JP2015127969A - Natural-language banking process server and natural-language banking process method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system enabling a natural-language text input to a mini-blog and the like to be used, as is, for a net transaction with a financial institution such as a bank.SOLUTION: A natural-language banking process server performs: receiving a natural-language text from natural-language input means such as a mini-blog connected to Internet; applying a morphological analysis to the received natural-language text to extract a predetermined keyword by using a keyword DB in which keywords used for transactions with a financial institution is pre-registered; generating requirement data required for a financial transaction on the basis of the extracted keyword and the registered information of a user wishing to do a transaction to send the requirement data to a system of a financial institution connected via a dedicated network; and, when the request for a transaction is complete, generating a natural-language text for responding to the user and responding to the natural-language input means.

Description

  The present invention relates to a natural language banking processing server and a natural language banking processing method for performing banking processing using natural language text input via the Internet.

  In recent years, instead of using a store window or an ATM when performing transactions with financial institutions such as banks, Internet banking has been widely used. However, the existing Internet banking system has different usage methods for each bank, different web form input practices, difficult to find functions that require many screen transitions, and rules such as user IDs and passwords. In many cases, the system is difficult to use unless there is a certain degree of knowledge of the personal computer or the web.

  For this reason, the system which can perform transaction with financial institutions, such as remittance and a loan, is considered, without requiring the input to a web form especially. For example, in Patent Document 1, an automatic remittance server analyzes a tone transmitted from a push button telephone, determines whether or not a loan to a user is possible based on the output digital information and registration information in a database, Disclosed is a system that includes voice generation means for prompting the user to input and announcing the availability of loans, etc., and allows the user to request remittance from anywhere using a mobile phone etc. ing.

  However, even in the method as disclosed in Patent Document 1, if the input method is a push button of a mobile phone or the like, and the response and the operation guide are simply voice output, there is a possibility that the operation will be different for each financial institution. In addition, there is no change in that an input operation is required for each financial institution having an account.

  On the other hand, a mini blog represented by Twitter (registered trademark) is a kind of blog, and is a service for writing on the web what is currently being done, miscellaneous feelings, etc. on the web with simple sentences. Mini blog has instant messenger (IM), social networking service (SNS), chat and other elements, and can easily post a short message of about 100 characters on the web from a personal computer or mobile phone. You can exchange and share information in real time. Since the posted content is a short text, it is easy to update, and as a result, almost real-time communication is often performed, so that the user can feel as if they are chatting. Due to these features, miniblogs are rapidly increasing the number of users because they can easily communicate and communicate information without knowing detailed personal computer knowledge or web page conventions.

JP 2004-199354 A

  In this way, texts in natural language can be easily accessed on the Internet in mini-blogs and the like (hereinafter “mini-blogs” is a generic term including not only mini-blogs but also IM / chat, SNS, and e-mail). What can be input is considered to be a great possibility not only as a means for communicating between humans but also as a means for communicating with a computer (input means and response means). This is because there is a possibility that operations and instructions for various applications on the Internet can be directly performed from a screen such as a miniblog used by the user every day. Therefore, in the present invention, in view of the above-mentioned problems of net banking and the possibility of mini-blogs, etc., natural language texts input to mini-blogs can be used as they are for online transactions of financial institutions such as banks. The purpose is to provide a system.

In order to solve the above problems, the natural language banking system of the present invention provides the following solution.
In a banking system via the Internet, a natural language banking processing server that accepts a processing request in a natural language, the natural language receiving unit receiving natural language text from the natural language input unit connected to the Internet, and the received Analyzing natural language text and using a keyword database in which keywords used for transactions of financial institutions are registered in advance, natural language analysis means for extracting predetermined keywords, and the extracted keywords and users who want to conduct transactions Processing request generation means for generating financial transaction request data based on the registration information and transmitting it to a system of one or a plurality of financial institutions connected via a network, and the natural language analysis means includes the Morphological analysis of received natural language text and parse into words And analyzing unit, on the basis of the keyword database and registered by categorized for each type of terms used the keyword in financial transactions, and having a keyword extraction section that extracts a predetermined keyword.

  According to the present invention, the natural language banking processing server analyzes the natural language text received via the Internet, extracts keywords related to transactions with the financial institution included in the text, and extracts the natural language text from the financial institution. Convert to processing request data for the system. With such a configuration, a user can complete a transaction with a financial institution only by inputting a natural language text from various means on the Internet without having to access each bank's net banking system. In addition, since a keyword database (DB) in which keywords used for transactions of financial institutions are registered in advance is used, by extracting those keywords from natural language text, analysis of natural language text and commands for financial institution systems are performed. Easy to convert to.

  In addition, the natural language banking processing server of the present invention may have a reply message generating means for generating a natural language text for replying to the user when the transaction request is completed. According to such a configuration, it is possible to reply in a natural language even when a transaction request is completed.

  The natural language input means of the natural language banking processing server of the present invention may be any one of a miniblog, an instant messenger, a chat system, and an e-mail.

  According to such a configuration, as a natural language input unit on the Internet, a miniblog represented by Twitter (registered trademark), an instant messenger (IM), a chat system, an SNS system, or an e-mail can be used. it can. Therefore, the user can send a transaction request to the financial institution system as if sending a message to a human using these daily communication tools.

