TWI723782B - Method for generating personalized interactive content and system thereof - Google Patents

Abstract

The disclosure is related to a method for generating personalized interactive content. The method is adapted to an interactive software procedure. While a system identifies a user, the system performs semantic analysis upon the input instruction received through a user interface so as to obtain key terms. A dimensional sentiment analysis is performed to acquire positive and negative values for the terms. The sematic sentiment of the terms can be recognized and replied by the system, and be the content to be deep learned. Referring to a personalized database, the positive and negative values of sentiment, personalized record and system database, a reply is generated according to a result obtained from the relational decision approach. The mentioned steps can be repeated and under a deep learning process. The key terms can be obtained and used to converge the future replies. The replies also form part of the personalized records.

Description

Personalized interactive content generation method and system

The disclosure book discloses a method and system for generating interactive content, in particular a method and related system for generating personalized interactive content based on query results of various databases after semantic analysis and dialogue.

In various social networks, in addition to providing the service of dialogue between people, an automatic dialogue chat service has been developed. The main function is to automatically provide responses to the questions in the vocabulary according to the content of the user’s inquiry. Can provide some standardized answers.

What’s more, there is also an automatic conversational chat service that can provide answers from querying the database for a specific service range. This type of chat bot mainly queries the corresponding database according to the needs of the user, and even covers location information. A more accurate result for the user. When the above-mentioned types of chat bots operate, the information behind them is the designed dialogue content. Generally, options are automatically generated, allowing the user to select items according to the options, and then the chat bot will continue to ask follow-up questions based on the options until the answer is obtained. . For example, if you first ask for the name of the restaurant, and then provide the restaurant branch selection, after confirming the branch, and then ask for the reservation time, the chat bot will assist the user to enter the restaurant’s reservation page or provide the restaurant’s phone number.

Figure 1 shows a schematic diagram of a chat robot providing restaurant information. This example shows an example of a chat robot looking for a restaurant. In the beginning, the software running on the electronic device was used to generate a In a dialogue window 10, the chat robot generates the first dialogue 101: "Where do you want to eat?", and then the user can respond to the second dialogue 102: "Anhe Road", the chat robot generates a third dialogue 103: "Anhe Road Restaurant", after querying the database, the query result 105 is displayed, which can indicate the restaurant on Anhe Road in various ways, and attach a map. In another way, the location of the user's device can be obtained through software to provide locality results.

It is different from the chat bots established in the conventional technology through dialogue design that requires a certain degree of restriction on the user’s words. For example, the chat bots used in the current known instant messaging software such as Line, WhatsApp, WeChat, etc., are different from The current instant messaging software uses the keyword mechanism input by the user to provide a response. The personal interactive content generation method and system disclosed in the disclosure can be complicated by the relational decision mode according to the user input. Semantic analysis, in which a personalized database is established through individualized behaviors, and then a semantic analysis is used to determine whether there are positive and negative emotional dimension judgments, and then keywords are formed through the associated decision-making model. When responding to user questions, Provide personalized answers, and provide results that match the time, place, and event based on the results of big data analysis.

According to one of the embodiments, the personalized interactive content generation method includes that when a user starts an interactive software program through the user device, the system obtains the user identity through the interactive software program, and then starts a user interface through the interactive software program Receive input instructions, the input instructions can be text or voice, and then perform semantic analysis on the text or the character string generated by the converted voice, including querying a vocabulary database, filtering irrelevant words, and obtaining the input instructions under a specific category One or more keywords. After that, query a personalization database according to the user's identity, and obtain a personalization record corresponding to the user's identity, which can record personal preferences, so you can compare the obtained one or more keywords, users Personalized records, and query a system database based on keywords The result obtained generates a converged response, and then outputs the response through the user interface, which can be displayed on the user device.

When the user enters a question through this interactive software program, the system will not only search for the designed answer based on the data, but also based on the user's past records, the results of big data analysis, and the answers based on real-time data. .

Further, the personalized interactive content generation method can ask the user whether the response is satisfied when the response is output. If not, the system can repeat the above method steps, receive a second input command (text or voice), and proceed. The second semantic analysis derives further keywords, so that the system can compare further keywords, personalize records, and query the system database to generate further responses after convergence.

In the personalized interactive content generation system, there are mainly servers, including a big data database, a personalized database, and a commercial database. According to the function, the server receives the data generated by the user device through the interactive software program executed by the user device through the communication unit, and obtains the user identity from the user device through an identity recognition module. After receiving the input command through the user interface, it can The semantic analysis module is used to perform semantic analysis on the input command to obtain keywords. After that, according to the user's identity, query the personalized database through the query module, and after obtaining the personalized record, you can compare one or more keywords, personalized records, and query the big data database and commercial data through the query module The result obtained by the library generates a response with a relational decision-making module, and outputs the response via the user interface through the feedback processing module.

