KR20240020624A - An artificial intelligence system that recommends music and sound content that users want - Google Patents

Abstract

An artificial intelligence system for recommending music and sound content desired by a user comprises: a communication unit connected to a user terminal through wired/wireless communication; a chatbot unit for obtaining detailed information and higher information necessary for sound source recommendation; an information extraction unit for extracting detailed information from conversation information received through the chatbot unit; a sound source content storage unit for storing sound source content; a sound source classification unit for classifying the sound source content stored in the sound source content storage unit to match the detailed information; a user information storage unit; a trend analysis unit for analyzing trends suitable for purpose or mood; a sound source recommendation unit for recommending sound source content to be found by a user; and a control unit for controlling each unit. The artificial intelligence system for recommending music and sound content desired by a user according to the present invention enables the user and the system to find sound source content for a desired purpose while asking and answering various questions through a chatbot, thereby enabling the user to quickly and accurately find sound source content suitable for the desired purpose or atmosphere.

Description

An artificial intelligence system that recommends music and sound content that users want {AN ARTIFICIAL INTELLIGENCE SYSTEM THAT RECOMMENDS MUSIC AND SOUND CONTENT THAT USERS WANT}

The present invention relates to an artificial intelligence system that recommends music and sound content desired by a user. More specifically, the present invention relates to an artificial intelligence system that recommends music and sound content that a user wants, and more specifically, not only when the user finds music or sound content that suits the user's taste, but also for a specific purpose even if it is not the user's desired taste. It is about an artificial intelligence system that recommends music and sound content desired by users and can help users easily find the right music or sound content.

As information and communication technology develops, we have entered an era where one can easily access and listen to music or sound content (hereinafter referred to as 'sound content') anytime, anywhere through electronic and/or information and communication devices such as portable terminals or computers. Today, users can easily access and enjoy music content that is not stored on their terminal anytime, anywhere using devices connected to the Internet, such as computers and smartphones.

As technology has developed that allows consumers to access a variety of sound source content through the Internet or online, more sound source content has been gathered and provided in one place, broadening the range of choices for sound source content. Sound source content is not used once and then discarded, but is continuously accumulated, and in line with the era of creation, the amount of newly created sound source content is rapidly increasing day by day.

As time goes by and an increasing amount of new sound source content is created and accumulated, it has become difficult for users to select or choose the desired sound source content. To solve this problem, companies that provide music services recommend music sources that suit the user's tastes and provide services that allow users to easily select music that suits their tastes without putting in a lot of effort.

Conventional technologies related to these services include technologies such as Republic of Korea Patent No. 10-0895009 and Republic of Korea Patent No. 10-1102823. The sound source content recommendation system according to these prior technologies is provided by analyzing and classifying sound sources, analyzing the user's tastes, and recommending music that is frequently listened to or similar music. For this purpose, the user's age, gender, frequently listened to music, etc. are created in a database and analyzed to recommend music that the user frequently listens to, music similar to this music, and music liked by people with similar tastes to the user's age or gender. It is common for services to be provided.

However, although this method is advantageous in recommending music you like, it is not very helpful in finding music content for a specific purpose that is unrelated to your taste. For example, if the music you want is an exciting dance song, but you need to play music that is good to listen to at a birthday party for your parents, or if you need to play music that suits a specific mood in a place with multiple age groups, the recommended sound source is In many cases, this is not helpful at all.

And with the widespread use of video content, we have come to an era where audio content is not just used to enhance the atmosphere or enjoy, but is also widely used as background music for various video media. People working in this field are not interested in video content, but rather Even when it is necessary to select a sound source that fits the mood, the sound source recommendation system according to the prior art is not helpful or rather only hinders the use of various contents.

For example, when looking for a sound source to use as background music for an advertising video or YouTube video, find a sound source that the main target audience of the video likes or fits the mood of the video and use it as background music, not based on your own taste. It is important. However, since the sound source content recommendation system according to the prior art recommends sound sources that the person previously liked or similar sound sources, it only hinders the user from finding a sound source suitable for a specific purpose.

Also, if one ID is used for both listening to one's favorite music source and finding a specific sound source, the algorithm becomes confused and the recommendation accuracy further decreases. For example, as described above, when a person who likes exciting dance songs searches for trot songs for the elderly, the algorithm according to the prior art determines that the user likes trot and provides exciting music even when used for personal listening to music. Confusion occurs when trot songs are recommended, and this confusion not only does not suit the user's taste, but also causes inconvenience.

Therefore, there is a need for a service that does not simply analyze the user's tastes and automatically recommend music content, but provides music content that suits the user's desired purpose.

KR 10-1102823 B1 KR 10-0895009 B1

The present invention is intended to solve the above problems. It not only recommends music content that suits the user's taste by analyzing the user's subscription information and the music that the user frequently listens to, but also allows the user to listen to music that is suitable for a specific purpose. When looking for sound content, the purpose is to provide an artificial intelligence system that recommends the music and sound content desired by the user through a chatbot that can quickly and accurately find and recommend sound source content that suits the user's desired mood or purpose.

The present invention, as a technical means for solving the above problems, is an artificial intelligence system that recommends music and sound content desired by a user, including a communication unit connected to a user terminal through wired and wireless communication, and a conversation format with the user terminal through the communication unit. A chatbot unit that acquires detailed information and high-level information necessary for sound source recommendation while exchanging questions and answers, an information extraction unit that extracts detailed information from conversation information received through the chatbot unit, and a sound source content storage that stores sound source content. A sound source classification unit that classifies the sound source content stored in the sound source content storage unit to match detailed information, and a storage unit containing the user's personal information and history information about the sound source content that the user has listened to over a certain period of time. It is composed of a user information storage unit, a trend analysis unit that analyzes trends that match the purpose or mood the user is looking for, a sound source recommendation unit that recommends the sound source content the user is looking for, and a control unit that controls each of the above units, When the user answers a question about detailed information related to the sound source content that the user wants to find in the chatbot unit, the information extraction unit extracts information about the detailed information from the answers, and the sound source recommendation unit sends the information to the information extraction unit. It is characterized by extracting sound source content that matches the detailed information extracted and providing it to the user.

