WO2002093415A1 - Method and device for voice data input in an electronic form - Google Patents

Abstract

The invention concerns a voice-controlled method for controlling a parameterable system designed to execute different actions on the basis of values adopted by several voice-input parameters. The invention is characterised in that it comprises steps which consist in: prompting a user to state an initial request; receiving and processing said request so as to identify in said request the presence of one or several groups of words; for each group identified in said request and associated with an admissible value attributed to one of said voice-input parameters, assigning said admissible value to said parameter; verifying whether there remains at least one voice-input parameter without assigned value; if this is the case, prompting the user to state a complementary request; otherwise, controlling the execution of an action (41, 53) attributed to the particular combination obtained from the values assigned to each parameter.

Description

 METHOD AND SYSTEM FOR VOICE ENTRY OF DATA INTO A FORM

The present invention relates to a voice command method, a configurable system, that is to say a system capable of performing distinct actions according to the values taken by several parameters, with voice command, and a computer program for executing distinct actions depending on the values taken by several parameters. In particular, the invention relates to any device authorizing the implementation of a computer voice command in order to execute actions under the conditions set out below. Telephone servers are a frequent example of such devices, but the invention also relates to means which do not impose the use of a public or private telephone network to route a speech signal to a computer executing the actions. There are known configurable voice control systems such as telephone servers, for example in the transport sector. Such a system is capable of performing multiple actions, for example reserving tickets, communicating timetables or practical information, canceling a reservation, under the command of a telephone contact. A control process must be implemented in order to correctly assign values to the various parameters defining the action desired by the interlocutor. For example, when booking a train ticket, the machine must be able to assign values to the parameters consisting of the city of departure, the city of destination, the date of the journey, the time of departure, the class, etc. In known methods, questioning by the user's automaton takes place according to a predetermined tree structure. The user is forced to follow a path, defined during the design of the server. Schematically, the user must move on a tree, in the mathematical sense of the term. Each question of the server is a node on which are linked a certain number of branches corresponding to the admissible answers for the question asked. For example, the machine first requests the city of departure, it processes the user's response using a voice recognition method and, if it does not recognize a valid departure location in the response, it restores the same question, if not, it goes to the next question relating, for example, to the destination. The server thus imposes its own logical path on the user. The process continues until the arrival on a leaf of the tree, that is to say a node to which no branch is attached. Such a tree-like process has drawbacks: it has restrictive ergonomics since the user must comply with a data order imposed on him by the server, even if he reasons naturally in a different order. The user must provide the data one by one, even if he has all of them in mind, which is not optimal from the point of view of the total time of use to obtain a precise action. In addition, the user cannot change his idea or need during the process and it is impossible for him to totally or partially modify his orders, unless this possibility was fully provided for from the original design and is offered to the user choice.

The design, development and implementation of a tree-like process requires a generally meticulous and cumbersome design phase, the study of all cases being very long due to the combinatorial explosion of the number of cases depending on the number of parameters involved. This phase is followed, in the case of a rich application, by the writing of a voluminous software, in which all the questions are asked, in which the test loops on the answers to each successive question depend on the others and in which all the possible paths are written, that is to say all the combinations of possible admissible responses. The system is vulnerable to a programming error due to the length of the code and that a failure of an operation in the software at a node in the tree structure results in an overall failure of all the system functions associated with all the branches from this node.

The object of the present invention is to allow voice control of a configurable system while solving the aforementioned drawbacks.

For this, the invention proposes a voice command method for controlling a configurable system capable of performing distinct actions as a function of the values taken by parameters. comprising several voice input parameters, at least one and, preferably, several predetermined admissible value (s) being assigned to each of said voice input parameters, characterized in that it comprises the steps consisting in: - inviting a user to state an initial request,

receive and process said request by an automatic speech recognition method so as to recognize the presence in said request of one or more group (s) of words belonging to a predetermined set of distinct groupings of words, ie that is to say to a dictionary, said groupings forming recognizable expressions, at least one of said groupings being associated with each of said admissible values,

- for each group recognized in said request and associated with an admissible value assigned to one of said parameters with voice input, assign said admissible value to said parameter,

- check if there is at least one voice input parameter with no assigned value, i.e. if the request is incomplete,

- if so, invite the user to state an additional request and resume the process from the aforementioned reception and processing step,

- if not, when the request is complete, order the execution by said configurable system of an action attributed to the particular combination obtained from the values assigned to each parameter. Within the meaning of the invention, a grouping of words consists of one or more words recognized as forming a whole by the speech recognition engine.