Preferably, the system divides the input instructions into primary keywords and secondary keywords through the semantic analysis module, and then performs an emotional dimension analysis on the input instructions to obtain the positive and negative degrees of one or more keywords. When the semantic emotion cannot be recognized, the key words formed by the input instruction are recorded and become the learning object of a deep learning library; if the semantic emotion can be recognized, the response is output according to the word and sentence rules and the emotional dimension.

Further, in the step of querying the personalized database, the system uses a query module to query whether there is an associated data model in the personalized database. When there is no associated data model, it creates a semantic meaning that has not been modeled Dimension records, generate feedback, and form related data models.

Preferably, the system can generate a response to the user's question and answer based on a personalized database, a big data database, and a commercial database in an operational process executed by an associated decision-making module and a feedback processing module.

In order to further understand the features and technical content of the present invention, please refer to the following detailed description and drawings about the present invention. However, the provided drawings are only for reference and description, and are not used to limit the present invention.

10: Conversation window

101: The first conversation

102: The second conversation

103: The Third Dialogue

105: result

22: Internet

24: User device

20: server

201: Big Data Database

203: Personalized Database

205: Commercial Database

30: Personalized interactive content generation system

31: User device

301: Communication unit

302: Identity Recognition Module

303: Semantic Analysis Module

308: Thesaurus

304: Query module

311: Big Data Database

313: Personalized Database

315: Commercial Database

305: Relational Decision Module

306: Feedback Processing Module

307: transfer module

32: customer service staff

51: Thesaurus

53: Deep Learning Library

90: System

Steps S401~S413: Operation process of the identity recognition module

Steps S501~S519: Operation process of semantic analysis module

Steps S601~S611: Query the operation process of the module

Steps S701~S711: the operation flow of the associative decision-making module and the feedback processing module

Steps S801~S811: Operation process of the transfer module

Steps S901~S913: Personalized interactive content generation process

Steps S1001~S1021: Example of the process of personalized interactive content generation

Steps S1101~S1115: Example of the process of personalized interactive content generation

Steps S1201~S1213: Example of the process of personalized interactive content generation

Steps S1301~S1311: Example of the process of personalized interactive content generation

Fig. 1 shows a schematic diagram of a dialogue example of a conventional chat robot; Fig. 2 shows a schematic diagram of an embodiment of the structure of a personalized interactive content generation system; Fig. 3 shows a schematic diagram of an embodiment of functional modules in the personalized interactive content generation system; Fig. 4 shows an individual The operation flow of the identity recognition module in the personalized interactive content generation system; Figure 5 shows the operation flow of the semantic analysis module in the personalized interactive content generation system; Figure 6 shows the operation of the query module in the personalized interactive content generation system Process; Figure 7 shows the operating process of the associated decision-making module and the feedback processing module in the personalized interactive content generation system; Figure 8 shows the operating process of the switching module in the personalized interactive content generation system; Figure 9 shows as A flowchart of an embodiment of a method for generating personalized interactive content; Fig. 10 shows one of the exemplary flow charts of the method for personalized interactive content generation; Fig. 11 shows the second exemplary flow chart of the method for personalized interactive content generation; Fig. 12 shows the exemplary flow chart of the method for personalized interactive content generation Figure 3; Figure 13 shows another example flow chart of a method for generating personalized interactive content.

The following are specific specific examples to illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be based on different viewpoints and applications, and various modifications and changes can be made without departing from the concept of the present invention. In addition, the drawings of the present invention are merely schematic illustrations, and are not drawn according to actual size, and are stated in advance. The following embodiments will further describe the related technical content of the present invention in detail, but the disclosed content is not intended to limit the protection scope of the present invention.

It should be understood that although terms such as "first", "second", and "third" may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are mainly used to distinguish one element from another, or one signal from another signal. In addition, the term "or" used in this document may include any one or a combination of more of the associated listed items depending on the actual situation.

The disclosure book is about a method of personalized interactive content generation and the system that implements this method, which is different from the chat established in the conventional technology that requires a certain degree of restriction on the user's language (standard fixed keywords) through dialogue design. For bots, such as chat bots used in instant messaging software such as Line, WhatsApp, WeChat, etc., the method disclosed in the disclosure book can create a personalized database based on user input words through personalized behavior and big data analysis, and then semantic analysis (semantic analysis) to determine the positive and negative sentiment dimensions, only through the correlation decision-making model Key words, in which more accurate and rapid semantic analysis results can be obtained with the help of neural network technology, and accurate keywords are formed. When responding to user questions, a personalized database is used to record each member or virtual member Personalized information to provide personalized answers, and in accordance with the results of big data analysis and decision-making, to provide results in line with time, place and event.

