DE102005037621A1 - Method and speech dialog system for determining at least one transaction for operating a background application - Google Patents

Abstract

By evaluating at least one action information item (AI) contained in a user utterance (BE), a speech dialog system (1) is used to determine at least one current transaction (T1-Tx) related transactions (T1-Tx) in the speech dialog system (1) at least one matching context information (KI) is assigned. The context information (KI¶letzt¶) of at least the most recently executed transaction (T¶letzt¶) is stored in the speech dialog system (1) and when determining the current transaction (T1-Tx) those transactions (T2, T3) are preferably taken into account whose assigned context information (KI) matches the stored context information (KI¶letzt¶).

Description

The The invention relates to a method and an associated speech dialogue system for Determining at least one transaction for operating a background application under evaluation of action information contained in a user statement in such a speech dialogue system, each with a background application a finite set of selectable ones Transactions is assigned and with which over one in the speech dialogue system provided input unit detects at least one user utterance and stored in the speech dialogue system.

IVR Systems for database access, which information accesses as well as a control of communication applications enable via voice communication, are known as interfaces to many computer-aided applications. Applications or background applications, e.g. a technical Device of Consumer electronics, a telephone information system (train, Flight, cinema, etc.), a computerized transaction system (home banking system, electronic goods order, etc.) are increasingly over such Speech dialogue systems as access systems ("user interfaces") operable Speech dialogue systems can realized in hardware, software or a combination thereof become.

Of the Dialogue course to achieve application-specific dialogue goals will over controlled such a speech dialogue system, which the interactions between a dialogue management unit and the individual user. This is indicated by the dialogue management unit an input unit and output unit.

A for example, user generated user utterance a speech signal is detected by the input unit and stored in saved the dialogue management unit. The input unit can For example, be followed by a voice recognition unit, via which a determination of action information contained in a user utterance carried out becomes. To output action prompts in the form of speech signals the output unit can be designed as a speech synthesis unit.

About the provided in the dialogue management unit speech recognition unit For example, action information will become out of the speech signal in the form of individual words or words chains obtained by comparison with in a parser loaded keywords or grammars are evaluated. Depending on the evaluation result becomes a transaction associated with one or more key terms started to operate a background information. Depending on the transactions will be over the dialogue management unit a dialogue history with the respective user to control the associated Background application added.

Currently are three approaches for the realization of such speech dialogue systems, namely the "Finite State Based" approach, the "Frame Based" approach and the "Agent Based" approach, the "Frame Based" or frame oriented Approach is currently the most widely used in practice. Since the dialogue process of frame-oriented speech dialogue systems not organized on a hierarchical tree structure this one big one Flexibility. Rather, each transaction is modeled on a framework that has multiple ones Gaps which are filled by the assignment of transaction parameters.

In order for a user to be able to voice his requests in spoken natural language, the use of speech recognition modules or units is required. Such speech recognition modules are for example from the DE 197 19 381 C1 and the DE 199 56 747 C1 known.

Furthermore, a procedure for action determination ( DE 196 15 693 C1 ), in which a user gives specific commands or makes a specific request to a speech dialogue system, without him the choice alternatives available are tediously enumerated. As a result, a direct access to a variety of functions of a speech dialogue system is possible, which requires only a dialogue, provided that the initial utterances of the user are incomplete or ambiguous or in contradiction to individual possibilities of the background application. In the described cases, the described speech dialogue system with the name DIANE intervenes and carries out a clarification dialog with the user in order to determine the user's intention and to ask for missing information units. The described method is based on the following abstract task model:
A background application is considered to be a finite set of transactions (T1, T2, .... Tx), where each transaction is associated with a finite set (which may also be empty) of transaction parameters (P1, P2, ..... Px). The transaction parameters are known to the speech dialogue system. For each transaction parameter, a grammar is provided which serves to capture the transaction parameter in the dialog.

Of the User can choose the desired Transaction and the assigned transaction parameters in one Naming a sentence or not. In the first case, the transaction done immediately In the second case, the still unknown parameters in Dialog with the user recorded. If by the user statement no Transaction is uniquely determinable, the system performs automatically a clarification dialogue for Determination of the desired Transaction through. The same applies to unclear or incomplete transaction parameter entries.