Such a method allows the control of the configurable system in natural language. In fact, the order of the parameter assignments is not imposed on the user. The latter can state a complete request at once, if he is experienced with the process or complete it several times. For each additional request, it can provide one or more data which are all taken into account. The processing time to obtain an action is therefore reduced. The user can choose the order of information that is natural to him, which improves ergonomics. The journey is no longer tree-like but reticular, a transverse path or a jump to any step of parameter assignment being possible at any time. The total number of parameters to be assigned to obtain a complete request can be predetermined initially, for example when only one type of action is planned, or can be determined during the process, depending on the values of a part of the parameters.

Preferably, said parameters consist of one or more selection parameter (s), making it possible to select a type of action to be executed from among several, and one or more information parameter (s), making it possible to specify the action to be executed at Within a given type of action, the aforementioned verification step consisting in:

- first, check whether there is at least one selection parameter with no assigned value and, if not, designate a batch of mandatory information parameters according to the value assigned to said audit or to each selection parameter, said batch comprising one, several or no information parameter (s), and check whether there remains, in said batch, at least one information parameter with voice input without assigned value. In this case, the value of the selection parameter (s) determines the type of action chosen and conditions the number and type of information parameters that must be specified for the query to be complete, that is to say that the action can be performed. These mandatory information parameters are automatically designated according to the type of action chosen. The action attributed to the particular combination obtained from the values assigned to each selection parameter and to each mandatory information parameter is finally executed by the configurable system. The selection parameter (s) can be entered by voice or other, the information parameter (s) also.

Advantageously, the aforementioned verification step consists, when there remains at least one selection parameter with voice input without assigned value, that is to say if there remains a selection of type of action to be performed, to be invited the user to state an additional request and resume the process from the reception and processing step mentioned above. According to a particular embodiment of the invention, the processing of an additional request can be followed by a step consisting, for any recognized group associated with an admissible value assigned to a parameter with voice input to which a value is already assigned, to check with the user if he wishes to assign a new value to said parameter and, if so, to assign said parameter to said admissible value, if not, to leave said parameter unchanged. Thus, if the user goes back on his choice, he can correct the parameter values already assigned, but he is not likely to do so by mistake. Such a verification step is lighter to implement in the context of the method of the invention than in the context of a tree-like method.

Some parameters which are not with voice input can be parameter with alphanumeric input. For example, in the case of a telephone server, one may also want to allow the transmission of certain parameters, not using the speech signal, but by exploiting the frequencies of the signal emitted using the keys of a telephone handset when this authorizes use in voice frequency mode (in English: Dual Tone Multiple Frequency). According to a particular characteristic of the invention, the parameters can also include at least one alphanumeric input parameter or one input parameter in voice frequency mode, the method comprising at least one step of receiving alphanumeric data, respectively one step of receiving signal in voice frequency mode, and at least one step of assigning a value to said or one of said alphanumeric parameter (s) as a function of said received alphanumeric data, respectively a step of assigning a value to said parameter input in voice frequency mode according to the alphanumeric data received in this mode. Such alphanumeric or voice frequency data can be entered via a touch-tone telephone or a system keyboard, for example in the case of a train ticket vending machine located in a station. A parameter can also be transmitted, either by voice input, or by input in voice frequency mode. According to another particular characteristic of the invention, said requests can be received via a data transmission network, for example a public or private telecommunications network or a data bus internal to a system. Advantageously in this case, the user is invited to state said requests by sending a spoken message to said user via said data transmission network. Thus, the system controlled by the method, for example, a telephone server, detects if a request expressed by a user is incomplete and automatically decides to organize a question / answer exchange to supplement the information.

Preferably, each invitation to state a complementary request is formulated as a function of the parameter (s) for voice input or for input in voice frequency mode with no value assigned. Thus the use of the method by a novice user, that is to say who initially had no idea of the data to be provided, is without difficulty, the method making it possible to guide the user parameter by parameter according to the data that 'he did not provide, so complete the request. A real dialogue is established between the user and the system. The latter is able to check whether the information given by the user is incomplete or inaccurate. Thus, the method according to the invention makes it possible to propose, with a single system and without involving several distinct interfaces, an experienced mode of use, employed by a user knowing in advance all the data expected by the system and providing all these data from the initial request, and a novice mode of use, employed by a user supplying the data one by one as they are requested, as well as all the possible intermediaries between these two modes. The transition between these different modes of use is automatic, transparent and continuous.