Realize the personalized interactive content generation method proposed in the disclosure book. In the application of the proposed interactive dialogue, the system will establish a personal database for each user through the database, if in each personalized interactive service process If there is relevant information in the database, there is no need for the user to repeat the answer, and the system does not need to ask again, but can directly respond to the question, resulting in a "jumping question" effect, while the system continues to operate and has always been dialogue Collect user’s personal information in.

In the step of generating personalized interactive content (the flow in FIG. 9), a software program is used to realize the dialogue. The system implementing this method is shown in FIG. 2 as a schematic diagram of an embodiment of the structure of the personalized interactive content generating system.

In this embodiment of the architecture, the server 20 provides several databases. The personalized database 203 is used to record the data of the users participating in the system and the data generated by the users through the system. For example, when the user starts an interactive software program on the user device 24, the server 20 of the system can obtain the user identity through the network 22, and compare the personalization database 203 to obtain the data generated by the user in the past. The user's history can be obtained and a personalized record can be formed. Moreover, when the user interacts with the system through an interactive software program, the data, such as text or voice, may also be input commands from other audio-visual content. After being converted into a string, semantic analysis is used to obtain meaningful keywords. Including the last response provided by the system, it can be converted into a record in the personalized database 203.

The interactive software program may be an interactive dialogue software that is executed in a computer device to obtain answers in a dialogue question and answer mode, which may be equipped with interactive chat and various forms of human-computer interaction mechanisms. In addition to ordinary personal computers and smart handheld devices, computer devices can still be widely used Various electronic devices with computing and communication capabilities, such as smart speaker devices, smart home networking devices, robots, smart air conditioners, and other devices with human-machine interface devices (HMI devices).

The big data database 201 provided by the server 20 records various types of data collected on the Internet and an open database provided by government departments, and the commercial database 205 provides various types of real-time data.

According to the embodiment, the big data database 201 is mainly big data formed by the system collecting information on the Internet through the software program in the server 20. In addition to collecting all kinds of data on the Internet, it can also Obtain open data provided by the government or a specific organization. However, the big data database 201 can obtain corresponding data according to the content to be served by the system, so that meaningful content for a specific purpose can be formed. For example, in the service project of tourism, the big data database 201 can collect data on the Internet according to tourism-related topics, including the content of tourism discussed on the social network, the tourism information discussed by various blogs, and the content of various tourist attractions websites. , And can include official information to form big data on the topic of tourism. Taking catering as an example, the server 20 collects data from the Internet, officials, and industry through software programs, which can also form big data related to catering. In this way, when the system needs to respond to users' questions about tourism, tourism, or specific fields (such as finance, people's livelihood, retail, etc.), the system can provide more useful and accurate information.

The commercial database 205 provided by the server 20 mainly provides real-time information in a specific field, so that the system can respond to the user's most real-time and correct information. For example, the commercial database 205 obtains real-time information such as room reservations, parking spaces, aircraft positions, etc. of the travel industry, and can respond to relevant information inquired by users through the system accordingly. Similarly, the commercial database 205 can obtain real-time reservation information of catering-related stores, real-time stocks, futures, funds, bonds and other information of financial service providers, real-time parking information provided by parking companies, and various livelihood commodities. Price is Time quotation enables the system to provide the user with real-time and correct information based on the user's inquiry through the server 20.

In order to realize the personalized interactive content generation method, a software program should be executed on the user device 24 on the user side. After being executed by the processor of the user device 24, the operating interactive software program will communicate with the system through the network 22. Through interactive dialogue, the response provided by the system is obtained, and the software program executed in the system server 20 will continue to calculate, and learn the individual of each user through semantic analysis, big data analysis, machine learning, and neural network technology. It can provide more appropriate personalized interactive services.

It is mentioned here that when semantic analysis and decision-making use neural network-like technology, one or more input commands (text or voice) input by the user through the user interface can be used to form input parameters, where the input layer ( In the layer composed of many neuron interaction links between the input layer) and the output layer (output layer), some output results are summarized by neural network-like technology, and the output results need to be confirmed or responded by the user. Learn, modify the parameters of the input layer, and converge until a result acceptable to the user is obtained.

