

5. The method according to claim 1, characterized in that The step of retrieving corresponding second target candidate content from the content database based on the historical interaction record and the content key information specifically includes: Based on a preset knowledge graph, semantically expand the key information of the content to obtain a key information enhanced set; constructing a second search condition based on the enhanced information in the key information enhanced set and the intention feature information; Retrieving a second candidate content set matching the second search condition from the content database; sorting the second candidate contents in the second candidate content set according to the historical interaction records to obtain a candidate content sequence; A second target candidate content is determined from the candidate content sequence.

6. The method according to claim 5, characterized in that The step of sorting the second candidate contents in the second candidate content set according to the historical interaction record to obtain a candidate content sequence specifically includes: Performing preference analysis on the historical interaction records to obtain historical preference feature information of the user; Extracting content feature information of each second candidate content in the second candidate content set; The similarity between the historical preference feature information and the content feature information is calculated, and based on the similarity, each second candidate content in the second candidate content set is sorted to obtain a candidate content sequence.

7. The method according to claim 1, characterized in that The step of recommending the target recommended content to the user terminal of the user specifically includes: Acquire device parameters of the user terminal, where the device parameters include device type and screen size; According to the device type and screen size, the format of the target recommended content is adjusted to obtain the target recommended content; The target recommended content is pushed to the user terminal for display.

8. A content recommendation device, characterized in that: include: An acquisition module is used to acquire the target input content of the user; A parsing module, used to parse the target input content using a preset language model to obtain the user's intention feature information and content key information; A judgment module, used to judge whether the user's intention is clear according to the intention feature information; A first retrieval module is used to retrieve corresponding first target candidate content from a preset content database based on the content key information if the user's intention is clear, and use the first target candidate content as target recommended content; A second retrieval module is configured to obtain the user's historical interaction record if the user's intention is unclear, retrieve the corresponding second target candidate content from the content database based on the historical interaction record and the content key information, and use the second target candidate content as the target recommended content; The recommendation module is used to recommend the target recommended content to the user terminal of the user.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the content recommendation method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the content recommendation method according to any one of claims 1 to 7 are implemented.

Images