Advantageously, the method according to the invention can, if necessary, also include the steps consisting in inviting the user to provide an identifier, in verifying the authenticity of said identifier, and in inviting the user to state said initial request only if the identifier is recognized as authentic. Thus, the security of the system is improved, for example when it allows the consultation of confidential data. Within the meaning of the invention, the actions executed can be of any type, for example, the transmission of a message to the user, for example an information message, the sending of an order purchase, sale, reservation, appropriate service, etc. The configurable system controlled by the method can itself be a control system for a second system, for example a machine, an automaton or a vehicle, and the actions executed are the sending of corresponding commands.

The subject of the invention is also a configurable voice-controlled system, characterized in that it comprises:

an execution means able to execute distinct actions as a function of the values taken by parameters comprising several parameters with voice input,

a signal production means capable of producing, for a user, at least one invitation signal to say an initial request and at least one invitation signal to say an additional request,

- a voice signal reception means able to receive at least one voice signal from said user, - a storage means in which are stored a set of groupings of words forming the expressions recognizable by said system, that is to say a dictionary of the system, and predetermined admissible values assigned to said voice input parameters, at least one admissible value being assigned to each of said voice input parameters, at least one of said groupings of words being associated with each of said admissible values,

- an automatic speech recognition engine capable of processing said voice signal,

a control unit capable of carrying out the operations consisting in: commanding said signal production means to produce, for a user, a signal for invitation to say an initial request,

- controlling said automatic speech recognition engine to process a spoken request received by the voice signal reception means, so as to recognize the presence in said request of one or more of said groupings of words, - for each group of words recognized and associated with an admissible value assigned to one of said parameters with voice input, assign said admissible value to said parameter,

- check if there is at least one voice input parameter with no assigned value, i.e. if the request is incomplete,

- in the affirmative, order said signal production means to produce, for said user, an invitation signal to utter a supplementary request, and resume the abovementioned operations from the automatic speech recognition operation,

- if not, if the request is complete, order said means of execution to execute an action attributed to the particular combination obtained from the values assigned to each parameter.

Advantageously, said parameters consist of one or more selection parameter (s) and one or more information parameter (s), the above-mentioned verification operation consisting in:

- first, check whether there is at least one selection parameter with no assigned value and, if not, designate a batch of mandatory information parameters according to the value assigned to said audit or to each selection parameter, said batch comprising one, several or no information parameter (s), and check whether there remains, in said batch, at least one intelligence parameter with voice input without assigned value, said action executed by the configurable system being attributed to the particular combination obtained values assigned to each selection parameter and to each mandatory information parameter.

Preferably, the aforementioned verification operation consists, when there remains at least one selection parameter for voice input, or for input in voice frequency mode, with no assigned value, to control said signal production means to produce, at destination said user, an invitation signal to say a complementary request, and resume the above operations from the automatic speech recognition operation. Advantageously, it can be provided that, when said automatic speech recognition engine recognizes, in a additional request, a group of words associated with an admissible value assigned to a voice input parameter to which a value is already assigned, said control unit controls said signal production means to produce, for said user, an invitation signal confirming that a new admissible value must be assigned to said parameter, said control unit being capable of receiving, via said voice signal reception means, a positive or negative response signal and, if so, assign said admissible value to said parameter, if not, to leave said parameter unchanged.

According to a particular characteristic of the invention, said signal production means can comprise a voice signal production means connected to a telecommunications network.

Preferably, said set of groupings of words recognizable by said system and said admissible values assigned to said voice input parameters can be organized in a dictionary comprising means for associating at least one of said groupings of words with each of said admissible values.

Advantageously, a data structure may also be provided in the system, for example software belonging to the control unit, comprising a list of said parameters with voice input and a list of said admissible values, said data structure comprising means to assign each of said admissible values to one of said parameters. The invention also provides a computer program containing instruction codes suitable for being read or stored on a medium, said codes being executable by a computer provided with signal reception means capable of receiving at least one voice signal. to execute distinct actions as a function of the values taken by several parameters comprising several parameters with voice input, at least one predetermined admissible value being assigned to each of said parameters with voice input, characterized in that said codes are executable to execute the operations consisting at :

- produce a signal inviting an user to state an initial request,

- receive said request by said signal reception means and process said request by an automatic speech recognition method so as to recognize the presence in said request of one or more group (s) of words belonging to a predetermined set of distinct groups of words, said groupings forming the expressions recognizable by the system, at least one of said groupings being associated with each of said admissible values,

- for each group recognized in said request and associated with an admissible value assigned to one of said voice input parameters, assign said admissible value to said parameter, - check if at least one voice input parameter remains without assigned value, this is ie if the request is incomplete,

- in the affirmative, invite the user to state an additional request and resume the above operations from the reception and processing operation, - if not, when the request is complete, execute an action attributed to the combination particular value obtained from the values assigned to each parameter.