In a preferred embodiment of the present invention, the information extraction unit determines information that matches the detailed information among the user's answers as confirmed information, judges the answer related to the detailed information as estimated information, and preferentially asks questions about the estimated information. After obtaining confirmed information, questions regarding other detailed information are sequentially asked through the chatbot unit, and the chatbot unit 1) utilizes user information among the sound source contents that match the detailed information acquired at each stage while talking with the user. 2) candidate songs that suit the user's taste, 2) candidate songs that other people listen to and prefer most through trend analysis, and 3) a sample list of candidate songs analyzed by the algorithm, and receive user evaluations to find the candidate song that the user is looking for. It is characterized by determining the direction of sound source content.

In a preferred embodiment of the present invention, when a new word that is not stored in detailed information is detected, the information extraction unit presents detailed information related to the new word to the user so that the user selects at least one or more words. The selected detailed information and search results related to the new word are stored together with the word to determine the meaning of the new word. When searches using the new word exceed a certain number of times, the new word is added to the detailed information category. It is characterized by adding.

A method of recommending desired music and sound content according to the present invention includes the steps of obtaining basic information through a chatbot; Step of obtaining additional information through chatbot; Step of obtaining high-level information through chatbot; A sample music provision step of providing sample sound source content that matches the obtained detailed information and higher level information; and a weight adjustment step.

In a preferred embodiment of the present invention, the 'basic information acquisition step through a chatbot' is characterized in that the user receives one specific information among a plurality of detailed information, and the 'additional information acquisition step through a chatbot' If the detailed information received and acquired in the 'basic information acquisition stage' is confirmed information, questions will be asked about the remaining detailed information sequentially or in a related order, excluding questions related to the detailed information already obtained. When asking a question related to the remaining detailed information, a partial list of the sound source content group specified by the detailed information obtained in the previous step is presented on the user's screen, and all or part of the list can be streamed as background music and listened to. It is characterized in that it allows the user to evaluate the list of the sound source content group.

In a preferred embodiment of the present invention, the sound source content back group presented to the user is a sound source content candidate group extracted based on the detailed information obtained in the previous step: 1) candidate songs that suit the user's taste, 2) other people through trend analysis. 3) This preferred candidate group is provided by classifying it into at least two of the candidate songs selected through algorithm analysis, and when an evaluated answer is received from the user, it is analyzed and a sound source content candidate group is created based on the detailed information obtained at a later stage. When providing, all or part of the recommended sound source content candidates are provided as sound source content of the user's preferred candidate group.

In a preferred embodiment of the present invention, the 'weight adjustment step' allows the user to re-search the sound source content candidates by changing the detailed information or higher information of the results provided in the 'sample music provision step', and among the detailed information Emotion information can be configured to re-search sound source content candidates by adjusting the intensity.

The artificial intelligence system that recommends the music and sound content desired by the user according to the present invention allows the user and the system to find sound source content for the desired purpose by asking and answering various questions through a chatbot, thereby providing a sound source suitable for the user's desired purpose or mood. It has the effect of allowing you to find content quickly and accurately.

1 is a diagram showing the configuration of an artificial intelligence system that recommends music and sound content desired by a user according to the present invention.
Figure 2 is a diagram showing a method of providing recommended sound sources using an artificial intelligence system that recommends music and sound content desired by a user according to the present invention.
Figure 3 is a diagram schematically illustrating the process of finding recommended sound sources in a tree in an artificial intelligence system that recommends music and sound content desired by a user according to the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Figure 1 is a diagram showing the configuration of an artificial intelligence system that recommends music and sound content desired by a user according to the present invention.

Referring to Figure 1, the artificial intelligence system that recommends music and sound content desired by the user according to the present invention includes a control unit 100, a communication unit 110 connected to the control unit 100, a chatbot unit 120, and information extraction. It is composed of a unit 130, a sound source classification unit 140, a sound source content storage unit 150, a user information storage unit 160, a trend analysis unit 170, a sound source recommendation unit 180, etc.

Music and sound content (hereinafter referred to as 'sound content') referred to in the present invention is a concept that includes all sound source content that has been created or can be created with digital sound sources, such as music and melodies as well as sounds of nature.

The communication unit 110 is connected to the user terminal 10 through wired or wireless communication and receives information necessary for recommending sound content, or transmits chatbot communication and/or sound source content to collect the information as a file or streaming. Do it. In addition, if necessary, if the sound source content is stored on another server or the sound source classification unit 140 and/or trend analysis unit 170 are provided on a separate server, the server where the sound source content is stored or It may also function to communicate with a server equipped with a sound source classification unit 140, etc.

The chatbot unit 120 functions to collect information necessary for music source recommendation while conversing with the user terminal 10 in a chatting format through the communication unit 110. Chatting by the chatbot unit 120 is not conducted by humans but by a preset artificial intelligence algorithm. Therefore, questions can be asked to the user according to the priority order according to a preset algorithm, and if the user sends the necessary information first through chat regardless of the preset order, additional questions can be asked through chat about the information not obtained from the received information. It plays a role in acquiring information. The method of obtaining information necessary for music source recommendation through the chatbot unit 120 will be described in detail later.