  In the natural language banking processing server of the present invention, the category of the keyword includes a destination keyword that specifies a transfer destination account, a command keyword that represents a transfer request or an inquiry request in a financial transaction, and an amount keyword category that represents an amount. You may make it contain at least.

  According to such a configuration, keywords used for financial institution transactions are stored separately in categories that indicate whether they are transfers or inquiries, transfer destinations, transaction amounts, etc. By determining which category the word in belongs to, it becomes easier to create data that constitutes a command for a financial institution system.

  In addition, the natural language banking processing server of the present invention, when an amount is registered in the transfer destination information corresponding to the destination keyword at the time of the transfer request, the registered amount as a monthly fixed transfer amount The included request data may be transmitted to the financial institution system.

  According to such a configuration, when a certain amount of money is to be transferred every month, such as rent or parking fee, a certain amount of money is transferred only by entering the destination keyword without entering the amount in text every time. Can be performed.

  Further, the natural language banking processing server of the present invention compares the inquiry request with the history information that has been referred to in the past at the time of the inquiry request, creates a difference between the two as new inquiry request data, and You may make it transmit to a system.

  According to such a configuration, since the inquiry history is stored in the database when inquiring usage details, etc., a new inquiry request is issued again for those parts that can be handled by previously inquired data. No, a new inquiry request is issued for only the part (difference) not in the inquiry history, and the part in the history and the newly inquired result are combined and returned to the user. By doing so, it is possible to reduce the load of inquiries to the financial institution system.

  In the natural language banking processing server of the present invention, the transfer destination corresponding to the destination keyword includes information on a predetermined fund-raising account, and transfers the fund for fund-raising to other users of the natural language input means. You may make it go to the fundraising account.

  According to such a configuration, the present system can be used as a means for collecting a donation by registering a bank account corresponding to the destination keyword as an account for a predetermined donation and disclosing it to other users. The user who makes a donation need only input the destination keyword and the amount registered in the system in a natural language. This is particularly effective when the natural language input means on the Internet provides means for communicating with many other users.

Further, the natural language banking processing server of the present invention can also be regarded as a natural language banking processing method in net banking on the Internet as described below, and can exhibit the same effects as described above.
A natural language banking processing method in a system in which a natural language input means existing on the Internet, a natural language banking processing server for processing a natural language, and a financial institution system are connected, the natural language input means comprising: A step of transmitting a natural language text input by a user for a financial transaction instruction; a step of receiving the transmitted natural language text by the natural language banking processing server; and a morphological analysis of the received natural language text. A step of extracting a predetermined keyword based on a keyword database in which keywords are categorized and registered for each type of term used in financial transactions, and the extracted keyword and transaction are performed. Request data for financial transactions based on user registration information And having a step of transmitting to the system of the connected one or more of the financial institutions over click, the.

  ADVANTAGE OF THE INVENTION According to this invention, the system which can utilize the natural language text input into the miniblog etc. as it is for the net transaction of financial institutions, such as a bank, can be provided. The user does not need to learn the operation procedure on the web page of the financial institution, and can give instructions such as transfer from the miniblog or the like that is always used with little screen transition.

It is a figure which shows the system configuration | structure and functional block of the natural language banking processing server 10 which concern on the 1st Embodiment of this invention. It is a figure which shows the image which converts a natural language text into the request | requirement (command) to a bank system. It is a figure which shows the image which converts another natural language text into the request | requirement (command) to a bank system. It is a figure which shows the outline of keyword DB12d which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the transfer destination registration table 14 which concerns on the 1st Embodiment of this invention. It is a figure showing an example of transfer history DB16 concerning a 1st embodiment of the present invention. It is a figure which shows the transfer processing flow by the natural language banking processing server 10 which concerns on the 1st Embodiment of this invention. It is a figure which shows the example of the transfer transmission message and reply message which concern on the 1st Embodiment of this invention. It is a figure which shows the system configuration | structure and functional block of the natural language banking processing server 10 which concern on the 2nd Embodiment of this invention. It is a figure which shows addition of the inquiry keyword to keyword DB12d which concerns on the 2nd Embodiment of this invention. It is a figure which shows an example of inquiry log | history DB17 which concerns on the 2nd Embodiment of this invention. It is a figure which shows the inquiry processing flow by the natural language banking processing server 10 which concerns on the 2nd Embodiment of this invention. It is a figure which shows the example of the inquiry transmission message and reply message which concern on the 2nd Embodiment of this invention.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention (hereinafter referred to as embodiments) will be described in detail with reference to the accompanying drawings. Note that the same number is assigned to the same element throughout the description of the embodiment.

  FIG. 1 is a diagram showing a system configuration and functional blocks of a natural language banking processing server 10 according to the first embodiment of the present invention. The financial institution system handled by the present invention will be described below using a bank system as a representative example. This system is dedicated to the natural language banking processing server 10 which is the core of the system, a natural language input means 20 such as a miniblog connected to the natural language banking processing server 10 via the Internet 19a, and the natural language banking processing server 10. It comprises a conventional bank system 30 or a plurality of bank systems 30 connected via a network 19b. The natural language banking processing server 10 is typically considered to be provided by a service company established by a bank or banking organization, but the banking system 30 connected to the server is not limited to the banking system of the same family. Other systems may also be used. The difference in interface with each bank system 30 is absorbed by the natural language banking processing server 10.