The use time of such a computer program is generally shorter than with a tree-like process, especially for an experienced user, since multiple assignments of parameters are possible for each request. This is advantageous for the reduction in communication costs which results for the user in the case of a telephone server.

In the case of a program providing an equivalent service by a tree-like process, the number of tests to be carried out to reach a node is generally proportional to the sum of the numbers of possible values for each parameter upstream of this node, while this number of tests is generally proportional to the number of parameters according to the invention. Reducing the number of tests to perform compared to reduces the overhead required for program execution. The design of the program is modular, which increases its robustness. Thus, a test step on a parameter that would be deficient can be short-circuited by providing a corresponding response from the initial request, so that the system remains functional despite possible errors affecting certain modules. The reliability and robustness of the system vis-à-vis programming errors are therefore improved. In addition, the length of the program according to the invention is overall reduced compared to that of a program providing an equivalent service by a tree-like process. The risks of programming errors are reduced.

Preferably, said parameters consist of one or more selection parameter (s) and one or more information parameter (s), the above-mentioned verification operation consisting of: - first, checking whether at least one parameter remains selection without assigned value and, if not, designate a batch of mandatory information parameters according to the value assigned to said or each selection parameter, said batch comprising one, several or no information parameter (s), and verify if there remains, in said batch, at least one intelligence input parameter with voice input without assigned value, said action executed being attributed to the particular combination obtained from the values assigned to each selection parameter and to each mandatory information parameter.

The invention will be better understood, and other objects, details, characteristics and advantages thereof will appear more clearly during the following description of a particular embodiment of the invention, given solely by way of illustration and not limiting, with reference to the attached drawing. On this drawing :

FIG. 1 is a diagram of a configurable system according to the invention,

FIG. 2 represents a part of a database of the system of FIG. 1,

FIG. 3 symbolically represents part of a grammar of the system of FIG. 1, FIG. 4 represents a part of a dictionary of the system of FIG. 1,

FIG. 5 is a general diagram of the operation of the system of FIG. 1,

FIG. 6 is a detailed diagram of step VI of FIG. 5. FIG. 1 represents a stock market information system by telephone which comprises a computer voice server 1 comprising a central control unit 2, an interface module 3 connected to a telecommunications network 4 and storage means 5 comprising a central memory and a hard drive. The voice server 1 is connected by the telecommunications network 4 to the voice frequency telephone handset 6 of at least one user, preferably of course to a plurality of users' telephone handsets. The voice server 1 is also connected to a stock market information database 7, for example by the telecommunications network 4 or by other means not shown.

The storage means 5 contain control software 8, an automatic speech recognition engine 9, a lexical database or dictionary 10, a database of parameters 11 and a voice simulation software 13. All of these elements are presented for example in the form of a plurality of data files, form part of the stock market information system and are configured to be used jointly by the server 1 to provide a stock market information service with voice control. The system allows a user equipped with the telephone 6 to orally formulate specific requests for information and to obtain, in response, the requested information. For this, the server 1 is able to receive oral requests from the user and to send information messages to the user via the interface module 3, the network 4 and the telephone 6. The content information messages is automatically adapted according to a plurality of system parameters. These parameters include a plurality of parameters with voice input that the system is able to position according to the oral requests received from the user and, optionally, one or more parameters with digital input in voice frequency mode, that the system is able to position. according to digital data sent by the user via the keyboard 12 of his telephone 6 to voice frequency. The system parameters are in predetermined number and include at least one selection parameter and at least one information parameter, both of which can be entered by voice or numerically. In the mode of very simple embodiment considered below, the system parameters include a PSI selection parameter and two information parameters PR1 and PR2, all with voice input.

With reference to FIG. 2, the parameter base 11 has two columns: column 14 contains a list of all the predetermined admissible values assigned to each of the parameters with voice input; column 15 contains, opposite each admissible value in column 14, the name of the parameter to which this value is assigned. The parameter base 11 therefore makes it possible to assign each admissible value to a predetermined parameter. Thus, the PSI parameter has two predetermined admissible values: PORTFOLIO and INFORMATION. Parameter PR1 has two predetermined admissible values: TITLE1 and TITLE2. Parameter PR2 has two predetermined admissible values: COURSE and VARIATION. The system is capable of providing two types of service to the user: a consultation service for his personal portfolio for which no option is available and an information service for two distinct stock titles TITLE 1 and TITLE2. For each stock security, the system can provide the user with data on the current price or data on its variation.

