TWI675303B - Conversational financial service device and system - Google Patents

Abstract

A dialog financial service device includes a transceiver module, a control module connected to the transceiver module, and an emotion recognition module connected to the control module. The transceiver module receives a voice message, and the control module transmits the voice message. To the emotion recognition module, and according to the emotion contained in the voice message recognized by the emotion recognition module, a reply message is generated correspondingly.

Description

Dialogue financial service device and system

The present invention relates to a financial service device and system, and more particularly to a dialog-type financial service device and system for obtaining financial services in a natural language dialogue mode.

In recent years, due to the rapid development of communication and network-related industries, people have increasingly demanded for the electronics and mobile industries. For the financial industry, in addition to providing online banking, it also provides mobile banking services, so that users can log in to online banking or mobile banking through various electronic devices such as computers and smart phones, thereby directly online. Financial transactions such as checking accounts, transferring funds, paying fees, foreign exchange, and remittances are carried out on the road.

However, in the existing technology, although online banking or mobile banking can bring many conveniences, when operating various financial services, you still need to keep your eyes and hands on the physical keyboard or screen to operate. Troublesome, and for the visually impaired or physically disabled users, the existing click interface is not friendly.

Therefore, an object of the present invention is to provide a dialogue financial service device for obtaining financial services by using a natural language dialogue method.

Therefore, the dialog financial service device of the present invention includes a transceiver module, a control module connected to the transceiver module, a natural language dialog module connected to the control module, and an emotion recognition module connected to the control module. The module receives a voice message, the natural language dialogue module is used to identify the content of the voice message, the emotion recognition module is used to recognize the emotion contained in the voice message, and the control module is based on the emotion recognition module and the natural language dialogue module The recognition result correspondingly generates a reply message, wherein the judgment weight of the recognition result of the emotion recognition module by the control module is higher than that of the natural language dialogue module.

In one embodiment, the emotion recognition module includes a text recognition unit and an emotional word database connected to the text recognition unit. The emotional word database stores a plurality of words or sentence patterns related to emotional text, and the text recognition unit captures The corresponding voice text in the voice message, and the emotion word database is used to determine the emotional text of the voice message, and an emotional probability score is generated, and the control module determines the content of the voice message based on the emotional probability score. mood.

In addition, the emotion recognition module may further include a voiceprint emotion recognition unit and an emotion feature database connected to the voiceprint emotion recognition unit. The emotion feature database stores plural voiceprint signals and prediction models related to emotions, and voiceprint emotions The recognition unit captures the corresponding voiceprint signal in the voice message, and determines the emotion of the voice message with the emotion feature database, and generates an emotion feature value, and the control module determines the emotion probability score and the emotion feature value Emotions contained in voice messages.

In one embodiment, the natural language dialogue module includes a natural language understanding unit and a dialogue generation unit. The natural language understanding unit is configured to perform a corresponding financial service according to the content of the voice message, and the dialogue generation unit is configured to perform a corresponding financial service based on the voice message. Generate corresponding conversation content.

Further, the natural language dialogue module further includes a language-to-text unit and a text-to-speech unit. The language-to-text unit is used to compare whether the text in the voice message conforms to at least one financial vocabulary, and the text-to-speech unit is used to convert the text. Convert to corresponding audio signal output.

In addition, another object of the present invention is to provide a dialogue financial service system for obtaining financial services using a natural language dialogue method.

The dialog financial service system of the present invention includes a client device that generates a voice message and a dialog financial service device that communicates with the client device. The dialog financial service device includes a transceiver module that receives the voice message, a A control module connected to the transceiver module, a natural language dialogue module connected to the control module, and an emotion recognition module connected to the control module. The natural language dialogue module is used to identify the content of the voice message. The emotion recognition module To identify the emotion contained in the voice message, the control module generates a response message corresponding to the recognition result of the emotion recognition module and the natural language dialogue module, wherein the control module judges the weight of the recognition result of the emotion recognition module. Will be higher than the recognition result of the natural language dialogue module.

In one embodiment, the emotion recognition module includes a text recognition unit and an emotional word database connected to the text recognition unit. The emotional word database stores plural words or sentence patterns related to emotional text, and the text recognition unit retrieves the word or sentence pattern. The corresponding voice text in the voice message, and the emotional word database to determine the emotional text of the voice message, and generate an emotional probability score value, and the control module determines the emotion contained in the voice message based on the emotional probability score value .