  The natural language banking processing server 10 is analyzed by a natural language receiving means 11 for receiving a natural language text input to a miniblog or the like via the Internet 19a, and a natural language analyzing means 12 for analyzing the received natural language text. From the processing request generation means 13 that generates and transmits a request (command or command) to the bank system 30 from the natural language text, the transfer destination registration table 14 that stores information of the transfer destination registered in advance by the user, and the bank system 30 The reply message generation means 15 for generating a reply message to the user in a natural language text, the transfer history DB 16 for storing the history information of the completed transfer, and the generated reply Reply message sending means 1 for sending a message to a natural language input means 20 such as a miniblog It consists of.

  Further, the natural language analysis means 12 morphologically analyzes the natural language text into words and performs morphological analysis (Morphological Analysis) for discriminating each part of speech. The word dictionary 12b, a keyword extraction unit 12c for extracting keywords for bank transactions such as transfer / inquiry, and a keyword DB 12d storing the keyword group. For the morpheme analysis unit 12a and the word dictionary 12b, a known morpheme analysis technique such as KAKASI, “Mekabu”, or “tea bowl” may be used.

  The keyword extraction unit 12c extracts keywords used for banking transactions from the natural language sentences analyzed by the morphological analysis unit 12a. The keyword to be extracted is a group of words that may be used for bank transactions, which will be described in detail later, and is registered in advance in the keyword DB 12d. For example, if the natural language text includes keywords such as “transfer” and “remittance”, it can be estimated that the message is a request for transfer to a predetermined transfer destination.

  The processing request generation unit 13 generates a processing request command (command) to the bank system 30 from the extracted keyword. The command includes data necessary for processing. For example, if the processing request is “transfer”, a word that can uniquely identify the transfer destination (hereinafter referred to as “destination keyword”) is searched in natural language text, and the destination keyword is registered as the transfer destination. It is checked whether it exists in the table 14. The destination keyword may be an arbitrary character string in addition to the personal noun, and the character string is linked to the bank account information of the actual transfer destination in the transfer destination registration table 14. If the destination keyword exists in the transfer destination registration table 14, the account information associated with it is included in the transfer request data for the bank system as the transfer destination.

  When the reply message generation means 15 receives the processing completion notification from the bank system 30, when generating a reply message in a natural language, the reply message generation means 15 displays the completed transfer information together with the text information when the transfer request is issued. To store. By leaving the transfer request completed in this way in the history, it can be used for later inquiry processing. The reply message is generated by embedding information such as a destination keyword and a transfer amount in a natural language reply message template (sentence pattern) prepared in advance. The reply message transmission unit 18 transmits the reply message generated by the reply message generation unit 15 to the natural language input unit 20 such as a miniblog via the Internet.

  The functional configuration described above is merely an example, and one functional unit or functional unit may be further divided, or a plurality of functional units may be configured as one functional unit. In each functional unit, a CPU (Central Processing Unit) built in the server device reads a computer program stored in a storage device such as a ROM (Read Only Memory) or a hard disk, and the computer program executed by the CPU Read and write necessary data such as tables from a database (DB; Data Base) stored in a storage device or a storage area on a memory, and in some cases, related hardware (eg, input / output devices, display devices, communication) This is realized by controlling the interface device. In addition, the database (DB) in the embodiment of the present invention may be a commercial database, but it simply means a collection of tables and files, and the internal structure of the database itself does not matter.

  FIG. 2 is a diagram showing an image of converting a natural language text into a request (command) to the bank system. As shown in the figure, the natural language text 40 shows a simple sentence “send 10,000 yen to the dad!” As an example. The morpheme analysis unit 12a sends this sentence to “dad” (noun), “ni” (case particle), “10,000 yen” (numerical), “send” (verb command form), “!” (Symbol). To break it up into words (morphemes). Among these, “dad”, “10,000 yen”, and “send” are transfer keywords registered in the keyword DB 12d. The transfer keyword is defined in advance as a keyword that is often used for transfer in a natural language sentence. When the keyword extraction unit 12c finds the transfer keyword from the words subjected to morphological analysis, the keyword extraction unit 12c checks the relationship before and after the keyword. Since “send” is registered as a term representing a command (transfer) in the keyword DB 12d, it can be detected first. Here, “!” Is defined in advance as an identifier indicating the end of an instruction (instruction end identifier). The presence of an instruction termination identifier makes it easy to identify the type of command. In this way, it can be determined that the natural language text 40 is for making a transfer request. In particular, “send” is called a transfer command keyword. Note that the part of speech that is a command keyword is not limited to a verb but may be a noun.