With reference to FIGS. 3 and 4, the dictionary 10 includes means for defining all of the statements that the voice recognition engine 9 of the system is able to interpret. For this, the dictionary 10 contains a list of lexical entries 16 which contains all the keywords that the system recognizes. Figure 4 contains part of this list, with the lexical entries INFORMATION, PORTFOLIO and ACCOUNT. The dictionary 10 also contains a list 17 of phonetizations, that is to say of predetermined groupings of one or more phonemes, forming all the expressions recognizable by the system. Each phonetization is a sequence of elementary sounds, comprising one or more syllables, associated with one of the keywords from list 16. There is at least one such phonetization for each keyword from list 16, but it can there are several when a keyword presents several possible variants of speech. In the example shown in Figure 4, the two phonetic groupings "in-for-ma-tion" and "in-fo" are associated with INFORMATION keyword, i.e. they are interpreted by the system as designating the same keyword. The phonetization “por-te- feuille” is associated with the keyword PORTFOLIO and the phonetization “comp-te” is associated with the keyword ACCOUNT. The dictionary 10 also includes a list of return values 18 comprising all the predetermined admissible values of each of the parameters with voice input, as well as empty values. A single return value from list 17 is associated with each keyword from list 16. In the part of the dictionary 10 represented in FIG. 4, the keyword INFORMATION is thus associated with the return value INFORMATION, the words -KEYS PORTFOLIO and ACCOUNT are both associated with the PORTFOLIO return value. We see that the dictionary 10 thus makes it possible to associate at least one phonetization with each admissible value of each parameter with voice input. With reference to FIG. 3, the dictionary 10 also contains blocks of keywords. A block of keywords is a set of one or more keyword (s) associated respectively with one or more admissible value (s) of the same parameter. There is at least one such block per parameter with voice input. In the example shown, block 19 is associated with the PSI parameter since it contains the key words PORTFOLIO and ACCOUNT; block 20 is also associated with the PSI parameter. Block 21 is associated with parameter PR2 since it contains the keywords COURSE and VARIATION. Block 22 is associated with parameter PR1 since it contains the keywords TITRE1 and TITRE2. Certain blocks, for example block 23, contain keywords which are not significant with respect to the parameters with voice input and are not associated with any parameter. Thus, the return values of the keywords HELLO and PLEASE are empty.

With reference to FIG. 3, the dictionary 10 also contains means defining the grammar of the voice command system. The grammar is a list of all the block sequences forming statements admissible by the system. It is shown diagrammatically by the arrows connecting the blocks to each other in FIG. 3, for example the arrow 24. The dictionary 10 comprising the list 17 of the recognizable phonetizations, the list 16 of keywords, the list 18 of the return values, the definition of blocks and definition of grammar of association of the blocks finally makes it possible to define all the vocal utterances that the system is capable of interpreting.

The dictionary 10 is used by the automatic speech recognition engine 9 in the following manner: when the server 1 receives a voice signal by the interface module 3, it compares this voice signal with the predetermined groupings of the list 17. S 'it does not recognize anything at the start of the signal, it does not take account of the start of the signal until it recognizes, for example by a statistical analysis, a first phonetization associated with a first keyword of the list 16. The procedure is then repeated with the continuation of the voice signal, phonetization by phonetization, until all the predetermined phonetizations present in the voice signal and all the corresponding keywords are identified. For example, the voice signal corresponds to the statement: “Hello, I want to consult my portfolio. »Engine 9 identifies the keywords HELLO, I WANT, CONSULT and PORTFOLIO. Note that the statement: "Hello, I want to consult my account" produces the same result.

Starting from the initial block 25 of the grammar of FIG. 3, the engine 9 then searches for the first keyword recognized HELLO among the blocks linked to the initial block 25. If it does not find it, it searches for this first word- key in the blocks connected to the blocks where the search has already taken place, and so on until it finds the first keyword, that is to say in block 23 in the example shown. Then it searches for the second recognized keyword I WANT in the blocks linked to block 23, then, if it does not find it, in the blocks linked to them, and so on until reaching the block output 28. In the example taken, the sequence of recognized keywords corresponds to the sequence of blocks 23-26-27-19. Then, the block sequence having been recognized, the search engine 9 returns the return value associated with each of the recognized keywords, that is to say in the present case, the return values: empty, empty, empty, PORTFOLIO.

When a sequence of recognized keywords does not correspond to any sequence of blocks provided for in the grammar, the statement cannot be interpreted by the system and an error message is generated. The operation of the stock market information system will now be described with reference to Figures 1, 5 and 6.