In addition, the emotion recognition module may further include a voiceprint emotion recognition unit and an emotion feature database connected to the voiceprint emotion recognition unit. The emotion feature database stores plural voiceprint signals and prediction models related to emotions, and voiceprint emotion recognition The unit captures the corresponding voiceprint signal in the voice message, and determines the emotion of the voice message with the emotion feature database, and generates an emotion feature value. The control module is based on the emotion probability score and the emotion feature value. Determine the emotion contained in the voice message.

In one embodiment, the natural language dialogue module includes a natural language understanding unit and a dialogue generation unit. The natural language understanding unit is configured to perform a corresponding financial service according to the content of the voice message, and the dialogue generation unit is configured to perform a corresponding financial service based on the voice message. Generate corresponding conversation content.

Further, the natural language dialogue module further includes a language-to-text unit and a text-to-speech unit. The language-to-text unit is used to compare whether the text in the voice message conforms to at least one financial vocabulary, and the text-to-speech unit is used to convert the text. Convert to corresponding audio signal output.

The function of the present invention is that: using natural language interaction to obtain financial services can be more convenient in operation, not only eliminating the trouble of the user having to stay on the physical keyboard or screen with both eyes and hands, and for people with disabilities or vision Users with disabilities can also easily and simply complete financial services.

Before the present invention is described in detail, it should be noted that in the following description, similar elements are represented by the same numbers.

Referring to FIG. 1, an embodiment of a dialogue financial service system 100 according to the present invention. The dialogue financial service system 100 can obtain financial services through a natural dialogue, which not only provides services to customers with special needs, but also solves the current problem. Users must maintain the trouble of keeping eyes and hands on the physical screen to increase the convenience of operation and life.

The dialog financial service system 100 includes a client device 10 and a dialog financial service device 20 that communicates with the client device 10.

The client device 10 may be a mobile device, a personal computer (PC), a tablet computer, a notebook computer (NB) and other electronic devices operated on the user terminal. The electronic device can be installed / opened by an application (APP), social software, An interface such as a webpage executes the interactive financial service of the present invention, that is, the user can talk to the interactive financial service device 20 through the electronic device to obtain financial services. The client device 10 generates a corresponding voice message according to the speech spoken by the user.

The dialog financial service device 20 is a back-end host or server located at the bank (or enterprise) side, which includes a transceiver module 21, a control module 22 connected to the transceiver module 21, and a control module 22 connected to the control module 22. Emotion recognition module 23.

The transceiver module 21 is a wireless transmission circuit, such as: radio frequency circuit (RFID), or near field communication (NFC), WiFi, Bluetooth, 3G / 4G network, etc., which is used to communicate with The client device 10 communicates to transmit signals to each other.

The control module 22 may be a microprocessor (MCU) or a central controller to control the operations and signal transmission between the modules. The control module 22 transmits the voice messages received by the transceiver module 21 to the emotion recognition. Module 23, and generates a response message corresponding to the emotion contained in the voice message identified by the emotion recognition module 23, and returns the response message to the user-end device 10 through the transceiver module 21 to respond to the user's request Narrative content.

In this embodiment, the emotion recognition module 23 includes a text recognition unit 231, an emotional character database 232 connected to the text recognition unit 231, a voiceprint emotion recognition unit 233, and an emotion feature connected to the voiceprint emotion recognition unit 233 Database 234.

The emotional word database 232 stores a plurality of words or sentence patterns related to emotional words, for example: angry emotion words (such as "swear words", "what are you doing", etc.), and confused emotional words (such as "Are you sure?", "Why?", Etc.), impatient emotional words (such as "Can you hurry?", "Very slow", etc.), happy emotional words (such as "Very Good," "Great", etc.), Not limited to the above. The text recognition unit 231 receives the voice message and extracts the corresponding voice text, and cooperates with the emotional character database 232 to generate an emotional probability score value according to the combination of grammatical sentence pattern and keywords, and according to a preset threshold ( threshold) to judge the above various emotional texts. For example, when the sentence pattern is a negative sentence or contains abusive words, the combination has a high probability of angry emotions (ie, a high emotional probability score).