Among the various neural network technologies that can be used, for example, Google ^{TM in the} United States once proposed to use a neural network system to accelerate the development of its translation services. Among them, SyntaxNet, a neural network architecture for semantic analysis, is used. It is an open source code form that can make natural semantic analysis more accurate and correct with the help of everyone, thereby making its translation, voice input and other services more usable. Therefore, this type of neural network architecture SyntaxNet can also be applied to semantic analysis in the personal interactive content generation method proposed in the disclosure.

The personalized interactive content generation system can implement interactive services for users at various ends through a computer server. For the software program that implements the personalized interactive content generation method, please refer to the software module embodiment diagram shown in FIG. 3 .

The figure shows that the personalized interactive content generation system 30 can be hosted by a computer, cloud Servers and clusters are implemented, and the user device 31 needs to execute software programs that interface with the system. On the user device 31, the user can interact with the system through voice dialogues, text, or can be combined with videos, pictures, etc., especially voice and text are more commonly used for semantic analysis (semantic analysis) method is better Interactive way.

When the user inputs voice or text through the software program executed by the user device 31, the commands and messages generated by the software program of the user device 31 will be converted into input strings, or can be encrypted and sent to the communication unit 301 Personalized interactive content generation system 30. In addition, it does not rule out that the software program on the system side performs the task of converting the string.

At the beginning, the personalized interactive content generation system 30 obtains the identity recognition information generated by the user device 31 through the identity recognition module 302, including authentication information required to execute interactive software programs, device information, or information required by the system to log in. Based on this information, the user's identity is obtained, which becomes the basis for performing authentication through the personalized database 313, inquiring about individualized records, and so on.

The system provides a semantic analysis module 303, which is connected to a vocabulary 308. When the user enters commands such as voice or text through the user interface initiated by the interactive software program running on the user device 31, the input string is converted to personalize the interaction. The content generation system 30 receives the input string through the communication unit 301, and then uses the semantic analysis module 303 to perform semantic analysis on the input text or the string obtained by converting the speech, and compares the word database 308 to obtain the input string One or more keywords, these keywords become the basis for the system to respond to users later.

The semantic analysis technology used here mainly refers to the analysis of a long string of text or content to get the meaning, which can be applied to text, voice or other audio-visual content. After comparing with a specific range of thesaurus 308, the thesaurus 308 can be classified Record the words used in different fields, and after comparison, you can get meaningful keywords in a specific field. Moreover, semantic analysis technology can also be based on big data algorithms, in which neural network technology can be used to analyze the information expressed in user input commands, and can eliminate incorrect semantics, based on machine learning. , Can judge more accurately Break out the keywords in the input command.

The query module 304 in the system assists each functional module to query the contents of the big data database 311, the personalized database 313, and the commercial database 315. Among them, the personalization database 313 can be queried according to the user's identity, and a personalized record corresponding to the user's identity can be obtained. According to the needs of semantic analysis, the big data database 311 can be queried, and then in the process of generating a response, compared with all The obtained one or more keywords, personalization records, and the results obtained by querying the big data database 313 and the commercial database 315 through the query module 304, are responded to by a relational decision module 305.

The associated decision-making module 305 can integrate various data, such as user identity, keywords, big data, commercial data, etc., to generate a response to the user input command, and then the feedback processing module 306 according to the personalized database 313 The obtained user identity is output to the user device 31 in response.

The personalized interactive content generation system 30 provides a transfer module 307, which is a software program for the system to provide online customer service for the user and convert the customer service staff 32.

When a user enters the system, the system will authenticate the user (also a consumer under a certain demand), which adopts the operation process of the identity recognition module in the personalized interactive content generation system as shown in FIG. 4.

For example, in step S401, after the identity recognition module implemented by software in the system is activated, a digital fingerprint data is created for the members or non-members of the system (step S403). This is when the user uses the system, the system collects the user Data, establish a database, or directly obtain the information of the computer device used by the user, as the basis for identifying the user, so that the system can still identify the user based on the history of some system use even when the user is not a logged-in member.

In this process, in step S405, the system determines whether the user is a new visitor? If it is a new visitor (a user who is not logged in), a virtual membership can be created based on the user's history. For example, in step S407, a virtual member is created for the user, where the so-called digital fingerprint is the virtual member When the user officially becomes a member of the system (return to account), such as step S409, the digital fingerprint becomes the member feature (step S413), which is to optimize the flow of the method for generating personalized interactive content. On the other hand, when the user is a member of the system, as in step S411, similarly, various behaviors of the member in the system form member characteristics (step S413). In the personal interactive content generation method, the response member can be effectively formed Reference material for the material.