The user establishes a telephone contact with the server 1, as represented by the arrow 29. In step 30, the system sends the user a self-identification signal 32, for example a spoken message produced by the voice simulator 13 or a bell. In step 31, the user enters on the keyboard 12 a personal identification code 34. In step 33, the server compares the personal code 34 with a predetermined list of codes of the users subscribed to the service, so as to check the user's right of access to the service. If the personal code 34 is not recognized, the invitation to identify signal 32 is sent again. If the personal code 34 is recognized, the server 1 sends the user a signal 36 of invitation to say an initial request in step 35, for example a spoken message.

The user then utters an initial request 39 in the microphone 38 of the telephone 6. In step 37, the server 1 receives via the network 4 and processes the request 39 by the automatic speech recognition engine 9, by the method of the aforementioned automatic speech recognition. The motor 9 returns to the control program 8 all the return values of the keywords of the recognized statement. Using the parameter base 11, the central processing unit 2 identifies each voice input parameter to which a return value is assigned and assigns each return value to the parameter to which it is assigned. Then the control unit checks whether a value is assigned to the PSI selection parameter. Otherwise, as indicated by arrow 55, the system returns to step 35.

Using the example of the aforementioned statement, step 37 leads to the assignment of the value PORTFOLIO to the selection variable PSI. This means that the user has chosen, as a type of service, the consultation of his portfolio, and not the consultation of information on securities. Each admissible value of the selection variable PSI is associated with a predefined batch of mandatory information parameters. For the PORTFOLIO value, this batch is empty. For the INFORMATION value, this batch contains the parameters PR1 and PR2. In step 40, the control unit checks whether there remains a mandatory information parameter with no assigned value. In the present case, there is none and the request is therefore recognized as complete. Consequently, in step 41, the system performs the action assigned to the PORTFOLIO value of the PSI parameter, that is to say that it sends the user a message 56 containing data on his portfolio. action. Then the server returns to step 35 to allow the user to make another service request.

Taking another example of a statement for the initial request 39, for example the statement: "I want information on the Titrël action", step 37 results in the assignment of the value INFORMATION to the selection parameter PSI and assigning the value TITLE 1 to the information parameter PR1. With reference to FIG. 6, in step 42, the control unit checks whether there remains a mandatory information parameter with no assigned value. If so, as in this example, we check if it is the parameter PR1 in step 43. If not, as in this example, we check if it is the parameter PR2 in step 44 If so, as in this example, in step 45, the server 1 sends the user an invitation signal 46 to formulate an additional request to specify whether he wishes to know the course of the action TITLE 1 or its variation.

The user responds with the statement of an additional request 48 which is processed by the voice recognition engine 9 in the same way as the initial request 39, in step 47. If the engine 9 does not return any return value , the additional request 48 has not been understood and the invitation signal 46 is returned, as indicated by the arrow 49. If the return values include the admissible value VARIATION or COURSE, this value is assigned to the parameter PR2. If the return values produced by the additional request 48 are assigned to parameters which already have an assigned value, the values of these parameters are modified accordingly.

Thus, if the user has changed his mind and asked in an additional request to consult his portfolio, the value PORTFOLIO is assigned to the PSI parameter and the system goes to the aforementioned step 40, as indicated by the arrow 51.

Otherwise, assuming that the value VARIATION has been assigned to the parameter PR2 in step 47, the system returns to step 42 in which it is verified that the mandatory information parameters are all assigned. The request is then recognized as complete. As indicated by arrow 52, the system then goes to step 53 for executing the action corresponding to the combination obtained from the parameters PSI, PR1 and PR2, that is to say that it sends to the user an information message concerning the variation of the Titrel share price. For this, the server 1 retrieves the corresponding information from the stock market information database 7, as indicated by the arrows 54.

If in step 43, it is determined that it is the mandatory information parameter PR1 which remains without assigned value, in step 145, the server 1 sends the user an invitation signal to formulate an additional request for specify what is the name of the action on which he wants information. The user responds with the statement of an additional request which is processed by the voice recognition engine 9 in step 147. If the engine 9 does not return any return value, the additional request has not been understood and the invitation signal is returned, as indicated by the arrow 149. Steps 47 and 147 can be confused with step 37.