  Next, since “dad” is registered as “destination keyword” in the keyword DB 12 d and accompanied by “ni” (case particle) representing the target object behind it, the transfer destination identifier Can be determined. The same is true even if "Dad" is followed by another case particle "He". Further, “10,000 yen” is a number representing the amount of money (yen) in the keyword DB 12d, so it is also called an amount keyword and is determined to be a transfer amount.

  By transmitting the analysis result as described above to the processing request generation unit 13, the processing request generation unit 13 can generate the transfer request data 60. That is, as illustrated, bank account data (bank code, store number, account type, account number) of the transfer source is generated from the user registration information 50. Similarly, bank account data (bank code, store number, account type, account number) of the transfer destination is generated from the transfer destination registration table 14. And if the amount keyword (numerical) registered in keyword DB12d is detected in the natural language text 40 analyzed previously, the number in it will be identified as a transfer amount.

  Here, even if there are a plurality of numerics, such as “Send 10,000 yen on March 3!”, Etc., it is accompanied by a “yen” or “¥” symbol, etc. Is determined as an amount. In addition, even if there are multiple amounts of money with a complicated syntax such as “If your account balance is 20,000 yen or more, send 10,000 yen!”, The command form of the registered verb Whether or not it is a transfer amount can be distinguished by whether or not it is immediately followed. In the case of this example sentence, “10,000 yen” following “Send me!” Is determined as the transfer amount. However, the extent to which such a complicated syntax with a condition is supported depends on the design of the keyword extraction unit 12c and the processing request generation unit 13.

  FIG. 3 is a diagram showing an image of converting another natural language text into a request (command) to the bank system. In this example, “March parking charge transfer #!” Is shown as the natural language text 40. In this case, however, “#!” Is defined as the instruction end identifier. In this example sentence, “transfer #!” Is a transfer command keyword indicating a transfer request to the bank system 30. However, although “parking lot” is the destination keyword of the transfer destination, there is no number as an amount keyword indicating the transfer amount. “March” is a number, but does not accompany “yen” or “¥”, so it is not recognized as an amount keyword, but is recognized as “other keywords”. “Other keywords” are often used for condition determination when generating a transfer instruction, but are also stored in the remarks column of the transfer history DB 16.

  When the amount keyword is not found in the sentence as in this example sentence, the keyword extracting unit 12c or the processing request generating unit 13 checks whether or not the fixed amount information is included in the destination keyword record of the transfer destination registration table 14. . In this example, since the “parking lot” is a fixed amount fixed every month, it can be included as fixed amount information such as 15,000 yen in the destination keyword record. However, if “March parking fee for 20,000 yen transfer #!” Is entered, “20,000 yen” in the input text has priority over the fixed amount information of 15,000 yen as the transfer amount. It may be done.

  FIG. 4 is a diagram showing an overview of the keyword DB 12d according to the first embodiment of the present invention. The keyword DB 12d is a database that stores a keyword group that may be used for transaction processing such as a bank from natural language text. Here, only keywords related to transfer are shown. As shown in the figure, keywords are classified into several categories. The categories include a destination keyword 100a that is a keyword that defines a transfer destination, a type keyword 100b that indicates the type of transfer, a financial institution keyword 100c that is a list of financial institutions and stores, an amount keyword 100d that indicates a transfer amount, and the like for the bank system. An instruction keyword (100e) for defining an instruction, a date keyword 100f representing a date and time, and a period are classified. In addition, a keyword often used in condition determination may be defined as a condition keyword.

  Among these, the command keyword 100e is a collection of keywords for defining a processing request (command) to the bank system, and is particularly important. Therefore, in order to distinguish from other keywords, it is desirable to define it together with the instruction end identifier described above. The instruction end identifier may be an arbitrary symbol or a combination thereof in addition to “!” And “#!” Already given as examples. In the example of the figure, as a keyword indicating a transfer command (transfer command keyword), a noun related to a transfer operation such as “send money”, “transfer”, “send”, or a verb command form or a continuous form is shown. . Instruction keywords other than transfer will be described in the second embodiment.

  The destination keyword 100a is a collection of keywords that can uniquely indicate a business partner such as a transfer destination. In this example, “Father” (father), “Grandma” (mother), “Taro” (eldest son), “Hanako” (eldaughter) and other personal nouns, “rent”, “parking fee”, “electricity fee” ”,“ Gas fee ”and the like are included in the destination keyword category because the transfer destination can be uniquely defined. In the latter case, when there are a plurality of “rents” or the like, they may be defined as “rent 1” or “rent 2”.

  The type keyword 100b is not a command for the bank system, but is used to classify the type of transfer. The financial institution keyword 100c is a collection of keywords of all bank names and branch names provided from the financial institution. The amount keyword 100d is a collection of keywords (mainly numbers) representing the amount used for the transaction. The date keyword 100f is a collection of keywords related to the date related to the transfer. For example, it is used as “March” for “Parking fee for March, transfer #!” Or “This month” for “Pay this month's rent #!” As described in FIG. Information (conditional keywords), or written in the remarks column of transfer information for later inquiries. It should be noted that the keywords shown in FIG. 4 are only a part, and various other keywords can be added.