The emotion characteristics database 234 stores a plurality of emotion-related voiceprint signals and prediction models. The voiceprint emotion recognition unit 233 receives the voice message and extracts the corresponding voiceprint signal, and cooperates with the voiceprint signal and prediction model stored in the emotion feature database 234 to determine the above-mentioned various emotions (including anger, doubt, impatience). , Happy, etc.) and generate an emotional characteristic value.

In other words, the character recognition unit 231 and the emotional character database 232 use characters such as emotional vocabulary to determine the emotion of a voice message, and the voiceprint emotion recognition unit 233 and the emotional feature database 234 use voiceprints such as pitch fluctuations to determine Emotions of voice messages, of course, two methods can be used to determine which one to use in accordance with different applications (that is, the emotion recognition module 23 includes only the text recognition unit 231 and the emotion word database 232, or the voiceprint emotion recognition unit 233 and emotion Feature database 234), therefore, the control module 22 determines the emotion contained in the voice message according to the emotional probability score generated by the text recognition unit 231 and / or the emotional feature value generated by the voiceprint emotion recognition unit 233, and generates the corresponding Reply to the message.

In addition, the interactive financial service device 20 may further include a natural language dialog module 24 connected to the control module 22 to determine the content of the voice message received by the transceiver module 21.

The natural language dialogue module 24 includes a language-to-text unit 241, a natural language understanding unit 242, a dialogue generation unit 243, and a text-to-speech unit 244.

The language-to-text unit 241 contains a financial vocabulary database, which stores a plurality of financial words related to financial services and products, such as: Yuxi card, LINE PAY card, etc. The language-to-text unit 241 is used to extract from the voice message Take the corresponding text and cooperate with the financial vocabulary database to check whether there is at least one relevant financial vocabulary in these texts to improve the accuracy of speech-to-text recognition. In addition, the language-to-text unit 241 can also contain a Chinese acoustic model, which can collect pronunciations of different users for the same financial vocabulary, in order to increase the adaptability to accents (such as: Taiwanese Mandarin), and to optimize the speech-to-text Correct conversion rate.

The natural language understanding unit 242 may contain a database of intents and instructions. The intent and instructions database stores plural control instructions related to financial transactions. The natural language understanding unit 242 may cooperate with the intents and instructions according to the content of the voice message. Control instructions in the database execute corresponding financial services, including personal account balance inquiry, bill inquiry, transaction details inquiry, remittance transfer, bill payment, income and expenditure records and inquiry, stock exchange price inquiry, arrival price reminder, transaction booking , Investment transaction execution, online customer service quiz, policy content inquiry, etc., but not limited to the above.

The dialogue generation unit 243 may include a brand character personality database storing preset answer content and character sentence structure of plural situations and user emotions, so as to provide a variety of assistant character personalities and generate dialogue content, so that users can decide services on their own Way of interaction.

The text-to-speech unit 244 is used to convert any text into a corresponding sound signal output. In this embodiment, the text-to-speech unit 244 can collect and build a brand-identifying voiceprint model for the synthesized voice of a specific character. Outputs sound signals for specific characters.

Furthermore, the dialog financial service device 20 may further include an identity recognition module 25 connected to the control module 22 for checking and verifying the identity of the user. The identity recognition module 25 includes an identity recognition unit 251, a personal database 252 connected to the identity recognition unit 251, a voiceprint recognition unit 253, and a user voiceprint database 254 connected to the voiceprint recognition unit 253.

The personal database 252 is a database of a user's identity verification, account password, and digital credentials, which stores a plurality of financial personal information of the user; the identification unit 251 is based on the content of the voice message (for example, the user speaks himself Account and password) and verify the identity of the user in accordance with the financial personal information of the personal database 252. The user's voiceprint database 254 stores the voiceprint of the voice message each time the user speaks; the voiceprint recognition unit 253 can use the voiceprint extraction method to build a personally identifiable model to check the user's voice in future use.

It is added that the identity recognition module 25 may also be located in another host or server (for example, an existing identity verification server), and the control module 22 may communicate with the host or server through the transceiver module 21, To verify user identity. In addition, the interactive financial service device 20 may also include a personal financial database (not shown) connected to the control module 22 to store user accounts, credits, financial transactions, and other information and records, or through a transceiver module Group 21 communicates, and stores the information and records in a database of another electronic device (for example, an existing personal data server), so the above is not limited to this embodiment.