Figure 5 shows the operation process of the semantic analysis module in the personalized interactive content generation system. This process provides a solution that can effectively obtain the meaning of a long string of text or content analysis, such as step S501, the system starts to be implemented by software Semantic analysis module. After the user enters text or voice signals through the user interface, the software program first breaks words according to the text, or according to the meaning and context of the string converted from the speech (step S503), and then compares the words After the library 51, in step S505, the input text is divided into main keywords and secondary keywords, and the words input by the user are subjected to dimensional sentiment analysis (step S507).

In the sentiment dimension analysis, the positive and negative degrees of keywords can be distinguished according to the meaning of the keywords, and the text information can be divided into strong, calm, etc., which can conduct a more detailed analysis of the needs of users, and also use the thesaurus 51. Word query obtained through deep learning.

In step S509, the system analyzes the sentiment dimension based on the extracted primary keywords and secondary keywords, and then determines whether the semantic sentiment can be identified? If it cannot be identified (No), this information (such as input text or speech keywords) will be recorded in step S511 and become the learning object of the deep learning library 53, and in step S513, the user interface can be used to reply. Understand the semantics, and then require the user to re-enter instructions, and related follow-up data also become the basis for deep learning.

If the system can recognize the emotional dimension of the text input by the user (Yes), in step S515, the software program of the system will obtain the analysis result of the words input by the user, which is used to directly provide personalized content in a specific interactive dialogue The purpose of skipping questions is based on the rules of words and sentences and emotion The dimension output response, such as step S517, outputs the dimension related to the response depth to the system, and may end this part of the personalized interactive dialogue (step S519). The various information generated during the period becomes the record of the user's digital fingerprints, characteristics and deep learning.

Figure 6 shows the operation flow of the query module in the personalized interactive content generation system. The query module is also a function implemented by the software program in the system. For example, in step S601, the system is implemented as a software program after starting the query module. The main purpose described in step S603 is to query whether there is an associated data model in the database according to the semantic dimension.

The data model is the type of interactive content. According to the semantic analysis, the user's question and answer category is obtained, such as travel, food, life, etc. The purpose is mainly to converge the decision tree (decision tree) and produce the effect of skipping questions. Meet the needs of personalization. The data model includes the original model based on the past history and the new model generated by the new dialogue.

If there is no associated model in the database (No), that is, in step S605, a semantic dimension record that has not been modeled is created, and in step S607, the system feeds back the sub-related data model to this and returns to step S609. The feedback forms a data model, and the purpose of this query is completed (step S611).

If it is determined in step S603 that the associated model is obtained according to the semantic dimension query result (Yes), it means that the same user’s dialogue is still in the model constructed by the past history, and the data model is still fed back (step S609), and End (step S611).

In the operation process of the associated decision-making module and the feedback processing module in the personalized interactive content generation system shown in Figure 7, the application of the personalized database (Figure 3, 313) and big data data (Figure 3) are provided. , 311) and the commercial database (Figure 3, 315) to form the decision-making and feedback processing flow.

In step S701, in the interactive content generation process, the system initiates the software program executed by the associated decision-making module and the feedback processing module, according to the above-mentioned semantic analysis of the text generated by the user in the interactive content generation method Data model to get the classification of interactive content, namely In step S703, the feedback with the highest weight is searched, and at this time, the personalized database is queried (step S705) and the big data database (step S707).

Decision-making methods. For example, the system establishes a category based on the data model, and establishes a decision tree (decision tree) respectively. The feedback with the highest weight will achieve the effect of skipping the question, that is, it does not need to make decisions according to the decision tree set by the original system. Can dynamically jump out of the decision tree and quickly get results.

When querying the personalized database, you can query the personalized records formed by the data and history generated by each user, including past semantic analysis to get meaningful personalized keywords, and you can get the positive and negative emotions of the compared keywords. Dimensions, and personalized records formed after the system responds.

When querying the big data database, you can query all kinds of information (big data) collected by the system through software programs on the Internet. In addition to all kinds of information on the Internet, it can also include open data. After integration, various types of big data are formed, which can provide users with more extensive and accurate data.

Then, in step S709, the system continues to provide real-time information provided by commercial databases, such as real-time travel information, real-time information in life, and real-time public issues, etc., based on the query content derived from the above-mentioned personalized database and big data database. , And finally produce the result of the relational decision and feedback (step S711), which can make an interactive response to the result obtained through the relational decision mode.

In addition to online interactive content services, the system also provides real-person customer service. The related process is shown in Figure 8 for the operation process of the transfer module. The transfer module is a software program that the system provides users with online customer service and converts the customer service staff.

When the interactive content generation process is executed, the user can transfer to customer service under specific conditions under specific consumer needs, such as step S801, the software program of the transfer module is also started during the interaction process, and in step S803, The system will judge whether the feedback content meets the consumer according to the type of feedback obtained from the interactive content.