  FIG. 5 is a diagram showing an example of the transfer destination registration table 14 according to the first embodiment of the present invention. As illustrated, the transfer destination registration table 14 is a table for obtaining account information of a transfer destination from the destination keyword 200a. In this example, the account information 200b is associated with “Grandma”, “Taro”, “Hanako”, and the like. As described above, the destination keyword may be a general noun such as “rent” in addition to the personal noun. Of course, it is also possible to transfer money to a destination not in the transfer destination registration table 14. In that case, the account information of the transfer destination is described in the input text. However, in this case, the destination keyword can be composed of a specific character string and used like an encryption. For example, a specific character string to be input is defined to be a true destination keyword when converted according to a predetermined rule. In a simple example, if the bank account is to be transferred to “A bank B branch ordinary account number 1234567”, the bank name and branch name are exchanged, the account number is inverted, and “B bank A branch ordinary account number 7654321” is input. Etc. By doing so, even if the user ID and password of a miniblog, etc. are leaked, unless this encryption rule is known, no transfer is performed illegally. As described above, since the transfer instruction can be issued with a character string, the possibility of unexpected security measures such as encryption is expanded.

  In the table of FIG. 5, in the remittance amount 200c, when the transfer amount can be fixed by the destination, the amount is described. The fact that the remittance amount 200c is blank indicates that the transfer amount cannot be determined from the destination keyword, so that the amount keyword must be included in the natural language input text. The remittance amount 200c is not necessarily one amount, and a plurality of amounts can be defined for each type keyword. For example, as shown in the figure, the amount transferred to “Grandma” is fixed as 50,000 yen when the type keyword is “celebration” and 10,000 yen when the type keyword is “birthday”. Further, as shown by the transfer designation date 200d, the transfer can be executed on a fixed date every month. By doing so, it is possible to instruct the transfer request for the rent and the parking fee with a very short text (in the shortest example, the transfer can be completed only with “rent this month, rent!”).

  FIG. 6 is a diagram showing an example of the transfer history DB 16 according to the first embodiment of the present invention. As shown in the figure, the transfer history DB 16 is a table in which the amount transferred to the transfer destination indicated by the destination keyword 300b and the transfer execution date are recorded in time series. In addition, detailed information at the time of transfer may be recorded. By leaving such a detailed history, it is useful for later inquiries and for investigating the cause of a failure. In the table of FIG. 6, in addition to the transfer date 300a, the destination keyword 300b, the transfer destination account 300c, and the transfer amount 300d, other related information included in the natural language input text at the time of the transfer request is displayed in the remarks column 300e. It shows adding. Further, when an inquiry is made later, an inquiry flag 300f indicating whether or not the information is referred to may be added. The operation of the system when an inquiry request is made will be described in detail in the second embodiment.

  FIG. 7 is a diagram showing a transfer processing flow by the natural language banking processing server 10 according to the first embodiment of the present invention. First, in step S10, the natural language input means 20 such as a mini blog transmits a message input in the natural language to the natural language banking processing server 10 (hereinafter simply referred to as the server 10 in this figure). Here, for example, when the natural language input means 20 is Twitter (registered trademark), the direct message (DM) is used for transmission to a predetermined account defined uniquely for the server 10. When the natural language input means 20 is an e-mail, the e-mail is transmitted to a predetermined e-mail address unique to the server 10. Further, when the natural language input means 20 is a chat system, it is transmitted to a predetermined chat ID unique to the server 10. The same applies to the other natural language input means 20.

  Upon receiving this natural language message in step S11, the server 10 checks in step S12 whether the sender's ID is in the user registration of this service. If there is a user registration, the process proceeds to step S13. . If there is no user registration, a predetermined error message is returned to the natural language input means 20, and the process is terminated. Next, in step S13, the server 10 performs natural language analysis on the received message. Specifically, as described above, the natural language analysis is performed by the natural language analysis unit 12 using the morpheme analysis unit 12a to decompose words into words that form sentences and determine the part of speech of the words and the dependency relationship between the words. analyse. Further, the keyword extraction unit 12c extracts keywords necessary for issuing the subsequent transfer instructions (step S14).

  If the transfer command keyword exists as a result of the keyword extraction (step S15; Yes), the process proceeds to step S16, and the transfer destination registration table 14 is read. Then, transfer request data is generated using the extracted keyword and transfer destination registration table 14 (step S17), and transmitted to the bank system 30 (step S18). In step S15, when the transfer command keyword cannot be detected (step S15; No), the process jumps to step S14a in FIG.

  In the bank system 30, when the transfer request data is received from the server 10 in step S19, whether or not the request data has an error and the request source registration data are authenticated. If the authentication is OK, the transfer request is executed (step S20). In step S <b> 21, when the bank system 30 normally completes the transfer process, a completion notification is transmitted to the server 10. When the balance is insufficient, a predetermined error notification is transmitted to the server 10.

  When the server 10 receives the completion notification in step S22, the server 10 generates a reply message (step S23) and transmits the reply message to the natural language input means 20 (step S24). Here, the reply message is generated by combining the data included in the completion notice received from the bank system 30 and the keyword information when the transfer request data is created with a pre-registered reply message template. Created in language. At this time, the transfer information is also recorded in the transfer history DB 16 (not shown). Finally, the natural language input means 20 recognizes that the transfer has been completed normally by receiving a reply message in step S25.