In an alternative embodiment, there is provided a step 50, visible in FIG. 6, in which it is checked whether the return values produced by the complementary request 48 are assigned to parameters already having an assigned value, for example the parameter PSI . In this variant, if the user has changed his mind and asked in additional request, in step 147, to consult his portfolio, the server 1 sends an invitation message to confirm that the service selection must be modified. The user sends a positive or negative confirmation signal, for example by the keyboard 12 or by answering "yes" or "no". If so, as indicated by arrow 51, the value PORTFOLIO is assigned to the parameter PSI and the system goes to step 40 above. If not, the system returns to step 42. Although the invention has been described in connection with a particular embodiment, it is obvious that it is in no way limited thereto and that it includes all the technical equivalents of the means described as well as their combinations if these are within the scope of the invention. In particular, a voice recognition engine of another type than that described above can be used. The dictionary 10 and the data structure 11 are only one example of the organization of the words recognizable by the system, the admissible values of the parameters, and the parameters with voice input. These two structures can be merged or arranged differently without departing from the scope of the invention.

In addition, in the context of the present invention, the steps and means for automatic speech recognition can be generalized to the recognition of shapes, characters or character strings. The parameters with voice input can be generalized corresponding to parameters with graphic, text or alphanumeric input. Correspondingly, a request can be generalized to a set of graphic, textual or alphanumeric data sent by the user and received by the system according to the invention by any suitable means, such as a textual or graphic interface. Correspondingly, the dictionary can be generalized, within the meaning of the invention, to a set of predetermined groupings of graphic, textual or alphanumeric data recognizable by the system according to the invention.