Therefore, after the user's identity is verified, the control module 22 will transmit the voice message received by the transceiver module 21 to the emotion recognition module 23 and the natural language dialogue module 24. The emotion recognition module 23 will pass the text Recognize the emotion contained in the voice message with the voiceprint, and the natural language dialogue module 24 will judge the content of the voice message, such as: whether a specific financial term is mentioned, or whether it is necessary to perform such as remittance transfer, bill payment, etc. Financial transaction services, and when the emotion recognition module 23 and the natural language dialogue module 24 respectively return their judgment results to the control module 22 for the comprehensive evaluation of the control module 22 and corresponding generation by the dialogue generation unit 243 Reply to the message.

It is specifically stated that the judgment weight of the emotion recognition module 23 will be higher than that of the natural language dialogue module 24, that is, when the emotion recognition result exceeds a preset threshold (Threshold), the emotion processing will be prioritized and the dialogue The default answer content for emotions in the brand character personality database in the generating unit 243 responds. For example, when the user responds "Wow! This is too expensive," the corresponding response message will be "Don't "Worry", when the emotional response priority processing is completed, the corresponding financial transaction service is executed according to the user's intention determined by the natural language dialogue module 24.

With reference to FIG. 2, FIG. 2 is a dialog financial service method according to the present invention. The method can be applied to a dialog financial service system 100 or any system or electronic device that can execute the following processes. The detailed process of this dialogue financial service method is as follows:

In step S10, the user executes a dialogue financial service through an interface of installing / opening an application (APP), social software, webpage, and the like, and starts a dialogue, for example, "Hello, I'm Wang Xiaoming, please help me."

In step S20, the client device 10 converts the content of the conversation into a voice message, and sends it to the conversational financial service device 20.

In step S30, the control module 22 first sends a voice message to the identity recognition module 25 for identity verification. The identity recognition unit 251 can judge the personal information of "Wang Xiaoming", and cooperate with the voiceprint recognition unit 253 to determine the voiceprint of its dialogue to comprehensively check the identity of the user. If the identity is correct, the control module 22 generates Reply to the message, for example: "Hello, Mr. Wang, I am happy to serve you."

In step S40, the user device 10 can then convert the content of the conversation (for example, "how much is my credit card bill this month?") Into a voice message and send it to the conversational financial service device 20.

In step S50, the control module 22 comprehensively evaluates and generates a response message according to the emotion contained in the voice message recognized by the emotion recognition module 23 and the content of the voice message judged by the natural language dialogue module 24, for example: " Your monthly credit card bill is $ 50,000. "

In step S60, it is transmitted back to the client device 10 through the transceiver module 21 to output the reply message. It is specifically noted that the output mode can be not only a voice mode, but also a corresponding text displayed through the interactive interface of the client device 10 at the same time, which is not limited to any mode.

In summary, the conversational financial service system 100 and method of the present invention can use the form of natural language interaction to more conveniently and easily obtain financial-related services, and can solve the problem that current users must maintain their eyes and hands on the physical screen. The trouble of operation, for example, can be used to complete daily financial services in a dialogue mode while driving, so it can indeed achieve the purpose of cost invention.

However, the above are only examples of the present invention. When the scope of implementation of the present invention cannot be limited in this way, any simple equivalent changes and modifications made in accordance with the scope of the patent application and the content of the patent specification of the present invention are still Within the scope of the invention patent.

100‧‧‧ Dialogue Financial Services System

10‧‧‧ end device

20‧‧‧ Conversational Financial Services Device

21‧‧‧Transceiver Module

22‧‧‧Control Module

23‧‧‧Emotion recognition module

231‧‧‧Character recognition unit

232‧‧‧Emotional Character Database

233‧‧‧Voiceprint emotion recognition unit

234‧‧‧emotional characteristics database

24‧‧‧ Natural Language Dialogue Module

241‧‧‧ Language-to-Text Unit

242‧‧‧ Natural Language Understanding Unit

243‧‧‧Dialog Generation Unit

244‧‧‧Text-to-speech unit

25‧‧‧Identification Module

251‧‧‧Identification Unit

252‧‧‧Personal Database

253‧‧‧Voiceprint recognition unit

254‧‧‧User Voiceprint Database

Steps S10 to S60

Other features and effects of the present invention will be clearly presented in the embodiment with reference to the drawings, in which: FIG. 1 is a schematic circuit block diagram of an embodiment of the dialog financial service system of the present invention; and FIG. 2 is a dialog of the present invention Flowchart of a financial services approach.