The information extraction unit 130 serves to extract information necessary for music source recommendation from the conversation conducted through the chatbot unit 120. The information extracted by the information extraction unit 130 may consist of 'confirmed information' and 'estimated information', and the chatbot unit 120 is configured to ask various questions in order to obtain as much confirmed information as possible. Here, 'confirmed information' refers to information that conforms to detailed information or higher-level information to be explained later, and 'estimated information' refers to information that does not conform to detailed information or higher-level information but is related to the above information. Other information or words are judged as unnecessary information to extract sound source content.

Because information collection through chatbots is done in a conversation format that users can feel comfortable with, information that is unnecessary for music recommendation may often be received. In a conversation through a chatbot, when the user answers to the question 'What kind of music are you looking for?' by saying 'I am looking for music (melody) with a bright (exciting) atmosphere like today's weather,' the information extraction unit 130 says ' Emotional information such as 'bright (exciting) atmosphere' and 'music' are extracted, judged as 'confirmed information', and used as information needed to extract the necessary sound source content. However, the user simply 'looks for music that suits today's weather' If you answer ', there is no 'confirmed information' required for music content recommendation, so in order to obtain the necessary confirmed information, you can ask an additional question, 'What is the weather like today?'. At this time, a confirmation that matches the detailed information among the user's answers is selected. Although there is no information, there is a mention of 'weather' related to the detailed information, so the 'weather' is judged to be estimated information and questions related to the estimated information are given priority, and through questions about the estimated information, confirmed information ( Once weather information is obtained, questions about other detailed information are sequentially asked.

In order to make the conversation more realistic, the chatbot unit (120) determines the customer's location information as necessary, checks the weather information for the area, and then says, 'There is a weather forecast that says the weather in the OO region is clear today, and the current weather is You can also ask questions of the form ‘Is it sunny?’ The user's location information can be easily determined when signing up to use the artificial intelligence system that recommends the music and sound content the user wants according to the present invention or by agreeing to check the location information during use. Information will be available through the Korea Meteorological Administration. Since the above-mentioned location information and weather information (estimated information) may differ from the actual situation, the user is asked questions to confirm, and recommended sound sources are extracted according to the user's answers (confirmed information). However, if the user does not have an additional answer for more than a certain period of time, recommended sound sources can be presented using the above estimated information. In addition to the above information, the information extraction unit 130 may also recommend a sound source that suits the user's taste by utilizing the user information stored in the user information storage unit 160 and the records (history information) that the user has listened to so far. In addition, through the trend analysis unit 170, music content that people of similar tendencies, gender, and age recently like can be recommended.

The artificial intelligence system that recommends the music and sound content desired by the user according to the present invention can be used in the same way as above when the user is looking for sound source content that suits the user's taste, but it is a sound source to be used for a specific purpose rather than the user's taste. Its characteristic is that it can be usefully used when searching for content.

For example, a user who mainly enjoys exciting dance songs can use it to find music suitable for grandmother's 60th birthday party, or to find background music needed to work on presentation documents or videos. In the case of music for this specific purpose, the user's taste is important, but it is more important to find sound source content suitable for the task in question, or the preference of the person listening to the sound source content is important. Therefore, as in the prior art, the user's taste or what he or she usually listens to is important. It is difficult to find sound source content that meets the purpose by relying solely on records. Rather, records and data such as user tastes are often an obstacle to finding accurate sound content that is objective and meets the purpose. Therefore, when searching for sound content for a specific purpose, new settings and conditions are needed that are less or less influenced by the user's tastes and suit the purpose, and the present invention can be usefully utilized in such cases.

Detailed information obtained through the chatbot unit 120 may include emotional information, place information, scene (mood) information, season (date) information, time information, era information, etc., as shown in FIG. 3. . However, the above detailed information represents an example for implementing the present invention, and various detailed information may be added as needed. And each detailed information category may include specific information corresponding to the detailed information. For example, as shown in Figure 3, 'emotional information' may include specific information such as 'scary, solemn, grotesque...', and 'place information' may include 'home, school, beach.' Specific information such as '...' may be included. In the same way as above, each category corresponding to detailed information includes specific information corresponding to a sub-concept. The functions of the chatbot unit 120 and the information extraction unit 130 will be described in more detail later.

Before providing sound source content to a user, the sound source classification unit 140 classifies sound sources stored in the sound source content storage unit 150 by genre or mood, and classifies and stores them in the sound source content storage unit 150. The criteria for classifying sound source content in the sound source classification unit 140 may be in a direction that matches the detailed information shown in FIG. 3. In other words, by adding indexes or weights related to emotional information, place information, scene information, etc. to all sound source content and displaying the measured value or score in the index, it is possible to have an evaluation value that matches the detailed information. Can be classified. Known methods can be applied to classify sound content by genre or mood, and in some cases, previously classified information may be used.

The user information storage unit 160 stores personal information stored when a user signs up to use the system for recommending desired music and sound content according to the present invention, and personal information stored by the user for a recent period of time using the system according to the present invention. History information about sound source content listened to may be included. Accordingly, the information extraction unit 130 can use the information to extract sound source content that suits the user's taste. Recommendation of sound source content using user information is mainly used to help users find music for personal listening, and it is desirable not to use it to find sound source content for a specific purpose, regardless of the user's taste, or to add low utilization. However, no matter how much sound source content is searched for a specific purpose, the individual tastes of the user or worker are bound to be reflected to some extent, and since users will also like sound source content that matches their tastes, even in searches for a specific purpose, the user's You can ensure that your tastes are reflected to some extent. The method of recommending sound content using user information and how to utilize it will be explained later.