From the DE 101 10 977 C1 For example, a method and an arrangement for providing help information for a user of a speech dialogue system for operating a background application are known in which predefined help information is determined and output to the user depending on the existing dialog status. The user is thus provided with context-sensitive help information by the speech dialogue system, depending on his respective dialog status.

Further are "unified Messaging Systems "known through which different messages on a communication platform stored, managed as well as access to these over different Allows access becomes. For example, you can Emails, SMS or even classic telephone, fax and answering machine functions here over connected so, for example, about such a "Unified Messaging System " Fax, send an email or email an SMS message to a mobile phone can be. Answering machine messages can listened to as usual or SMS and e-mails via "Text-To-Speech" (TTS) be read out. about a voice dialogue system can Thus, the individual messages heard while on the move, answered, forwarded or deleted become. Especially with such a variety of functions the background application "Unified Messaging System "is it for a user often very tedious, to move from one function to another move.

In order to Such speech dialogue system in the long run by a wide circle of users Accepted is besides a secure speech recognition as well the use of speech dialogue systems required the user not unnecessary Spend time on the right question of the speech dialogue system to wait or ask the right question by any manipulation force.

The The object of the present invention is thus a method for determining at least one transaction for operating a background application and an associated Specify a speech dialogue system in which a unique identification the one desired by the user Background application fast and user-friendly, in particular already possible on the basis of reduced action information. The object is achieved by a method according to claim 1 and a Speech dialogue system according to claim 16 solved.

Of the essential idea of the method according to the invention for the determination at least one transaction for operating a background application and evaluation of at least one contained in an expression of use Action information in a speech dialogue system, in each case a background application a finite set of selectable ones Transactions is assigned and with which over one in the speech dialogue system provided input unit detects at least one user utterance and stored in the speech dialogue system, is that individual correlated transactions in the speech dialogue system at least one matching each Context information is assigned and the context information at least the last executed Transaction is stored in the speech dialogue system. In the investigation In a transaction, those transactions are preferably preferred considered, whose associated context information matches the stored context information. Advantageous are determined by the assignment of at least one context information preferred to transactions in a meaningful context those transactions in the search for the current by the User addressed by the user utterance Transaction considered and thus both time-consuming and computationally intensive Investigation costs significantly reduced. Further advantageous can by way of of contextual Grammars already shortened key terms sufficient to determine the action desired by the user.

Further advantageous embodiments of the method according to the invention, in particular a voice dialog system for carrying out the method are given in the further claims.

The The present invention will become more apparent from an embodiment explained in more detail with reference to figures.

It demonstrate:

1 in a schematic block diagram by way of example a speech dialogue system,

2 in a further block diagram by way of example the individual components of a transaction and

3 in a flow chart by way of example the individual steps of the method according to the invention.

In 1 is a schematic block diagram exemplified a voice dialogue system 1 represented, via which a user B can serve at least one background application HA. The operation of the background application HA takes place here by way of example on the basis of individually selected and in the speech dialogue system 1 stored transactions T1 to Tx, which are assigned to one or more functions of the background application HA. This in 1 illustrated speech dialogue system 1 includes an input unit for this purpose 2 , a dialogue management unit 3 and an output unit 4 , where the dialogue management unit 3 each with the input unit 2 , the output unit 4 and the background application HA.

In the dialogue management unit 3 is a storage unit 5 for storing the transactions T1 to Tx for operating the background application HA and other parameters provided. The dialogue management unit 3 also has a speech recognition unit 6 , a speech synthesis unit 7 , a parser 8th and a control unit 9 on, which are interconnected for example via a data bus DBS. The background application HA is also connected to the data bus system DBS, which in an alternative embodiment is connected via an interface module (not in FIG 1 shown) with the dialogue management unit 3 can be connected. The input unit 2 of the speech dialogue system 1 is with the speech recognition unit 6 and the output unit 4 with the speech synthesis unit 7 connected.

By the user B, a user statement BE is generated, which has action information AI. In this case, a user statement BE can have one or more action information AI, which can be arranged within the user statement BE in different orders. About the input unit 2 is a user statement BE of the user B, preferably recorded as a voice signal and in the dialogue management unit 3 stored. In a preferred embodiment, the user utterance BE is digitized or in digital form of the memory unit 5 supplied and via the speech recognition unit 6 in a manner known per se, the action information (s) AI contained in the user statement BE is determined as well as in the memory unit 5 stored.