  FIG. 8 is a diagram illustrating an example of a transfer transmission message and a reply message according to the first embodiment of the present invention. This figure shows a case where the direct message (DM) function of Twitter (registered trademark) is used as the natural language input means 20. However, it is assumed that the user has already registered for using this system. As shown in the drawing, @autoFurikomi is an account for sending a message to the natural language banking processing server 10, and @hasegawa is this user's account. A direct message (DM) is a message that cannot be viewed by a third party to a specific user who is following (registering the user so that other users' posts can be displayed on his timeline). This is a direct sending function. Usually, DMs do not necessarily follow the person who is sending the message, but they must follow each other to use this system. And However, even if a direct message (DM) is not used, a tweet (a sentence of 140 characters or less posted by the user, each assigned a unique URL) can be used privately. .

  In addition, taking advantage of the characteristics of Twitter (registered trademark), a specific bank account number can be disclosed via a tweet and used for collecting funds and collecting funds. In this case, the follower issues a transfer request using this system, but the destination keyword may be registered on the user side who collects the donation so that the transmission message of the transfer can be simplified as much as possible. For example, if you register the term “XX donation” as a destination keyword, the person who makes the donation simply tweeted “1000 yen for XX donation!” To @autoFuromi and easily registered to the account registered for XX donation. Can be transferred to.

  FIG. 8 shows an example of several transmission messages (400 to 404) for a transfer request and an example of a reply message generated by the natural language banking processing server 10 for each transmission message. For example, the transmission message 400 indicates that the transfer can be completed simply by inputting “Send to Taro!”. As described above, when the natural language banking processing server 10 detects that the word “Taro” is the destination keyword and that the sentence representing the amount keyword is not included in this sentence, the transfer destination registration table 14 Then, the bank account information of “Taro” and the fixed amount (for example, 100,000 yen) are read out, transfer request data for the bank system 30 is created, and transmitted to the bank system 30.

  When the natural language banking processing server 10 receives the completion of the transfer request from the bank system 30, it generates a reply message 500. Here, some sentence patterns are prepared in the reply message text, and when the user's profile information, transfer destination information, transfer contents information, etc. can be obtained from natural language input text, etc. The reply message pattern may be changed according to the information. By doing in this way, the user can feel the friendliness like a member of a follower with respect to this system instead of a tasteless reply message. For example, in another example in the figure, the transmission message 401 includes information such as “grandma” and “congratulations of Kouki”, and therefore, in the reply message 501, a text suitable for it is selected. Creating such a “warm” reply message can be realized by the reply message generating means 15 having a simple artificial intelligence function.

  Note that this is the same in the other reply message examples (502 to 504) in the figure, but not only the contents of the transmitted message are repeated, but also a measure such as conversion to more easily understandable information for confirmation. It can also be included. For example, in the reply message 502, “current month” in the transmission message is not returned as a parrot, but the date when the transfer request is made is determined and replaced with “March”. In addition, “Rent” in the transmission message is replaced with “Rent ¥ 80,000” and returned. In addition, a transfer date can be specified as in the transmission message 403 (in this example, “March end date” is specified). Of course, there may be a reply message pattern that enumerates only necessary items, such as the reply message 504. Although not shown in the figure, when there are a plurality of registered accounts of the transfer source of the user, it can be specified from which account the transfer is made. In this case, a transmission message such as “From account 1 to ~ yen transfer!” May be input.

  FIG. 9 is a diagram showing a system configuration and functional blocks of the natural language banking processing server 10 according to the second embodiment of the present invention. In the second embodiment, not only a transfer request to the bank system but a case of an account balance or usage statement inquiry request will be described. However, since the functional block for the inquiry process is basically not significantly different from the functional block for the transfer process described with reference to FIG. 1, only the parts different from FIG. 1 will be described below.

  The difference between FIG. 9 and FIG. 1 is that the database (table) to be referred to (read) when the processing request generation unit 13 generates the inquiry request data is not the transfer destination registration table 14 but the transfer history DB 16, and the reply message. The database that the generation means 15 reads / writes (updates or references) is the inquiry history DB 17. The reply message generation means 15 may also refer to the transfer history DB 16. In the transfer history DB 16, all text information when transferred by this system is recorded. The reason why the processing request generation unit 13 or the reply message generation unit 15 refers to the transfer history DB 16 at the time of inquiry is to enable reference to detailed text information transferred in the past. The reason why the processing request generation unit 13 also refers to the inquiry history DB 17 is because information that has been inquired in the past may be reused.

  In normal net banking, the inquiry conditions for referencing past usage details can only be selected from a menu prepared by the system. For example, “for this month”, “for last month”, “from XX month to XX month”, “unchecked cases”, “latest 10 items”, and the like. However, in this system, it is possible to create an arbitrary inquiry condition (search condition) by taking advantage of the fact that all transfer requests and inquiry requests are made in text. For example, “Check all payments over 10,000 yen!”, “Total the total amount of electricity, gas, and water paid last month!”, “What is the total credit card usage over the past year?” Free inquiry is possible.