The trend analysis unit 170 serves to analyze and utilize recent trends when extracting sound source content that suits the purpose or atmosphere the user is looking for. For example, if a user is looking for an 'atmospheric ballad', the chatbot is used to analyze the purpose of finding the music. If the purpose is to listen personally, music that the user usually enjoys or similar music is recommended. . Also, when recommending music content, you can recommend all or part of the music that people of the same taste, age group, or gender group recently enjoyed listening to through the trend review section.

However, if you are looking for sound source content for a specific purpose rather than for personal listening, it may be useful to find and recommend sound source content that meets that purpose through trend analysis. For example, in the case mentioned above, when a young person in his 20s is looking for music for a grandmother in her 80s, and the purpose of the search is 'music to play to my grandmother' through a chatbot, the chatbot searches for the grandmother's age and gender. By asking questions, you can analyze and recommend the music that seniors of the relevant age and gender listen to a lot recently.

To this end, the trend analysis unit 170 serves to analyze recent trends by utilizing not only the data of other users using the service, but also the announced music source rankings and/or the number of broadcasts or downloads provided on other service providers' platforms, etc. .

The sound source recommendation unit 180 comprehensively analyzes the information (detailed information and top information), user information, and trend information acquired through the chatbot described above, and ultimately recommends the sound source content that the user is looking for. Users can choose to limit the number of recommended contents, and can also improve accuracy by evaluating their preferences among recommended sound sources.

Figure 2 is a diagram showing a method of providing recommended sound sources using an artificial intelligence system for recommending music and sound content desired by a user according to the present invention, and Figure 3 is a diagram showing a method of recommending music and sound content desired by the user according to the present invention. The process of finding recommended sound sources in an artificial intelligence system is outlined as a tree.

The artificial intelligence system that recommends music and sound content desired by the user according to the present invention is divided into two modes when recommending sound content. One is purely a mode (taste mode) where the user searches for music he or she likes or similar music to listen to personally, and the other is a mode where the user experiences new music that is different from the user's usual tastes or is unrelated to the user's tastes. This is a mode to find sound source content needed for a specific purpose (specific purpose mode). The present invention is particularly useful for application to the 'specific purpose mode', and in the case of the 'taste mode' where the user searches for music content he or she likes to personally listen to, similar to the prior art, the type of music the user usually listens to is used. While analyzing and providing recommended sound sources, the present invention can be used as an auxiliary method of finding new music using the chatbot unit 120 or recommending music that may be liked through trend analysis. Therefore, the following description will focus on how to utilize the present invention in 'specific purpose mode', and will explain the case where the present invention can be applied in 'taste mode' as an example.

Whether the purpose of using the service according to the present invention is preference mode or specific purpose mode, the user's purpose can be confirmed by having the user select in advance or by asking a question through a chatbot. In addition, the results of a user's search for sound source content for a specific purpose will not be reflected or affected when the user searches in preference mode later, so that accurate sound source content that better matches the user's purpose can be recommended.

Looking at the method of using the system for recommending sound source content desired by the user according to the present invention with reference to FIGS. 2 and 3 and the configuration of FIG. 1 described above, the present invention includes a mode selection step (S200), basic information through a chatbot, Acquisition step (S210), additional information acquisition step through chatbot (S220), upper level information acquisition step through chatbot (S230), providing samples of recommended sound source content (S240), step of adjusting the weight of recommended sound source content (S250) and a step of selecting and providing a sound source that meets the purpose (S260).

In the 'mode selection step (S200)', the user determines the purpose for which he or she is searching for sound source content. In other words, the user determines whether the purpose of using the service is 'preference mode' or 'specific purpose mode', which can be selected or set in advance by the user as described above, or the chatbot asks the user a question and makes the choice or decides. You can do it.

In the 'step of acquiring basic information through a chatbot (S210)', the user delivers basic information about the sound source content he or she is looking for through a chatbot. The information about the sound source content refers to information including at least one specific information belonging to the plurality of detailed information described above. To this end, the chatbot unit 120 asks 'What kind of music are you looking for?' the moment the chat window opens. Ask questions to start a conversation, such as: When basic information is received from the user through a chatbot, the information extraction unit 130 extracts the necessary information from the information sent by the user and asks additional questions, or additional information is required after determining whether the information is sufficient to recommend sound source content. If so, ask additional questions.

If users frequently utilize the system for recommending sound source content related to the present invention, they will be able to identify which information provides the most useful results. Therefore, as the frequency of use increases, more accurate information will be provided. In this way, if sufficient and accurate information required for music source recommendation is provided, additional questions may be reduced or, in some cases, unnecessary. However, since it is generally difficult to provide accurate recommended sound source content with basic information, the chatbot unit 120 attempts conversations to obtain various additional information.

The 'step of acquiring additional information through chatbot (S220)' is a step of asking questions about information that is useful or necessary for recommending accurate sound source content among the information that was not obtained as basic information.

In Figure 3, information of the sound source content recommendation system according to the present invention is illustrated in a tree format. In Figure 3, it can be seen that various detailed information for recommending sound source content and genre, instrument, etc., which are higher level information of the detailed information, are included. In addition to genres and instruments, high-level information may include information by country and language, and of course, other additional information may be included as needed. And of course, various details can be added as needed. The basic information and additional information provided by the user through the chatbot generally correspond to detailed information, and the chatbot unit 120 also mainly proceeds in the direction of acquiring detailed information in the basic information acquisition stage and the additional information acquisition stage. do. However, if the user provides the top information as basic information in advance, the top information is limited in advance during the process of extracting the recommended sound source in the next step. At this time, the chatbot unit checks whether the user wants to limit the top information in advance, and if the user wants to specify and limit the top information, it limits the top information in the process of extracting the recommended sound source in a later step. If not, In the final stage, the upper level information can be restricted.