About the output unit 4 may also be provided to the user B output prompt or information prompt AP, for example in the form of a voice synthesis unit 7 generated synthesized speech signal via the output unit 4 be issued. As a result, the user B is informed, for example, about the current status of the background application HA to be operated or the actions carried out by the system on the basis of the last user statement BE, or the user B initiates the delivery of at least one further user statement BE. For this purpose, a clarification dialog which alternates between output prompt AP and user statements BE can be carried out.

In 2 are shown in a schematic representation by way of example the individual components of such a transaction T1 to Tx, which for the operation of the different functions of the back office HA HA in the dialogue management unit 3 are provided. A background application HA is assigned a finite set of such transactions T1-Tx whose selection is triggered via a respective grammatical routine GR assigned to a respective transaction T1-Tx. In this case, the transaction T1 is assigned one or more of such grammars G1-Gx, which are used to determine the transactions T1 to Tx associated with the stored action information AI in the parser unit 8th getting charged. In the parser 8th the individual keywords of the grammars G1-Gx characterizing the respective transaction T1 are compared with the action information AI. Here, the grammars G1 - Gx By way of example, contextual grammars G1 * - Gx * are replaced or extended, which have abbreviated key terms, taking into account the context of meaning between the individual transactions T1 to Tx.

Furthermore, the in 2 exemplified first transaction T1 a Vorbedingungsroutine VBR, which before the further execution of a transaction T1 - Tx in the control unit 9 is executed to verify the existence of preconditions such as the successful execution of another transaction T2. For this purpose, the preconditioning routine VBR has one or more precondition parameters VBP and associated parameter prompts PP. Through the in the control unit 9 executed preconditioning routine VBR, for example, the activation or deactivation of another transaction T2 checked and selected depending on the result of the associated Parameterprompt PP and the speech synthesis unit 7 as well as the output unit 4 issued to the user B. If the precondition defined by the precondition parameters VBP for the execution of the considered first transaction T1 is met, the further execution of the determined first transaction T1 is initiated without an output of a parameter prompts PP.

According to the invention, the transactions T1 - Tx which are connected in a meaningful context are in the speech dialogue system 1 each associated with at least one matching context information KI and thus summarizing the meaningful relationship having transactions T2, T3 to a respective transaction group TG1 - TGx with matching context information KI. In 1 By way of example, a second and a third transaction T2, T3 are combined to form a first transaction group TG1.

Especially a background application HA such as a "Unified Messaging System "with a Variety of similar Functions related transactions T1 - Tx, so many of the transactions T1 - Tx such matches exhibit. For example, the first transaction T1 = "Email read "and the second Transaction T2 = "Email delete "can by the assignment the context information KI = "Email Processing " a first transaction group TG1 = "Email processing" summarized become. Thus, the contextual information KI becomes the contextual shortcut different transactions T1 - Tx among each other for the current Determination of a transaction T1 to Tx evaluated and thereby the Time and computational effort for the implementation significantly reduced in the preliminary investigation.

In the storage unit 5 is the context information KI _last at least the last in the dialog management unit 3 executed transaction T _last and taken into account in the current determination of a transaction T1 - Tx, in such a way that preferably those transactions T2, T3 are taken into account in the determination whose associated context information KI with the stored context information KI _last match. For the determination of the current transaction T1 - Tx, first of all that transaction group TG1 - TGx is considered to which the last executed transaction T belongs _last . As a result, the transactions T1-Tx of the determined transaction group TG1-TGx are particularly advantageously prioritized in the selection process, and thus those transactions T1-Tx are executed which have the largest "hit probability".

For checking the context information KI, the transaction T1-Tx is respectively assigned a context information routine KIR that matches the context information KI associated with the transaction T1 with that in the memory unit 5 stored _last KI context information last in the dialogue management unit 3 completed transaction T _last checked. If the last stored context information KI _last and the context information KI assigned to the first transaction T1 agree, the considered first transaction T1 is preferably taken into account in the determination of the current transaction, ie the action information AI obtained from the user statement BE via that in the parser unit 8th executed grammar routine GR of the considered transaction T1 evaluated. The context information KI may, for example, have the parameter type "string", which represents a generic term representing the meaning relationship between the transactions T1 to Tx.