  Although another specific example is also shown in FIG. 13 described later, the reason why this is possible is that detailed text information at the time of transfer is stored in the transfer history DB 16. In addition to responding to requests such as querying only unqueried information, refer to past query history at the time of a query request as well as some previously queried information in this query request. If it is, it is possible to reduce the query load by using it. For example, if the current inquiry is a “use statement for the past one year” and if the inquiry history up to the previous month has already remained in the inquiry history DB 17, only the inquiry for the current month is referred to the bank system 30 and the previous month. What is necessary is just to synthesize | combine with the data of a minute. Needless to say, such detailed personal information is subjected to sufficient security management to prevent unauthorized leakage.

  FIG. 10 is a diagram showing addition of a query keyword to the keyword DB 12d according to the second embodiment of the present invention. In FIG. 4 which has already been described, keywords related to the transfer process are classified by category. In FIG. 10, a keyword related to “inquiry” (hereinafter sometimes referred to as “inquiry keyword”) is added to these keywords. ing. In the figure, the underlined words are added for inquiry processing. For example, “usage details” and “account balance” are added to the type keyword 100b. In the amount keyword 100d, not only a word including “yen” but also a word representing a condition such as “above” or “below” is added. “?” Is added to the instruction termination identifier.

  Further, as a matter of course, a term (inquiry instruction keyword) that implies an inquiry or a search is added to the instruction keyword 100e. Furthermore, words representing the past, such as “last month”, “last month”, “last year”, “last week”, and “last week”, are added to the date keyword 100f. Both are terms used for inquiry (search) conditions, not at the time of transfer. The keyword DB 12d can be adapted to other processes simply by adding such keywords. For example, in addition to bank transfer and inquiry, if a keyword such as “registration”, “change”, “stop” is added, a function that can partially change or stop registration information for this service can be added. Of course, the keyword itself may be added. In this way, it becomes possible to make a request to the system using words that are usually used by the user. The keyword to be added may be a word unique to a certain region or community, such as dialect or gal.

  FIG. 11 is a diagram showing an example of the inquiry history DB 17 according to the second embodiment of the present invention. Although the inquiry history DB 17 has already been described, the inquiry history DB 17 is a record of the history inquired by the user in the past. The history information includes a transaction date 600a, a transaction type 600b, a customer 600c, a transaction amount 600d, a transfer history 600e, an inquiry history 600f, and the like as shown in the figure. The case where the inquiry history 600 f is “present” indicates that the data has been inquired in the past, and if there is a further inquiry request, the inquiry request is not sent to the bank system 30. A reply can be made from this history information. The reason why the transfer history 600e is provided is to include the text information used at the time of transfer in the transaction with "Yes" in the transfer history 600e. Alternatively, a link may be made to text information at the time of transfer. By doing in this way, the text information at the time of transfer can be used for an inquiry and a search.

  FIG. 12 is a diagram showing an inquiry processing flow by the natural language banking processing server according to the second embodiment of the present invention. This figure shows processing performed by the natural language banking processing server 10 when receiving an inquiry processing request. Steps S10 to S13 are the same as those in FIG. Steps S14a and after are different parts in the inquiry process, but from the case where the natural language message does not include the transfer command keyword in FIG. 7 (step S15; No), it may be considered that it is connected to step S14a in this figure. .

  The natural language banking processing server 10 analyzes the received natural language message, and if there is no inquiry keyword in step S14a, returns an error message to the natural language input means 20 assuming that there is no instruction keyword to support. If there is a query keyword, the process proceeds to step S15a to check whether there is a query history. Specifically, the inquiry history DB 17 is read to check whether there is a past inquiry history having the same keyword. If there is a past history, the difference between the past inquiry history information and the current inquiry request is extracted in step S16a. For example, if the current inquiry is “use details for the past one year” and the inquiry history DB 17 already has the inquiry history up to the last month, only the inquiry for this month is inquired to the bank system 30 and the last month's amount is inquired. Combine with the data.

  Next, in step S17a, if there is history data in the past, a part or all of the data is used, and if there is no history data in the past, new inquiry request data is generated. Then, the inquiry request data generated in this way is transmitted to the bank system 30 (step S18a). When the bank system 30 receives the inquiry request data, it executes it and sends a completion notice (steps S19a, S20a, S21). Subsequent processing (steps S22 to S25) is the same as that in FIG.

  FIG. 13 is a diagram illustrating an example of an inquiry transmission message and a reply message according to the second embodiment of the present invention. This figure shows an example of a typical transmission message and reply message in response to an inquiry request. The transmission message 405 is an account balance inquiry request. A reply message 505 is sent back together with the date and time when the user's bank account reference is completed as shown in the figure. When the user has a plurality of bank accounts, a keyword that can identify the bank account is included in the transmitted message. If you want to check the balance of all bank accounts, specify "All".

  The transmission message 406 is a usage detail inquiry request. At the time of use statement inquiry, the period is usually designated, but in this example, “from January 1 to today” is designated. If no period is specified, a predetermined default period is used.