For example, if the user provides 'fun pop songs that are good to listen to at an amusement park' as basic information, the chatbot unit 120 asks whether to limit the top information to 'pop songs' and replies that the user limits it to pop songs. If there is a response, in the subsequent step of extracting the recommended sound source, the top information is fixed as a pop song and the recommended sound source is extracted. However, if the top information is not determined in advance, recommended sound sources of various genres are extracted and provided, and the user can specify the desired top information by viewing or listening to samples of the recommended sound sources. Therefore, in order to utilize a wider range of diverse information, it is desirable not to specify in advance the upper information, which is the top of the tree. If the higher information is predetermined, faster results can be obtained by limiting the upper information at the beginning. There will be. The main purpose of the present invention is to help users simply find sound source content that is good to listen to in any situation or time without knowing exactly what sound source content they want, so even in the process of searching or recommending sound source content, the tree structure of FIG. 3 is used. It is desirable to narrow the scope by going from the bottom (detailed information) to the top (higher information).

In the 'additional information acquisition step through chatbot (S220)', after the chatbot unit 120 goes through the basic information acquisition step described above, it asks additional questions to obtain additional detailed information from the tree in FIG. 3. This is the step. At this time, in the additional information acquisition step, if the questions regarding the basic information already obtained in the above-described 'basic information acquisition step through chatbot' are confirmed information, the remaining detailed information is asked except for the questions related to the confirmed detailed information. Ask questions sequentially or in order of relevance.

For example, when a user provides basic information such as ‘Recommend music that is good to listen to while taking a walk at the beach,’ the information extraction unit 130 converts the place information ‘beach’ and the scene information ‘walk’ into confirmed details. Recognize and extract 'primary sound source content candidates' that can be candidates. The user is then asked for additional details, such as 'What kind of emotion (mood) are you looking for in a song?' And if the user answers 'peaceful atmosphere', sound source content matching the peaceful atmosphere is extracted from the first sound source content candidate group (second sound source content candidate group).

After the secondary sound source content candidates are extracted, questions regarding other detailed information can be asked sequentially and repeatedly. In other words, after the secondary sound source content candidates are extracted, the question 'What season are you looking for music that suits?' or 'What kind of day are you looking for music that suits?' After asking a question and obtaining season information from the answer regarding the season information, a tertiary sound source content candidate group matching the season information is extracted from the secondary sound content content group.

As described above, when a user searches for sound source content to personally enjoy music, recommended sound sources are provided using personal information stored in the user information storage unit 160, current location information, weather information, etc. However, when searching for sound content for a specific purpose, it is not related to the current weather or location, but rather requires specific weather information for that purpose, so you can ask about weather information through a chatbot.

Acquiring information about the detailed information and extracting the sound source content candidates as described above is carried out repeatedly until the extracted sound source content reaches a certain number or less, or until all questions about the preset detailed information are completed, or until the user asks no more questions. The process continues until no information is provided or a recommendation is made without further information, and the nth sound source content candidates are finally extracted and presented to the user.

As described above, if you use the method of reducing the candidate group as each step progresses, the search will proceed in a direction that suits your purpose by viewing the list of recommended sound sources extracted at each step or listening to samples during the conversation through the chatbot. It is easy to check whether something exists and to use functions such as going back to the previous step. However, of course, if necessary, a method of extracting the final recommended sound source after combining all detailed information and high-level information may be used.

Meanwhile, detailed information can be added by the system administrator as needed, but the algorithm can also add it automatically. In language, various buzzwords and abbreviations appear depending on the era, but it is difficult for managers to check and add all newly created buzzwords or abbreviations one by one. However, users who frequently use these languages may also use these languages when conversing with a chatbot. Therefore, when a new expression that is not included in the existing detailed information is received, the information extraction unit 130 detects it, indexes it, and stores it separately. Additionally, additional detailed information obtained through conversation with a user who searches using the word and the user's final selection result are reflected and stored along with the newly introduced word. After that, when the number of searches using the newly introduced detailed information exceeds a certain number, and the detailed information sent by the user through the chatbot and the user's selection (satisfaction rate) of the sound source recommended by the algorithm exceed a certain level, the word is selected. Add it to the details and make it available to other users.

For example, when a user asks a question through a chatbot, ‘Recommend a good song to listen to while walking on the beach,’ the word ‘Kalsahan’ is a new song that is not stored in the existing detailed information, as shown in Figure 3. If it is information, the chatbot can ask additional questions about what 'kalsamhan' is. Today, algorithms have become more advanced and the ability to recognize the meaning of language or text has increased, and artificial intelligence algorithms continue to learn newly used terms, so in most or significant cases, the chatbot unit 120 and information extraction The unit 130 can recognize that the word 'kalsahan' represents emotional information, and in many cases, can understand its meaning. However, in order to confirm more accurate information or in the case of a completely new word, the chatbot unit 120 may ask additional questions about the meaning of the word 'kalsamhan' through chat. However, since the exact meaning cannot be determined through these questions, repeatedly asking questions may cause inconvenience to the user, so rather than asking repetitive questions, the chatbot unit 120 presents words that are close to the literally recognized meaning. Move to the next step by resolving uncertainty by presenting good emotional information. That is, when the word 'kalsahan' is received by the chatbot unit 120, a simple question is asked to determine whether the word is related to emotional information or scene information, and if it is related to 'emotional information', the emotional information is asked. Present all or part of the relevant details (calm, happy, lyrical, etc.) so that at least one piece of similar emotional information can be selected. Since the new word may be an existing word that expresses the same emotion, or there may be a word that applies a combination of existing emotions, the user may select multiple details.