In addition, each transaction T1-Tx is assigned a transaction parameter routine TPR which, in addition to the determined transaction parameters P1-Px, has transaction parameter prompts TPP, a parameter grammar PGR and a value determination information WEI. The transaction T1 is specified in terms of content by its transaction parameters P1-Px whose values are determined via the transaction parameter routine TPR. This will be done via in parsers 8th executed parameter grammar PGR from the action information AI, the values of the transaction parameters P1 - P1x determined. The transaction parameters P1-Px, which are yet to be determined for the execution of the transaction T1-Tx, are sent to the user B by issuing the respectively assigned transaction parameter prompts TPP and his in a further case user information BE contained action information AI in the context of a clarification dialogue.

Furthermore, value determination information WEI assigned in each case to the individual transaction parameters P1-Px is determined in which manner the determination of the transaction parameters P1-Px should or can take place. For example, in order to determine the transaction parameters P1-Px, besides the execution of a clarification dialog with the user B, they can already be found in the speech dialogue system 1 present parameter values or status information are evaluated. For this purpose, the dependency of a transaction parameter P1-Px is evaluated by the presence or absence of a given dialog status. Thus, with the aid of value determination information WEI, different ways of determining the value of a transaction parameter P1-Px can be predetermined, whereby unnecessary user queries can be avoided in the context of a clarification dialog.

To check the validity of a changed transaction parameter, a constraint routine CR is provided per transaction T1-Tx, which includes trigger parameters TR, logical conditions LB and action instruction prompts AA. The modified parameter value of a transaction parameter P1-Px is thus ascertained for validity via the contraint routine CR with the aid of the predetermined trigger parameters TR and the logical condition LB and communicated to the user B in the event of a deviation via action instruction prompts AA. For example, it is specified as a logical condition LB to form the sum of two transaction parameters P1, P2 and to check whether this exceeds, for example, an upper limit P _max . When the upper limit value P _max is exceeded, the associated action instruction prompt AA of the contraint routine CR is output. Alternatively, depending on the evaluation of the logical condition LB, a predetermined system function can also be initiated.

Over a Nachbedingungsroutine NBR is checked if all for the execution the respective transaction T1 - Tx required post-conditions NB, for example, those for the execution at least required transaction parameters P1 - Px, now present, i. the gaps are modeled as a frame modeled transaction T1 and with the help of chosen Transaction T1 the assigned function of the background application HA can now be started. Thus, via the Nachbedingungsroutine NBR those post-conditions NB checks, which at the end of the dialogue must be present for a handover all for the Operation of the background application HA required information about the To ensure transaction T1. Can be advantageous by the Nachbedingungsroutine NBR different complex Nachbedingungen NB for different transactions T1 - Tx individually defined and checked become.

Finally, points a transaction T1 - Tx a system action routine SAR, with the help of the output of specific action prompt AP or system action SA implemented becomes. The system action routine SAR comprises, for example, this System trigger parameter TRP and precondition information VBI as well as predetermined system actions SA. In case of a change the system trigger parameter TRP is via the precondition information VBI verifies the initiation of the given system actions SA and, if appropriate, the assigned system actions SA started. With the aid of the system action routine SAR, the user B is thus informed about a change the dialog status or the status of the background application HA.

In 3 is the inventive method for determining at least one transaction T1 - Tx in the speech dialogue system 1 illustrated by way of example in a flow chart. By means of a in the control unit 9 executed selection routine AR, the in 3 shown five process steps S1 - S5 processed. First of all, in a first step S1, the correlated context information KI is assigned to the transactions T1-Tx, which are in a context of meaning.

Furthermore, in the context of the execution of a transaction T1-Tx, the context information KI assigned to it as the last-assigned context information KI is _last in the voice dialog system 1 or the storage unit 5 the dialogue management unit 3 stored (second step S2).

After the determination of the action information AI from the respectively detected user statement BE (third step S3), those transactions T1-Tx are determined whose assigned context information KI _last matches the last-assigned context information KI (fourth step S4).

In a fifth and last step S5 are the finite amount of transactions T1 - Tx the previously determined transactions T2, T3 for the determination of the current Transaction T1 - Tx preferably used.

• 1. Benutzeridentifizierung im Sprachdialogsystem (Login)
• 2. Abfrage neuer Nachrichten
• 3. Abfrage neuer Nachrichten einer bestimmten Person
• 4. Senden von Nachrichten.