Claims (10)

A dialog financial service device includes: a transceiver module for receiving a voice message; a control module connected to the transceiver module; a natural language dialog module connected to the control module; To identify the content of the voice message; and an emotion recognition module connected to the control module, the emotion recognition module is used to identify the emotion contained in the voice message, and the control module is based on the emotion recognition module and the nature A response message is generated corresponding to the recognition result of the language dialogue module, wherein the judgment weight of the control module for the recognition result of the emotion recognition module is higher than the recognition result of the natural language dialogue module. The dialog financial service device according to claim 1, wherein the emotion recognition module includes a text recognition unit and an emotional word database connected to the text recognition unit, and the emotional word database stores plurals related to emotional text Vocabulary or sentence pattern, the text recognition unit captures the corresponding speech text in the voice message, and cooperates with the emotion word database to determine the emotional text of the voice message, and generates an emotion probability score. The control module is The emotion contained in the voice message is determined according to the emotional probability score. The dialog financial service device according to claim 2, wherein the emotion recognition module further includes a voiceprint emotion recognition unit and an emotion feature database connected to the voiceprint emotion recognition unit, and the emotion feature database stores a plurality of Voiceprint signal and prediction model related to emotion, the voiceprint emotion recognition unit captures the corresponding voiceprint signal in the voice message, and cooperates with the emotion feature database to judge the emotion of the voice message, and generates an emotion feature value, The control module determines the emotion contained in the voice message according to the emotional probability score value and the emotional characteristic value. The dialogue financial service device according to claim 1, wherein the natural language dialogue module includes a natural language understanding unit and a dialogue generation unit, and the natural language understanding unit is configured to execute corresponding financial services according to the content of the voice message. Service, the dialogue generating unit is used to generate corresponding dialogue content according to the voice message. The dialog financial service device according to claim 4, wherein the natural language dialog module further includes a language-to-text unit and a text-to-speech unit, and the language-to-text unit is used to compare whether the text in the voice message is Matching at least one financial vocabulary, the text-to-speech unit is used to convert text into corresponding sound signal output. A conversational financial service system includes: a consumer device that generates a voice message; and a conversational financial service device that communicates with the consumer device. The conversational financial service device includes a transceiver module that receives the voice message. A control module connected to the transceiver module, a natural language dialogue module connected to the control module, and an emotion recognition module connected to the control module, the natural language dialogue module is used to identify the voice message Content, the emotion recognition module is used to recognize the emotion contained in the voice message, and the control module generates a response message corresponding to the recognition result of the emotion recognition module and the natural language dialogue module, wherein the control module The judgment weight for the recognition result of the emotion recognition module will be higher than the recognition result of the natural language dialogue module. The dialog financial service system according to claim 6, wherein the emotion recognition module includes a text recognition unit and an emotional word database connected to the text recognition unit, and the emotional word database stores plurals related to emotional text Vocabulary or sentence pattern, the text recognition unit captures the corresponding speech text in the voice message, and cooperates with the emotion word database to determine the emotional text of the voice message, and generates an emotion probability score. The control module is The emotion contained in the voice message is determined according to the emotional probability score. The dialog financial service system according to claim 7, wherein the emotion recognition module further includes a voiceprint emotion recognition unit and an emotion feature database connected to the voiceprint emotion recognition unit, and the emotion feature database stores a plurality of Voiceprint signal and prediction model related to emotion, the voiceprint emotion recognition unit captures the corresponding voiceprint signal in the voice message, and cooperates with the emotion feature database to judge the emotion of the voice message, and generates an emotion feature value, The control module determines the emotion contained in the voice message according to the emotional probability score value and the emotional characteristic value. The dialogue financial service system according to claim 6, wherein the natural language dialogue module includes a natural language understanding unit and a dialogue generation unit, and the natural language understanding unit is configured to execute corresponding financial services according to the content of the voice message. Service, the dialogue generating unit is used to generate corresponding dialogue content according to the voice message. The dialog financial service system according to claim 9, wherein the natural language dialog module further includes a language-to-text unit and a text-to-speech unit, and the language-to-text unit is used to compare whether the text in the voice message is Matching at least one financial vocabulary, the text-to-speech unit is used to convert text into corresponding sound signal output.