In this way, when new detailed information is introduced, the category and meaning of the detailed information are determined through a simple question, and the atmosphere of the sound source content finally selected by the user by searching for the word is combined and stored. And even when other users search with the same word, the same process is repeated a certain number of times, and if the user performs the search above a certain level without trial and error, the word 'Kalsahan' is added to the category of emotional information and presented or utilized. You can. In other words, by analyzing the search behavior of multiple users, if users mainly select search results corresponding to the meaning of 'calm (70%) and pleasant (30%)' when searching for the word 'calamhan', the above ' The detailed information 'Kkalsamhan' is judged to mean that calmness and pleasure have a ratio of 7:3, and recommended sound sources are presented in the same ratio or by recommending music with a relatively pleasant atmosphere among calm music. . By doing this, even when an uncertain new word enters the detailed information, it is possible to proceed smoothly to the next step under certain conditions. When the uncertain word is repeated a certain number of times, the algorithm determines what it means and adds it to the category for more accurate use. It is possible to recommend music content that fits the user’s needs. If the meaning of a newly introduced word through the above process shows a match rate exceeding a certain value (e.g. 80% or 90%) with existing specific detailed information, the word is combined with the existing detailed information and has the same meaning. It can be used as

When the chatbot unit 120 asks a question to obtain detailed information, candidate answers may be presented in advance to make it easier for the user to make a selection. For example, while asking a question about additional detailed information through the chatbot unit 120, ‘What kind of emotional (mood) song are you looking for?’, among the detailed information in FIG. By providing selection options such as ', grotesque, dreamy...' through a chatbot, the user can select the desired atmosphere from among the above selection options. At this time, if there are many selection options presented, all or part of the selection options may be provided according to priorities analyzed in advance through an artificial intelligence algorithm.

In general, when taking a walk on the beach, emotional information such as 'calm, romantic, happy, peaceful...' is likely to be selected, and emotional information such as 'spicy, scary...' is likely to be selected. low. Therefore, when presenting a selection option, at least emotional information such as 'splenic' or 'scary' could be presented as a lower priority or omitted. However, of course, users can still convey emotional information (e.g. 'scary', etc.) that is not presented in the selection options through the chatbot. In addition, the artificial intelligence algorithm learns the priority of selection options through the number of times users select them or through trend analysis, and can present additional detailed information options that users select most often from the same or similar selection options.

Meanwhile, the artificial intelligence system that recommends music and sound content desired by the user according to the present invention uses the sound source specified in the previous step when conversing through a chatbot in the step of acquiring the basic information and the step of acquiring additional information described above. A partial list of content candidates can be presented on the user's screen, and all or part of the list can be streamed and listened to as background music. Accordingly, a process may be further provided in which the user selects some songs from the provided list or evaluates sound source content played as background music. If there is a song that the user wants to hear among the provided list, he or she can select the song from the list and listen to it.

For example, 'a song that is good to listen to while walking on the beach' is obtained as basic information and the first sound source content candidate group is extracted, and in the next step, the chatbot unit 120 selects 'a certain atmosphere (emotion)' to obtain additional information. By asking the question, 'Are you looking for music?', a list of primary sound source content candidates is provided to the user's chat window using the basic information that is a good song to listen to while walking on the beach obtained in the basic information acquisition step, and a specific song from the list is provided. Songs can be played sequentially or according to the user's selection. Playback of sound source content consists of only part of the song's introduction or climax and continues until the user sends a response to additional information or stops playback. Users can select or evaluate one or two good songs from the list of primary sound content candidates provided before or after answering the chatbot's questions.

The list of primary sound source content candidates provided by the sound source recommendation unit 180 does not provide all lists of primary sound source content candidates extracted in the basic information acquisition step, but only provides a partial sample list. In addition, the sample list is characterized by randomly providing a plurality of types among the extracted primary sound source content candidates. For example, when extracting the primary sound source content candidates, 1) user information stored in the user information storage unit 160 is used to select candidate songs that suit the user's taste and 2) trend analysis regardless of the user's taste. Through unit 170, recent trends are analyzed, and 3) candidate songs are most listened to and preferred by other people under similar conditions, and 3) candidate songs analyzed purely by the algorithm are classified into at least two categories and provided at random. Then, the user can evaluate only the provided list (title) or listen to all the sound content in the list and evaluate. The evaluation can be done in a simple and quick way, such as high, middle, or low, or by selecting a few songs. And it is desirable that the list of candidates provided be comprised of a relatively small number, such as 3 to 6 songs. This is because the candidate list is not a recommendation song in the true sense, but a kind of test to more accurately determine the direction the user is looking for, and if a large number of songs are presented, there is a risk that the search will take a long time or cause confusion for the user, so there are relatively few songs. This is using samples.

The candidate songs 1) to 3) above may be provided simultaneously during the chatting process, or some of them may be provided sequentially. In other words, all three samples can be provided during each stage of the chat to confirm the direction the user prefers. The first provides candidate songs that suit the user's taste, and the second provides candidate songs through trend analysis. It may be done in a way such as providing .

Through the above method, the direction of the song the user is currently looking for can be more accurately inferred. For example, if the songs selected by the user are similar to the user's usual tastes, when extracting sound source content candidates at a later stage, the user's tastes are taken into consideration and the extracted sound sources are used, or at least the proportion of songs that the user is likely to like is increased. Proceed in the direction. And, if the songs selected by the user have a large number of sound source content groups extracted through trend analysis, the subsequent steps are made to consist of songs that are more consistent with the trend analysis or to increase the proportion of these songs. Candidate songs analyzed purely by the algorithm can be composed of sound source content analyzed based on tempo or representative melody by analyzing sound source data, and while proceeding with the above method, the algorithm analysis is supplemented using the user's selection results. As time goes by, it can be improved in a way that helps in classifying sound sources or selecting recommended songs, and also provides users with the opportunity to experience a variety of sound content other than their own favorite songs and songs that others like. can do. When providing a list of sound source content candidates at each stage in the sound source recommendation section, it would be helpful for a more objective search if it did not indicate exactly what criteria the sound source content was extracted by. However, after the selection is completed, you may be notified by what criteria the sound source content is classified, providing an opportunity to check the sound source content according to other standards.