In the following text, the structure and the parameters P1-Px of an exemplary embodiment will be described below Transactions T1 to T4 within the speech dialogue system 1 described in more detail. By way of example, a background application HA is considered to be a "unified messaging system", which as a rule provides a telephone connection between the user B and the voice dialogue system 1 exists, ie the input unit 2 is as a microphone unit of a telephone and the output unit 4 formed as a speaker unit of the same. The basic functions of such a Unified Messaging system include the following functions:

• 1. User identification in the speech dialogue system (Login)
• 2. Query new messages
• 3. Query new messages of a specific person
• 4. Sending messages.

To operate the said basic functions of the background application Unified Messaging System, four transactions T1-T4 are modeled at dialoglevel, one transaction T1-T4 being assigned one of the basic functions by way of example, ie the first transaction T1 is used to identify the user based on the second transaction T2 listening to new messages; via the third transaction T3, listening to messages of a particular person; and by means of the fourth transaction T4, sending messages via the speech dialogue system 1 controlled.

First transaction T1: "User identification"

About the dialogue management unit 3 Using the grammar routine GR, the action information AI obtained from the user statement BE is compared with the grammars G1-Gx or their key terms required for executing the first transaction T1, and the associated transaction T1 is started if they match.

For the further execution of the first transaction T1 "user identification", it is necessary that the user B is not yet in the speech dialogue system 1 is logged in. This also corresponds to the precondition parameter VBP for the execution of the first transaction T1, which is checked via the preconditioning routine VBR. If user B is already in the speech dialogue system 1 logged in, the execution of the first transaction T1 is interrupted and the user B issued a corresponding action Prompt AP. If the user is not yet in the speech dialogue system 1 logged on, the first transaction T1 is executed further and the transaction parameter routine P1R determines the transaction parameters P1 = "user name" and P2 = "password" assigned to the first transaction T1 "user identification." In the following, the telephone number of user B is assumed as the user name.

The user B gives as a user statement BE his phone number and the associated password in the speech dialogue system 1 one. The user utterance BE is via the input unit 2 detected and based on the speech recognition unit 6 the telephone number contained in the user statement BE and the password and determined as action information AI in the memory unit 5 saved. By means of the transaction parameter routine TPR, the first and second transaction parameters P1 = "user name" and P2 = "password" required for executing the first transaction T1 are determined from the stored action information AI.

The postcondition routine NBR checks the existence of the first and second transaction parameters P1, P2 required for executing the function of the background application HA "Unified Messaging System" assigned to the first transaction T1 "user identification", and then calls the associated function. Further, an action prompt AP indicating the status of the identification process to the user B via the output unit 4 output.

Finally, or at any arbitrary point in time during the transaction processing, a context information item KI associated with the first transaction T1 is optionally stored in the memory unit 5 saved.

The described dialogue sequence is briefly summarized by the following schematic table representation:

Second transaction T2: "Query new news "

For example, the user B gives the following user statement BE in the speech dialogue system 1 a: "I want to listen to my new messages." The user utterance BE is via the input unit 2 detected and based on the speech recognition unit 6 determines the action information AI contained in the user statement BE and in the storage unit 5 saved. Possible action information AI can here be the terms "new", "messages" and "listening".

On the context information routine KIR each of a transaction is initially T1 - Tx associated context information AI with the most recently saved context information KI compared _latter and that the observed transaction T2, T3 preferably used in a match for the determination of the current transaction T2. Alternatively, by means of the in the control unit 9 selection routine AR executed most recently saved context information KI to be loaded _last and after reference to the context information AI of the individual transactions T1 - Tx those transactions T2, are provided T3 with matching context information AI for the determination of the current transaction T2, T3.

Subsequently, the respective grammar routine GR of the related context, ie the same context information KI having transactions T2, T3 is started and the match of the obtained from the user statement BE action information AI with the required to execute the second or third transaction T2, T3 grammars G1 - Gx or their keywords are checked and started the identified transaction T2 if they match. In the present case, the second transaction T2 "query new messages", for example, the context information "Unified Messaging System: message query" assigned, which in turn when the second transaction T2 in the storage unit 5 the dialogue management unit 3 as last assigned context information KI is _last saved.

Before the determined second transaction T2 "query new messages" is executed further, is checked in analogy above on the basis of the assigned preconditioning routine VBR, whether the first transaction T1 "user identification" was successfully performed, i.e. the user B logged in or logged in the speech dialogue system is.