When the feedback type matches the consumer (Yes), these records form the feedback data (step S807), and the system also provides the consumer to directly establish an online customer service connection (step S809). After the connection is established with the customer service center, the transfer ends Flow (step S811).

On the other hand, when the judgment of the feedback type does not meet the consumer (No), it means that more information is needed to determine the issue of the customer service. Continue to step S805, and the system continues to guide the consumer. If the judgment is to guide the question-and-answer consumer (Yes ), continue to generate feedback data (step S807), and establish an online customer service connection (step S809); conversely, if the system determines that the question-and-answer consumer is not to be guided (No), the online customer service connection is directly established in step S809.

Figure 9 then describes the flow of an embodiment of the method for generating personalized interactive content. In this flow, at the beginning, as in step S901, the system 90 receives user authentication information through the interactive software program executed in the user device to authenticate the use User identity, including virtual members that can implement the identity recognition module shown in Figure 4, even if the user is not a logged-in member of the system, the system can still create a virtual member based on the digital fingerprint of each user. For example, when the interactive software program is started on the user's device, the user can be required to log in to authenticate the user (or the user can be identified without logging in). This process allows the user to enter authentication information, or Automatically log in to the system based on previous records.

Then, in step S903, the user can input voice, text, etc. input commands through the user interface initiated in the user device, and then in step S905, start to analyze the semantics of the input commands, where the input commands can be text or voice. At the same time, if the system receives text, it directly queries the vocabulary. If it is voice or other digital content, the software program in the system can convert the command into a string and send it to the system to query the vocabulary. After that, the irrelevant words are filtered to obtain one or more keywords under a specific category. In one embodiment, after semantic analysis is performed through the software program in the system as described in FIG. 3, in step S907, one or more keywords in the input instruction can be obtained, and these keywords become the basis for the subsequent system to query the database .

Then, in step S909, the system 90 queries the personalized database according to the user's identity to obtain a personalized record. This personalized record can record the information retrieved by the user during the course of using the system service in the past. For example, when the user's demand is to find a parking space, the system can obtain the parking information that the user has obtained through the system in the past to determine the user's vehicle form (size, car or locomotive) and acceptable parking fees, etc. Therefore, the system will be able to accurately provide parking space information that meets the user’s preferences on this basis in this interactive service.

When the system 90 obtains the user's personalized record in step S909, then as in step S911, the software program in the system will compare the keywords with the personalized record to converge, and then query the system database, including the data that can be matched from the big data database. Inquiry condition data, and can obtain real-time information according to the commercial database, and obtain the inquiry result according to the key words. Finally, in step S913, the response is output to the user device, and the result is displayed through the user interface in the user device. It is worth mentioning that if the user terminates the interactive software program after the response is output, the response can be accepted. This response can form the personalized data and become a part of the personalized database.

According to one of the embodiments, in this method, the interactive software program may be an interactive dialogue software that is executed on a computer device to obtain answers in a dialogue question and answer mode. Therefore, the user can input requirements through natural words in an interactive dialogue mode. got the answer.

The above steps can be repeated many times until the user terminates the process. In this way, when the system 90 outputs a response to the user through the user interface, if the user needs further information, the user can continue to inquire in an interactive dialogue repeatedly, including continuing to generate Enter the instruction, the system will continue to receive the second input instruction and perform semantic analysis. From the second input instruction, further keywords are obtained. After comparing the further keywords, personalization records, and querying the system database, the system will generate a pass convergence Further response afterwards.

After repeating the above method steps, further keywords can be used to obtain a further response after convergence. Similarly, the obtained response will form a record in the personalized database.

The following FIG. 10 shows an example flow chart of a method for generating personalized interactive content. In this example, a user finds a parking space through an interactive software program executed on the user's device as an example.

At the beginning of this method, the system can obtain the user's identity first. The user enters the word "find a parking space" through a user interface, such as an interactive dialogue software interface (step S1001), (in other embodiments, voice input can be used, not limited to text input), and then the input command is sent to After the system receives this input command (text or voice), it queries the commercial database in the system, and the system cooperates with the location information sent by the user device to provide the nearest parking space through the user interface (step S1003), for example, provide parking Parking space information such as location (such as map), type, price, etc., and navigation services can be provided when necessary. Then, the system can actively ask "Is it satisfied?" through the user interface (step S1005), or the user can decide whether to continue the query.

When the user is satisfied (Yes), the dialogue will be ended (step S1007). Among them, the way for users to express satisfaction is to close the software program directly, and the system can propose options for users to decide whether to continue to inquire or to have satisfied results. In this way, this response will form a record of the user's personal information and become one of the references for the next parking space query.