For example, after detailed information (basic information) is obtained regarding the first question, a list of 2 to 6 sound source content candidates is presented, and when the user selects a specific song from the list of sound source content candidates, ' You can display a message such as ‘This is a song similar to OO’s taste’ or ‘This is a song that fits the recent trend.’ If the selected list contains songs similar to the user's tastes, the user can adjust the direction by reselecting songs that fit the trend. In the system according to the present invention, no matter how objective the information is and the recommended sound source content candidates are presented, there is a high possibility that anyone will select a song that best suits their taste. Therefore, even in the process of providing and selecting a list of sound source content candidates as described above, the user has some influence in choosing to proceed in a direction that suits the user's taste or proceed in a direction that matches the trend. It is to make it possible.

In another embodiment, at each stage of obtaining detailed information, it does not tell what category the selected sound source content corresponds to, but when all detailed information is obtained and the recommended sound source content is confirmed and presented, it informs what category the recommended sound source content falls into. You can also view recommended music content that falls into other categories. This method is a type of weight adjustment, and will be further explained in the weight adjustment step, which will be described later.

The 'higher level information acquisition step through chatbot (S230)' is the step of applying the higher level information desired by the user among the extracted sound content. As shown in FIG. 3, the top information may include information about genres, information about musical instruments, etc. In addition, top information by country or language may be included. And in some cases, higher level information such as sounds of nature or birds may be included in addition to music.

The upper information acquisition step (S230) may be performed by providing selection options for higher information on the user's screen after the chatbot unit 120 completely specifies the detailed information. Alternatively, as described above, the user may select high-level information in advance or after specifying detailed information during a conversation with the chatbot according to the chatbot's suggestions.

When high-level information is specified in the high-level information acquisition stage, only recommended sound source content matching the high-level information is left and provided to the user in the sample sound source content provision stage, which will be described later.

In the 'step of providing samples of recommended sound source content (S240)', the top information is specified in the above-described top information acquisition step, and after specifying a song matching the top information among the candidate songs extracted through detailed information, the extracted song is provided to the user. It provides a list of recommended songs and sample sound source content where you can listen to the corresponding sound source content.

Recommended songs provided to the user are provided by sequentially listing a preset number of candidate songs or, if the user does not specify the number, a specified number in advance. Users can change the number of recommended songs at any time. Additionally, the sample sound content provided may provide some, but not all, of relatively low-quality sound sources, or may be equipped with a recording prevention function or a function to prevent copying or transmission to other devices.

Users can purchase high-quality sound sources by listening to the sample sound source contents provided in the 'sample providing step of recommended sound source content (S240)' and purchasing songs they like. However, if there is no song you like among the sound sources provided in the sample sound source content, you can change the options for the detailed information or higher level information made in the previous step or adjust the weight (step of adjusting the weight of the recommended sound source content, S250). Through the 'step of adjusting the weight of recommended sound source content (S250)', a more precise search is possible by adjusting the weight of detailed information.

In other words, through the process of acquiring basic information and additional information through a chatbot, the detailed information and high-level information are determined as 'a pleasant pop song that is good to listen to while walking on the beach in the morning', and the selected songs are extracted based on the user's favorite tastes. If you cannot find a song that meets your desired purpose in the sample sound content or there is no song you like, you can obtain new recommendation results by adjusting the weight with the sound source content according to the detailed information, top information, and/or trend analysis results. there is.

For example, the intensity of emotional information among detailed information can be adjusted using the 'strong, medium, weak' or 'high, medium, low' method. If the detailed information is 'fun', you can change it to other emotional information such as 'funny' or 'touching', but among 'fun', you can adjust the range of recommended songs by using methods such as 'strong' or 'weak' or 'high, medium, low', etc. Or you can change it. At this time, if there is a weight adjustment of detailed information, the algorithm can be done using values that analyze the tempo, beat, frequency, etc. of the sound source. As an algorithm for analyzing the tempo of a sound source, for example, known technologies such as Republic of Korea Patent No. 10-0895009 or similar technologies may be used.

You can also receive other recommended songs by adjusting 'morning' to 'day (afternoon)'. Adjusting the weight of sound content can also be done at the stage of providing detailed information. However, if you adjust the weight information while talking through a chatbot, the search process will become overly complicated, and unless you completely understand both the song and the algorithm for the mood you want, it will actually be a hindrance to the user. However, after going through all the steps, receiving sample music and listening to it, it will be easier to adjust the weight and be able to set the direction accurately.

If the recommended song is determined to be less exciting than expected by combining detailed information and top information, such as 'a fun pop song that is good to listen to while walking on the beach in the morning', change the time zone information to 'day (afternoon)' or 'fun'. It can be easily predicted that if you increase the intensity, a more exciting song will come out. And if you want a song that is a little quieter than the recommended song, you can lower the intensity of 'Happy', change the scene information to 'Meditation' instead of walking, or select dawn or night instead of morning or day in the time zone information to enjoy more atmospheric music. It will be easy to predict what will be recommended.

In addition, in the step of providing a sample of recommended sound source content (S240), it is possible to inform the user whether the recommended list is a result close to the user's taste or according to a trend, and allow the user to select a recommended list of a different category. For example, in the step of obtaining detailed information, if a list of sound source content candidates is provided and the user selects or evaluates a song they like, in most cases, sound source content that is close to one's taste will be selected. Therefore, in most cases, it is highly likely that the recommended music content that is finally presented will be music content that is close to one's taste. In this case, the user can receive or check a list or sample of sound source content close to the trend in the weight adjustment step (S250). In this way, you can check the differences between your own tastes and provide an opportunity to select new songs that fit the trend.