Via the post-condition routine NBR of the second transaction T2, the function ("polling new messages") of the background application HA "Unified Messaging System" assigned to the second transaction T2 is subsequently called, and the associated information prompt AP together with the messages to the user B via the output unit 4 output.

The dialog structure on which the second transaction T2 is based can thus be illustrated as an example as follows:

Third transaction T3: "Query new messages from a specific person "

The submitted user statement BE "I would like to listen to the new messages from Mr. Maier" will turn over the input unit 2 detected and based on the speech recognition unit 6 determines the action information AI contained in the user statement BE, which in the storage unit 5 is stored. In the dialogue management unit 3 is stored after the execution of the second transaction T2 as the last stored context information KI _last "Unified Messaging System: message _query ".

By means of the respective context information routine KIR of the transactions T1-Tx, the correspondence with the context information "Unified Messaging System: Message polling" is checked and, based on the coincidence of the context information KI of the third transaction T3 with the last stored context information KI, the _{latter is} preferably used.

Subsequently, will the respective grammar routine GR of the third transaction T3 is started and the match the user statement BE obtained action information AI with the execution of the third transaction T3 required grammars G1 - Gx or whose keywords are checked and if they match this continues the third transaction T3.

In an analogous manner, the preconditioning routine VBR also checks the successful execution of the first transaction T1 "user identification" in the context of the third transaction T3 and from the stored action information AI via the transaction parameter routine TPR as transaction parameter P1 = "name" the name of the person ("Mr. Maier ") whose new messages are to be output to user B. Similarly, the context information KI" Unified Messaging System: message query "in the memory unit 5 as last present context information KI _last saved.

Using the following table, the described steps of the third transaction T3 are illustrated by way of example using a table:

Fourth transaction T4: "Send a message "

Around a message about record the unified messaging system and then log on a selected one To send person is the fourth transaction T4 provided.

This is initiated by the user B by means of, for example, the user's statement BE = "I would like to send a message to Mr. Maier." The user utterance BE is again transmitted via the input unit 2 detected and based on the speech recognition unit 6 determines the action information AI contained in the user statement BE and in the storage unit 5 saved.

Taking into account in the dialogue management unit 3 after the execution of the third transaction T3 last stored context information KI _last "Unified Messaging System: Message query" are first the grammar routines GR of the second and third transactions T2, T3 checked for matches with the stored action information AI and then the other transactions T4 - Tx used.

Finally will the grammar routine GR of the fourth transaction T4 is started and agreement the user statement BE obtained action information AI with the execution of the fourth Transaction T4 required grammars G1 - Gx or their keywords are checked and based their agreement with the stored transaction information the fourth transaction T4 continued.

When Prerequisite is again the successful registration of the User required in the system, based on the precondition routine VBR is verified.

To send a message to the given recipient "Mr. Maier", it is first necessary that the user BE the message in the speech dialogue system 1 enters. The message is through the dialogue management unit 3 After the user has entered the message, the user has the option of being able to check the correct input by listening to the recorded message, for which user B can specify P2 = "lists" via a second transaction parameter. whether he wants to listen to the recorded message or not.

Also should be through the speech dialogue system 1 in case of an erroneous inclusion of the user B support he drive that a re-recording of the message is initiated. If the recorded message is correct, P3 = "Confirm Message" is to use the language based on a third transaction parameter dialog system 1 that the recorded message can be processed further.

alternative is also a by providing the value determination information WEI indirect confirmation of correct reception of the message, i. without user intervention possible. For this becomes the dependency between the second and third transaction parameters P2 = "lists" and P3 = "confirm Message "recycled, and Although such that over the validation routine WEI automatically returns the value of the third one Transaction parameter P3 = "Confirm Message "is set to" Yes "and thus further processing the recorded message without replay is implemented, provided the second transaction parameter P3 = "Lists" has the value "No".

To confirmation the error-free recording of the message we by means of the transaction parameter routine TPR is the value of the fourth transaction parameter P4 = "receiver", namely the Recipient of recorded message, determined.

Finally, by means of a fifth transaction parameter P5 = "Confirm Sending" by the user B, the confirmation for sending the recorded message to the in the dialogue management system 3 stored receiver ("Mr. Maier") queried.

In addition, still can the delivery parameters regarding the recorded message as further transaction parameters P6 - Px ( "Special Settings "), for example the confidentiality, urgency, encryption, etc., by means of a further clarification dialogue be queried by user B.