If the user expresses the need to continue inquiring in a specific way (No), the system may request other information to be input (step S1009), for example, the user may further provide the vehicle type, acceptable fee range, and parking area. When the system receives the user’s input as an "electric vehicle" (step S1011), after semantic analysis, the system knows that this type of vehicle needs to be charged. Therefore, if the data in the commercial database matches, the system provides the relevant information about the nearest rechargeable parking space. Information (step S1013). Similarly, is the inquiry satisfied? (Step S1015), if the answer is satisfied (Yes), the dialogue ends (Step S1007). In the same way, all relevant responses can become records in a personalized database.

If the user still needs to continue to inquire (No), in one embodiment, the system can continue to generate guidance questions (step S1017) until satisfactory parking space information is generated (step S1019), and the obtained responses form personalized data Recording of the library (step S1021).

Compared with FIG. 10, it may be the first time the user uses the system to inquire about the parking space, and the flow example shown in FIG. 11 shows that the system has obtained personalized information.

At the beginning, the user asks to find a parking space through a software program executed in the user’s device, and the system receives the “find parking space” command (step S1101), because the system has already obtained the user from the previous process (Figure 10) The vehicle that may be driven is an electric vehicle. Therefore, the system can directly generate the nearest rechargeable parking space here (step S1103). It can also include information such as the price and location of the parking space. Is it satisfactory after inquiry? (Step S1105), if the response is satisfied (Yes), the dialogue can be ended (Step S1107).

The user can still continue the conversation. The answer may be unsatisfactory (No). In this example, the system actively asks for price information (step S1109) and can generate guidance questions (step S1111) until satisfactory parking space information is generated. (Step S1113), the final or in-process content will form a record of the personalized database (Step S1115).

From the process shown in Figure 11, it can be seen that since the personalization database of the system has recorded the history of the users in the past, it can cooperate with the big data database and the commercial database under the same issue to quickly provide useful responses, which can effectively shorten the interactive process. .

Figure 12 then shows a further example of the process.

The same problem of finding a parking space, when the user uses the system again to find a parking space, he enters the "find parking space" command through the user interface, and the system receives this command (step S1201), and obtains the past history by querying the personalized database. And through semantic analysis to get keywords, and query big data database and commercial database to get immediate parking space information, you can quickly generate the latest chargeable parking space information and the price within a certain range through the user interface (step S1203) , It can also include various parking space related information, so that you can save some interactive processes.

Is the inquiry satisfied? (Step S1205), if the answer is satisfied, the dialogue will be ended (Step S1207); otherwise, the system can guide the question (Step S1209) to ask the user for other needs, Until satisfactory parking space information is generated (step S1211), these information can be recorded in the personalized database (step S1213).

In addition to the above examples of finding parking spaces, the system can provide restaurant-finding services under the same concept. When the user enters to find a restaurant, the system will query the past personalized records according to the user’s identity, which can accurately provide the user’s needs. The restaurant, including restaurant location, price, type, habit of eating time, etc., because of the learning mechanism of the system, many inquiries can be omitted. If no consistent information is obtained at the first time, other suggestions can be provided according to preferences.

The relevant implementation example describes a service process related to personal credit as shown in Figure 13. The software program in the system obtains the user's past loan records by querying the personal database, including the personal credit evaluation data that has been submitted. When asking credit-related questions, you can ignore some of the steps in the interactive dialogue, which has the effect of skipping questions.

According to this example, in step S1301, the system receives a user input of "please perform credit evaluation" through the user interface of the interactive software program, which can be input commands such as text or voice input. In step S1303, the system receives the input command according to the input The command is converted into a string, and after the keywords are obtained through semantic analysis, the personalized database is queried.

In this step, if the system has already recorded relevant data about the user’s personal credit in the personalized database, you can skip the subject and directly provide credit information, such as step S1305, without the need to re-inquire the user about the credit information. Information on loans such as mortgages or car loans.

However, the system can still ask "Is there any other guarantee" again in step S1307? Or other information that needs to be supplemented by the user, especially information about the amount of credit and interest rates. If not, proceed directly to step S1311, and the system proposes the loan amount and interest rate based on the previously collected data; if the user indicates that there is still information to be given, as in step S1309, the system receives the relevant data of other collateral again and recalculates, as in step Step S1311, respond to the loan amount and interest rate.

The personalized interactive content generation method realizes the effect of directly skipping questions in the dialogue, According to the parking lot and credit evaluation examples in the manual, if the system has obtained user-related information, it can directly provide parking lot or directly evaluate house loans and car loans.