After adjusting the weights through this method, sample music can be provided again and the steps described above can be repeated. In this way, by repeating the sample providing step (S240) and the weight adjustment step (S250) several times, you will be able to find the song you want more accurately or in a variety of ways.

In the step of adjusting the weight (S250), only information or items set in the detailed information can be selected, and intensity adjustment is also possible only in preset items. For example, it is possible to adjust the intensity of emotional information, but it is not possible to adjust the intensity of location information or time information. Therefore, based on Figure 3, intensity adjustment will be mainly done in detailed information corresponding to 'emotional information'. Therefore, in the weight adjustment step (S250), the strength and weakness will be adjusted mainly for emotional information, and other information will be changed. However, this only represents one embodiment of the present invention, and it is of course possible to apply it to other detailed information as long as it is detailed information that can adjust the weight to strength or weakness.

In the same way as above, users can check the recommended sound source list with various detailed information and select sound source content that suits their purpose.

As described above, in the detailed description of the present invention, specific embodiments have been described, but of course, various modifications are possible without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the claims and equivalents thereof as well as the claims described below.

100: control unit
110: Department of Communications
120: Chatbot Department
130: Information extraction department
140: Sound source classification unit
150: Sound source content storage unit
160: User information storage unit
170: Trend analysis department
180: Music recommendation section

Claims (7)

In an artificial intelligence system that recommends music and sound content desired by the user,
A communication unit connected to the user terminal through wired or wireless communication, a chatbot unit that exchanges questions and answers with the user terminal in a conversational format through the communication unit, and obtains detailed information and higher information necessary for sound source recommendation, and a conversation received through the chatbot unit. An information extraction unit that extracts detailed information from the information, a sound source content storage unit that stores sound source content, a sound source classification unit that classifies the sound source content stored in the sound source content storage unit to match the detailed information, and the user's personal information. A user information storage unit that stores information including information and history information about sound source content that the user has listened to over a certain period of time, a trend analysis unit that analyzes trends that match the purpose or mood the user is looking for, and a sound source that the user wants to find. It is composed of a sound source recommendation unit that recommends content and a control unit that controls each unit,
When the user answers a question about detailed information related to the sound source content that the user wants to find in the chatbot unit, the information extraction unit extracts information about the detailed information from the answers, and the sound source recommendation unit sends the information to the information extraction unit. An artificial intelligence system that recommends music and sound content desired by the user, characterized in that it extracts sound source content that matches the details extracted and provides it to the user.
In claim 1,
The information extraction unit determines the information that matches the detailed information among the user's answers as confirmed information, judges the answers related to the detailed information as estimated information, and preferentially asks questions about the estimated information to obtain confirmed information. Questions regarding other details are sequentially asked through the chatbot.
The chatbot unit selects 1) candidate songs that match the user's tastes by utilizing user information and 2) candidates that other people listen to and prefer a lot through trend analysis among the sound source contents that match the detailed information obtained at each stage while talking with the user. 3) Artificial intelligence that recommends music and sound content desired by the user, characterized by determining the direction of the sound source content the user is looking for by providing a sample list of candidate songs analyzed by the algorithm and receiving the user's evaluation system.
In claim 2,
When a new word that is not stored in detailed information is detected, the information extraction unit presents detailed information related to the new word to the user, allowing the user to select at least one or more words, and combines the selected detailed information with the A user stores search results related to a new word together with the word to determine the meaning of the new word, and adds the new word to the detailed information category when searches using the new word exceed a certain number of times. An artificial intelligence system that recommends desired music and sound content.

Basic information acquisition step through chatbot;
Step of obtaining additional information through chatbot;
Step of obtaining high-level information through chatbot;
A sample music provision step of providing sample sound source content that matches the obtained detailed information and higher level information; and
A method of recommending music and sound content desired by the user, including a weight adjustment step.
In claim 4,
The 'basic information acquisition step through chatbot' is characterized by receiving one specific information among a plurality of detailed information from the user,
In the 'additional information acquisition step through chatbot', if the detailed information received and acquired in the 'basic information acquisition step' is confirmed information, additional questions are asked regarding the remaining detailed information, excluding questions related to the detailed information already obtained. Proceeds and allows the user to enter additional details to receive,
When asking a question related to the additional detailed information, a partial list of the sound source content group specified by the detailed information obtained in the previous step is presented on the user's screen, and all or part of the list is streamed as background music. A method of recommending music and sound content desired by a user, characterized in that it allows the user to listen and evaluate the list of the latter group of sound source contents.
In claim 4,
The sound source content candidate group presented to the user is a sound source content candidate group extracted based on the detailed information obtained in the previous step. 1) candidate songs that suit the user's taste, 2) candidate songs preferred by other people through trend analysis, and 3) algorithm analysis. Among the candidate songs selected through , we categorize them into at least two categories and provide them.
When an evaluated answer is received from the user, it is analyzed and a sound source content candidate group is provided based on the detailed information obtained at a later stage. The sound source content is provided according to the results of the user's selection or evaluation of all or part of the recommended sound source content candidate group. A method of recommending music and sound content desired by a user, characterized by providing a Tsusubo group.
In claim 6,
The ‘weight adjustment step’ is,
In the results provided in the 'sample music provision step', when information that changes detailed information or higher information is received from the user, a new sound source content candidate group is extracted and provided according to the changed information. Music and sound content desired by the user is provided. Recommended method.

Images