  The transmission message 407 is an inquiry request for collecting past electricity charges. The natural language banking processing server 10 searches the transfer history DB 16 for the past one year based on the keyword “electricity bill”, and extracts and tabulates corresponding data. In this case, only transfer using this system can be counted, but in the case of automatic debit such as account transfer, it is also possible to count using another keyword. For example, “Tokyo Denryoku”, “Denkidai”, and the like are also included in the automatic debit account debit information, and can be counted as keywords.

  The transmission message 408 is an example to which inquiry conditions “1 year”, “100,000 yen or more”, and “payment” are added. Since the keywords “1 year”, “above”, and “payment” are included in the keyword DB 12d described with reference to FIG. 10, such a search is also possible. Specifically, from the usage details for the past year, the transaction type extracts the portion of payment (transfer), and further collects transactions with a transaction amount of 100,000 yen or more. If all payments are from the system, it is only necessary to search the transfer history DB 16 without making an inquiry request to the bank system 30. As described above, the banking system using the natural language of the present invention enables a flexible inquiry not found in the conventional net banking.

  In the first and second embodiments of the present invention described above, the bank system is used as the destination for giving instructions in the natural language, and the transfer request and the inquiry request are taken as specific examples. However, the present invention is other than the transfer and inquiry. Obviously, it can be applied to other services of banks (for example, requesting materials, changing account information, inquiries, etc.). Moreover, it can be easily applied not only to banks but also to systems of other financial institutions. For example, when applied to a system of a credit card company, the inquiry about the available balance of the card and the inquiry of the card usage details are almost the same as in the case of the bank system. In addition, it is possible to easily respond to requests such as application for personal loans, transfer of cash repayment, and change to revolving payment.

  As mentioned above, although this invention was demonstrated using embodiment, it cannot be overemphasized that the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above-described embodiments. Further, it is apparent from the scope of the claims that the embodiments added with such changes or improvements can be included in the technical scope of the present invention.

DESCRIPTION OF SYMBOLS 10 Natural language banking processing server 11 Natural language reception means 12 Natural language analysis means 12a Morphological analysis part 12b Word dictionary 12c Keyword extraction part 12d Keyword DB
13 processing request generation means 14 transfer destination registration table 14a account registration table 15 reply message generation means 16 transfer history DB
17 Inquiry history DB
18 Reply message transmission means 19a Internet 19b Dedicated network 20 Natural language input means 30 Banking system 40 Natural language text 50 User registration information 60 Transfer request data 100a to 100f Category classification in keyword DB 200a to 200d Transfer registration table items 300a to 300f Transfer history DB item 400-408 Transmission message 500-508 Reply message 600-600f Query history DB item

Claims (8)

In a banking system via the Internet, a natural language banking processing server that accepts processing requests in natural language,
Natural language receiving means for receiving natural language text from the natural language input means connected to the Internet;
A natural language analyzing means for analyzing the received natural language text and extracting a predetermined keyword using a keyword database in which keywords used in transactions of financial institutions are registered;
A processing request generation means for generating request data for financial transactions based on the extracted keyword and registration information of a user who intends to conduct a transaction, and transmitting the data to one or more systems of the financial institutions connected via a network When,
With
The natural language analysis means is based on the morphological analysis unit that morphologically analyzes the received natural language text and decomposes it into words, and the keyword database in which keywords are categorized and registered for each type of terms used in financial transactions. A natural language banking processing server, comprising: a keyword extracting unit that extracts the predetermined keyword.   The natural language banking processing server according to claim 1, further comprising a reply message generation unit configured to generate a natural language text for reply to the user when the transaction request is completed.   The natural language banking processing server according to claim 1, wherein the natural language input means is one of a miniblog, an instant messenger, a chat system, and an e-mail.   The keyword category includes at least a destination keyword for specifying an account of a transfer destination, an instruction keyword indicating a transfer request or an inquiry request in a financial transaction, and an amount keyword category indicating an amount. 4. The natural language banking processing server according to any one of 3 above.   When an amount is registered in the transfer destination information corresponding to the destination keyword at the time of the transfer request, request data including the registered amount as a fixed amount of monthly transfer is transmitted to the financial institution system. The natural language banking processing server according to claim 4, wherein:   The inquiry request is compared with historical information that has been inquired in the past, the difference between the two is created as new inquiry request data, and is transmitted to the system of the financial institution. 4. The natural language banking processing server according to 4.   The transfer destination corresponding to the destination keyword includes information on a predetermined donation account, and causes other users of the natural language input means to perform a transfer for donation to the predetermined donation account, The natural language banking processing server according to claim 5 or 6. A natural language banking processing method in a system in which a natural language input means existing on the Internet, a natural language banking processing server for processing natural language, and a financial institution system are connected,
The natural language input means transmits a natural language text input by a user for a financial transaction instruction;
The natural language banking processing server is
Receiving the transmitted natural language text;
Extracting the predetermined keyword based on a keyword database registered by categorizing and classifying the received natural language text into words by morphological analysis and keywords for each type of terms used in financial transactions;
Generating request data for financial transaction based on the extracted keyword and registration information of a user who intends to conduct a transaction, and transmitting the data to a system of one or more financial institutions connected via a network;
A natural language banking processing method characterized by comprising:

Images