Claims

1. Voice control method for controlling a configurable system (1) capable of performing distinct actions as a function of the values taken by parameters comprising several parameters with voice input (PR1, PR2, PSI), at least one predetermined admissible value being assigned to each of said voice input parameters, characterized in that it comprises the steps consisting in: - inviting a user to state an initial request, - receiving and processing said request (39) by an automatic speech recognition method so as to recognize the presence in said request of one or more group (s) of words belonging to a predetermined set (17) of distinct word groupings, said groupings forming recognizable expressions, at least one of said groupings being associated with each of said admissible values, - for each group recognized in said request and associated with an admissible value assigned to one of said voice input parameters, assign said admissible value to said parameter, - check if at least one voice input parameter remains without assigned value, this is ie if the request is incomplete, - if so, invite the user to state an additional request (48) and resume the process from the aforementioned reception and processing step, - if not, when the request is complete, order execution by said configurable system of an action (41, 53) attributed to the particular combination obtained from the values assigned to each parameter. 2. Method according to claim 1, in which said parameters consist of one or more selection parameter (s) (PSI) and one or more information parameter (s) (PR1, PR2), the aforementioned verification step consisting in: - first, check (37) if there is at least one selection parameter with no assigned value and, if not, designate a batch of mandatory information parameters according to the value assigned to the audit or to each selection parameter, said lot comprising one, more or no parameter (s) of information, and check (40, 42) if there remains, in said batch, at least one parameter of intelligence with voice input without assigned value, said action executed by the configurable system being attributed to the particular combination obtained from the values assigned to each selection parameter and to each mandatory information parameter. 3. The method of claim 2, wherein the aforementioned verification step consists, when there is at least one selection parameter with voice input without assigned value, to invite (45, 145) the user to say a complementary request. and resume the process from the aforementioned reception and processing step. 4. Method according to one of claims 1 to 3, wherein the processing of an additional request is followed by a step (50) consisting, for any recognized group associated with an admissible value assigned to a voice input parameter to which a value is already assigned, to check with the user if he wishes to assign a new value to said parameter and, in the affirmative (51), to assign to said parameter said admissible value, if not, to leave said parameter unchanged. 5. Method according to one of claims 1 to 4, wherein said parameters also include at least one alphanumeric input parameter, said method comprising at least one step of receiving alphanumeric data and at least one step of assigning a audit value or to one of said alphanumeric parameter (s) as a function of said alphanumeric data received. 6. Method according to one of claims 1 to 5, wherein said requests are received via a data transmission network (4). 7. The method of claim 6, wherein the user is invited to state said requests by transmitting a spoken message (36) to said user via said data transmission network. 8. Method according to one of claims 1 to 7, in which each invitation (46) to state a complementary request is formulated as a function of the parameter or parameters with voice input without assigned value. 9. Configurable voice-activated system characterized by the fact that it includes: - an execution means (3) able to execute distinct actions as a function of the values taken by parameters comprising several parameters with voice input (PR1, PR2, PSI), - a signal production means (3, 13) suitable producing, for a user, at least one invitation signal (36) to utter an initial request and at least one invitation signal (46) to utter an additional request, a voice signal reception means (3) capable of receiving at least one voice signal from said user, - a storage means (5) in which are stored a set of word groupings (17) forming the expressions recognizable by said system and predetermined admissible values (18) assigned to said voice input parameters, at least one admissible value being assigned each of said voice input parameters, at least one of said groupings of words being associated with each of said admissible values, - an automatic speech recognition engine (9) capable of processing said voice signal, - a control unit (2) capable of performing the operations consisting in: - controlling said signal production means to produce (35), intended for a user, an invitation signal to utter an initial request, - controlling said automatic speech recognition engine to process (37) a spoken request (39) received by the voice signal reception means, so as to recognize the presence in said request of one or more of said groupings of words, - for each group of words recognized and associated with an admissible value assigned to one of said parameters with voice input, assign said admissible value to said parameter, - check if there is at least one voice input parameter with no assigned value, i.e. if the request is incomplete, - in the affirmative, order said signal production means to produce, for said user, an invitation signal to utter a supplementary request, and resume the abovementioned operations from the automatic speech recognition operation, - if not, if the request is complete, order said execution means (3) to execute an action (41, 53) assigned to the particular combination obtained from the values assigned to each parameter. 10. The system as claimed in claim 9, in which said parameters consist of one or more selection parameters (PSI) and one or more information parameters (PR1, PR2), the above-mentioned verification operation consisting in : - first, check (37) if there is at least one selection parameter with no assigned value and, if not, designate a batch of mandatory information parameters according to the value assigned to the audit or to each selection parameter, said batch comprising one, more or no information parameter (s), and checking (40, 42) if there remains, in said batch, at least one intelligence parameter with voice input with no value assigned, said action executed by the system configurable being assigned to the particular combination obtained from the values assigned to each selection parameter and to each mandatory information parameter. 11. The system as claimed in claim 10, in which the abovementioned verification operation consists, when at least one selection parameter with voice input has no assigned value, in controlling said signal production means to produce, intended for said user. , an invitation signal to utter an additional request, and resume the above operations from the automatic speech recognition operation. 12. System according to one of claims 9 to 11, in which, when said automatic speech recognition engine recognizes, in a supplementary request, a grouping of words associated with an admissible value assigned to a voice input parameter to which a value is already assigned, said control unit controls said signal production means to produce, for said user, a signal for invitation to confirm that a new value admissible must be assigned to said parameter, said control unit being capable of receiving, via said voice signal receiving means, a positive or negative response signal and, if so, assigning said admissible value to said parameter, if not, leave said parameter unchanged. 13. System according to one of claims 9 to 12, wherein said signal production means comprises a voice signal production means (3) connected to a telecommunications network (4). 14. System according to one of claims 9 to 13, in which said set of groupings of words (17) recognizable by said system and said admissible values (18) assigned to said parameters with voice input are organized in a dictionary (10) comprising means for associating at least one of said groupings of words with each of said admissible values. 15. The system as claimed in claim 14, comprising a data structure (11), whether or not distinct from said dictionary, comprising a list (15) of said voice input parameters and a list (14) of said admissible values, said data structure comprising means for assigning each of said admissible values to one of said parameters. 16. Computer program (8, 9, 10, 11, 13) containing instruction codes suitable for being read or stored on a medium, said codes being executable by a computer (1) provided with reception means signal (3) capable of receiving at least one voice signal to perform distinct actions according to the values taken by several parameters comprising several parameters with voice input (PR1, PR2, PSI), at least one predetermined admissible value being assigned to each of said voice input parameters, characterized in that said codes are executable for carrying out the operations consisting in: - Producing an invitation signal (36) for a user state an initial request, - receiving said request (39) by said signal reception means and processing said request by an automatic speech recognition method so as to recognize the presence in said request of one or more group (s) of words belonging to a predetermined set (17) of distinct groupings of words, said groupings forming expressions recognizable by the system, at least one of said groupings being associated with each of said admissible values, - for each group recognized in said request and associated with an admissible value assigned to one of said parameters with voice input, assign said admissible value to said parameter, - check if there is at least one voice input parameter with no assigned value, i.e. if the request is incomplete, - if so, invite the user to state an additional request and resume the above operations from the reception and processing operation, - if not, when the request is complete, execute an action attributed to the particular combination obtained from the values assigned to each parameter. 17. The computer program as claimed in claim 16, in which said parameters consist of one or more selection parameter (s) (PSI) and one or more information parameter (s) (PR1, PR2), the verification operation. mentioned above consisting in: - firstly, checking (37) if there remains at least one selection parameter with no assigned value and, if not, designating a batch of mandatory information parameters according to the value assigned to said audit or to each selection parameter, said batch comprising one, more or no information parameter (s), and checking (40, 42) if there remains, in said batch, at least one intelligence parameter with voice input without assigned value, said action executed is attributed to the particular combination obtained from the values assigned to each selection parameter and to each mandatory information parameter.