Further become dependent of system trigger parameter TRR regarding the present transaction parameters P1 - Px by means of the system action routine SAR for the different system trigger parameters Systemactions SA assigned to TRR. For example, at a second having the value "yes" Transaction parameter P2 = "Lists" the output of the started message and with a value of "no" third Transaction parameter P3 = "Confirm Message "or one the value "No" has the fifth transaction parameter P5 = "Confirm Sending" deleting the recorded message.

In the context of the fourth transaction T4, the context information routine KIR also contains the context information KI "Unified Messaging System: Messaging" in the memory unit 5 as last present context information KI _last saved.

The postprocessing routine NBR of the fourth transaction T4 invokes the assigned function ("send message") within the Unified Messaging system and generates the associated information prompt AP and via the output unit 4 output.

The modeling of the fourth transaction T4 is shown schematically below using a table:

Based on the explained specification of four transactions T1 to T4, a dialog between a user B and the speech dialogue system will be described below by way of example 1 described. System (AP1): Welcome to the Unified Messaging System. Please identify yourself in the system. User (BE1): 123456 System (AP2): Phone number 123456, please enter your password. User (BE2): 333666 System (AP3): You are logged in to the system. What can I do for you? User (BE3): I would like to hear all new messages from Mr Maier. System (AP4): Messages from Mr. Maier ... a moment please ... System (AP5): You have two new messages from Mr. Maier: First message ...; Second message ... User (BE4): I would like to send a message to Mr. Maier. System (AP6): Please enter your message and press the # key as you type. User (BE5): Dear Mr. Maier, I take ... (# pressed) System (AP7): Do you want to listen to your recorded message? User (BE6): No System (AP8): Do you want to send the message normally? User (BE7): Yes System (AP9): Your message has been sent.

Based the following tabular presentation becomes the change the respective transaction parameters P1 - Px in the context of the described Dialogues tabulated:

BE4: I would like to send a message to Mr Maier.

BE5: Dear Mr. Maier, I'll take ... (# pressed)

BE6: No.

BE7: Yes

By means of the precondition parameters VBP and the associated context information KI, a quick identification of the function of the background application or of the transaction T1-Tx assigned to it by the user B is thus realized. In particular, by the storage of the last available context information KI _last in the memory unit 5 the dialogue management unit 3 and the initially limited evaluation of the obtained with the received action information AI in terms of a matching context information KI having transactions T2, T3 a transaction group TG1 is the ease of use and ease of use of the speech dialogue system 1 clearly increased.

The The invention has been described above with reference to an embodiment. It goes without saying that numerous Modifications and modifications are possible without thereby leave the inventive concept underlying the invention becomes.

1

Voice control system

2

input unit

3

Dialog management unit

4

output unit

5

storage unit

6

Voice recognition unit

7

Speech synthesis unit

8th

parser

9

control unit

AA

action prompts

AI

action information

AP, AP1 - AP9

output prompt

AR

selection routine

B

user

BE, BE1 - BE7

user utterance

CR

Constraint routine

DBS

data bus system

G1 - Gx

grammars

G1 * - Gx *

contextual grammars

GR

Grammar routine

HA

Background application

KI

context information

AI _last

last saved context information

KIR

Context information Routine

LB

logical condition parameters

NBA

Nachbedingungsaktionen

NBP

Nachbedingungsparameter

NBR

postcondition

P1 - Px

transaction parameters

P _max

upper limit

PGR

parameter grammars

PP

parameter prompt

S1 - S5

first until fifth step

SA

system Action

SAR

System Action Routine

T1 - Tx

transactions

T _last

last executed transaction

TG

transaction group

TPP

trigger parameters

TPR

Transaction parameters routine

TRR

System parameters Trigger

VBI

precondition information

VBP

Vorbedingungsparameter

VBR

Vorbedingungsroutine

VBR

Vorbedingungsroutine

VB

precondition

WEI

Valuation Information

Claims (19)