In summary, according to the personal interactive content generation method proposed in the disclosure, when the user enters a question through the user’s device and the initial interactive software program, the system will not only search for the data and get the designed answer, but also It can be based on semantic analysis, user identity, past history, and big data analysis, as well as answers based on real-time data. When the user still cannot get a satisfactory answer, the system can repeat the method steps, receive a second input command, perform a second semantic analysis to obtain further keywords, so that the system can compare further keywords, personalize records, and After querying the system database, a further response after convergence (jump question) is generated.

The content disclosed above is only the preferred and feasible embodiments of the present invention, and does not limit the scope of the patent application of the present invention. Therefore, all equivalent technical changes made using the description and schematic content of the present invention are included in the application of the present invention. Within the scope of the patent.

S501: Semantic Analysis Module

S503: Hyphenation

S505: The relationship between primary keywords and secondary keywords

S507: emotional dimension

S509: Is it possible to identify semantic emotions?

S511: Record

S513: Cannot understand the meaning of the reply

S515: Output according to word rules and emotional dimensions

S517: Output dimension

S519: End

51: Thesaurus

53: Deep Learning Library

Claims (8)

A method for generating personalized interactive content includes: (a) obtaining a user identity when starting an interactive software program, wherein the interactive software program is executed on a computer device and provides a user with a dialogue question and answer method Answer: (b) Receive an input command through a user interface initiated by the interactive software program, where the input command is a text or a voice; (c) execute a semantic meaning on the text or a string generated by converting the voice Analysis includes querying a vocabulary database and filtering irrelevant words to obtain one or more keywords in a specific category in the input instruction; (d) querying a personalized database based on the user’s identity to obtain A personalized record corresponding to the user's identity, where the personalized record is the data and history generated by the user corresponding to the user's identity, including personalized keywords and responses derived from past semantic analysis; (e) comparing the one Or more keywords and the personalized record to generate a response based on the result of querying a system database based on the one or more keywords; and (f) output the response through the user interface; where, in When step (f) outputs the response and terminates the interactive software program, the response forms a record in the personalized database; then when step (f) outputs the response to the user’s input command, repeat step ( b) to (e), to receive a second input command, semantically analyze the second input command to obtain further keywords, after checking the further keywords, the personalized record, and querying the system database, Produce a further response after converging. The method for generating personalized interactive content according to claim 1, wherein the semantic analysis is implemented by a semantic analysis module to achieve the following steps: Perform word segmentation according to the input instruction; compare the thesaurus to distinguish the input instruction into a primary keyword and a secondary keyword; perform an emotional dimension analysis on the input instruction to obtain the positive and positive values of the one or more keywords Negative degree; if the semantic emotion cannot be identified, the one or more keywords formed by the input instruction will be recorded and become the learning object of a deep learning library; if the semantic emotion can be identified, the response will be output according to the word and sentence rules and the emotional dimension . The method for generating personalized interactive content according to claim 1, wherein, in the step of querying the personalized database, a query module is used to query whether there is an associated data model in the personalized database. The personalized interactive content generation method according to claim 3, wherein when there is no associated data model, an unmodeled semantic dimension record is created to generate feedback to form an associated data model. The method for generating personalized interactive content according to claim 1, wherein the system database includes a large data database and a commercial database, and the big data database is recorded from an open network provided by a government agency. The database collects and corrects various types of data; the commercial database provides various types of real-time data. The method for generating personalized interactive content according to claim 5, wherein in the operation process executed by an associated decision-making module and a feedback processing module, according to the personalized database, the big data database and the The commercial database generates this response. The method for generating personalized interactive content according to claim 1, wherein, if the interactive software program is terminated after outputting the further response, the further response forms a record of the personalized database. A system for executing the method for generating personalized interactive content as described in claim 1, the system including a server, wherein the server is provided with a large data database, a personal resource A database, and a commercial database, where a humanized interactive content generation method running on the server includes: the server receives a user device through a communication unit and an interactive software program executed on the user device Generated data; obtain a user identity from the user device through an identity recognition module; receive an input command through a user interface initiated by the interactive software program; pass a semantic analysis module to the input command Perform semantic analysis to obtain one or more keywords in the input command; according to the user's identity, query the personalized database through a query module to obtain a personalized record corresponding to the user's identity. Personalized records are data and history generated by a user corresponding to the user’s identity, including personalized keywords and responses derived from past semantic analysis; comparing the one or more keywords, the personalized record, and the The query module queries the results of the big data database and the commercial database, generates a response with an associated decision-making module, and outputs the response through the user interface through a feedback processing module.

Images