Method for determining at least one transaction (T1-Tx) for operating a background application (HA) by evaluating at least one action information (AI) contained in a user statement (BE) in a speech dialogue system ( 1 ), wherein in each case a background application (HA) a finite set of selectable transactions (T1 -Tx) are assigned and in which a speech application in the system ( 1 ) provided input unit ( 2 ) at least one user utterance (BE) and in the speech dialogue system ( 1 ), characterized in that transactions (T2, T3) in the speech dialogue system ( 1 ) in each case at least one matching context information (KI) is assigned, - that the context information (KI _last ) at least the last executed transaction (T _last ) is stored in the speech dialogue system and - that for determining the current transaction (T1 -Tx) preferably those transactions ( T2, T3) whose associated context information (KI) matches the stored context information (KI _last ). Method according to claim 1, characterized in that that the transactions in a meaningful relationship (T2, T3) to a transaction group (TG1) with matching ones Context information (AI) are summarized. Method according to claim 1 or 2, characterized that the current transaction (T1 - T4) by means of one each Transaction (T1 - Tx) associated grammar routine (GR) and associated grammars (G1 - Gx) by evaluating the action information contained in the user statement (BE) (AI) is determined. Method according to one of claims 1 to 3, characterized in that in the speech dialogue system ( 1 ) a respective number of transaction parameters (P1 - Px) is assigned to a respective transaction (T1 - Tx). Method according to one of the preceding claims, characterized in that the at least one action information (AI) contained in the stored user statement (BE) is recorded by means of a speech recognition unit ( 6 ) is determined. Method according to one of the preceding claims, characterized characterized in that at least one of the transactions (T1 - Tx) at least a precondition parameter (VBP) is assigned. Method according to Claim 6, characterized in that the presence of the at least one assigned precondition parameter (VBP) by means of an associated Preconditioning routine (VBR) is verified. Method according to claim 7, characterized in that that by the Vorbedingungsroutine (VBR) the activation or deactivation another transaction (T1) is checked. Method according to claim 8, characterized in that, depending on the result of the check, a speech dialogue system ( 1 stored, assigned Parameterprompt (PP) and via the output unit ( 4 ) or the execution of the currently determined transaction (T2, T3, T4) is continued. Method according to one of the preceding claims, characterized characterized in that at least one of the transactions (T1 - Tx) at least a Nachbedingungsparameter (NBP) is assigned. Method according to claims 10, characterized in that that after the execution the currently determined transaction (T1, T2, T3, T4) an associated Nachbedingungsroutine (BNR) is started and by means of this the presence of the assigned Nachbedingungsparameter (NBP) checked as well associated post-conditional actions (NBA) are initiated. Method according to one of claims 4 to 11, characterized that the parameter value of a transaction parameter (P1 - Px) by means of a transaction parameter routine (TPR) is determined, wherein at least a transaction parameter (P5) a value determination information (WEI), which indicates the investigation path. Method according to one of Claims 4 to 12, characterized that at a change a transaction parameter (P1 - Px) of a transaction (T1 - Tx) the validity of the changed Transaction parameter (P1 - Px) is checked. Method according to claim 13, characterized in that that the review of validity of the changed Transaction parameters (P1 - Px) is performed by means of a constraint routine (CR), which trigger parameters (TR), at least one logical condition (LB) and action instruction prompts (AA). Method according to one of claims 1 to 14, characterized in that the user (B) of the speech dialogue system ( 1 ) during the execution of the currently determined transaction (T1 - Tx) of a background application (HA) given information or Aktionsprompts (AP) are displayed. Method according to claim 15, characterized in that that the given information or action prompts (AP) with Help of a system action routine (SAR) can be determined which Trigger parameter (TRR), precondition information (VBI) and default Systemactions (SA). Method according to Claims 1 to 16, characterized that the individual transactions (T1 - T4) each one of the associated with context information (KI) dependent grammar (G1 * - Gx *) is assigned. Speech dialogue system for carrying out the method according to one of claims 1 to 17, comprising at least one input unit ( 2 ), a dialogue management unit ( 3 ) and an output unit ( 4 ), characterized in that - the dialogue management unit ( 3 ) a control unit ( 9 ) to associate at least one match context information (KI) to transactions in a meaningful relationship (T2, T3), - that a memory unit ( 5 ) is provided for storing at least the context information (KI _last ) of the last executed transaction (T _last ) and - that in the control unit ( 9 ) is executed, which preferably selects those transactions (T2, T3) for the determination of the current transaction (T1 - Tx) whose associated context information (KI) matches the stored context information (KI _last ). Speech dialogue system according to claim 18, characterized in that that the background application to be used (HA) as "Unified Messaging systems "trained is.