Speech recognition system and method for operating a speech recognition system with a mobile unit and an external server

Abstract

Speech recognition system for capturing user voice input for controlling the equipment (7a to 7d) of a vehicle, comprising a mobile unit (1) arranged in the vehicle and an external server (10), wherein the mobile unit (1) comprises: a speech capture unit (2) with a microphone (2a) for capturing voice input data, a storage unit (4) on which speech model data (SM) comprising at least one set of expressions (Ai) are stored, a speech recognition unit (3) by which a recognized text can be generated from the captured voice input data using the speech model data (SM), a positioning unit by which the destination of a route currently active in a navigation system can be determined as a future position of the mobile unit (1), and a data interface (6a).a mobile communication connection with a data interface (6b) of the external server (10) can be established at least temporarily as a data connection, whereby in areas with incomplete mobile network coverage it is not always guaranteed that the speech recognition system can exchange data with the server at all times in order to transmit the future position of the mobile unit (1) to the external server (10), wherein the external server (10) comprises a database (9) with event data to which time data, geographic position data and expressions are assigned, the external server (10) is configured to generate update data (AD) to the speech model data (SM) by comparing the time data assigned to the event data with a current time and by comparing the geographic position data assigned to the event data with the destination of the route.to generate expressions that include at least the event data of the database (9) and to transfer them from the data interface (6b) of the server (10) to the data interface (6a) of the mobile unit (1), whereby the relevance of the event data is determined based on the destination of the route, whereby event data is considered relevant where the geographic position data is not too far from the destination of the route, and the transferred expressions (Aj) can be added to the expression set (Ai) based on the update data (AD).

Description

The invention relates to a speech recognition system comprising a mobile unit and an external server. The mobile unit comprises a storage unit on which speech model data comprising at least one expression set with expressions is stored. It further comprises a speech recognition unit, by means of which a recognized text can be generated on the basis of the speech input data to be acquired, and a data interface, by means of which a data connection to a data interface of the external server can be established at least temporarily. The invention further relates to a method for operating a speech recognition system comprising a mobile unit and an external server. In this case, speech model data which comprise at least one expression set with expressions are stored on a storage unit of the mobile unit. A recognized text can be generated on the basis of the speech model data for acquired speech input data. In this case, a data connection is established at least temporarily with a data interface of the external server.Speech recognition systems find a broad spectrum of application for operating systems. This applies equally to the operation of devices and to the completion of remote verbal tasks by means of voice computers, for example telephone banking. The operation of devices of a vehicle via voice inputs has also become established as an essential feature of current user interfaces in the vehicle. For example, to input complex facts, corresponding information is interrogated sequentially by the user by a speech dialogue system.It is of decisive importance here that the recognition and interpretation of a voice input of the user takes place as free of errors and quickly as possible. For this purpose, various methods have already been proposed, but the large computing capacity required for implementation often complicates use, for example in the field of mobile devices or in vehicles. One approach to solving this problem is to offload the computing tasks to a server accessible via the Internet. However, this presupposes that the speech recognition system can exchange data with the server at any time, which is not ensured at any time, in particular in areas with intermittent coverage by mobile radio networks.Various solutions have been proposed which provide for dynamic adaptation of the speech recognition system to the respective requirements and thereby improve speech recognition. In particular, the scope of the speech inputs recognizable by the system can be extended.In the method described in WO 2011 / 054324 A1, a grammar description is created on the basis of existing structured data, for example from the Internet, and is transmitted to a speech recognition application.DE 199 42 869 A1 describes a method in which a speech control system learns new speech patterns by adaptive transcription. For example, radio station names received via RDS (radio data system) are transcribed and stored so that they can then be recognized as commands. New voice patterns can also be transmitted over the Internet.In the case of the speech recognition arrangement in a vehicle proposed in DE 10 2009 017 177 A1, a voice command is transmitted to an off-board speech recognition system. Off-board, a phonetic representation of the voice command and a meaning are determined and transmitted to the on-board system. Here, the phonetic representation and meaning of the voice command may be stored. The on-board system thus learns from the results of off-board speech recognition.The system described in EP 2 273 491 A1 for providing a voice control for a navigation device provides that the voice recognition is carried out on the basis of a dynamically generated grammar. The size of the grammar can be limited so as not to make the requirements of the system too large. Updates can be provided by a server, in particular depending on the geographical context of the voice input.EP 2 109 097 A1 describes a method in which user-dependent speech models are stored in a device of the user. Speech recognition is performed using these speech models. The speech models can be synchronized with speech models of a server.US 2015 / 0120287 A1 describes a method in which a change in the context is detected for a mobile device and a speech processing model is determined which is required in the changed context. A server obtains this language processing model for use in the mobile device.US 2014 / 0108018 A1 describes a method for updating local speech models in the device of a user. A language model for another language may be loaded when it is recognized that the user is in another country with another language.The object of the present invention is to provide a speech recognition system and a method for operating a speech recognition system, by means of which speech recognition can be carried out reliably and in a resource-saving manner.According to the invention, this object is achieved by a speech recognition system having the features of claim 1 and a method for operating a speech recognition system according to claim 8. Advantageous embodiments and refinements emerge from the dependent claims.In the speech recognition system of the above type, the external server comprises a database of event data with time data and expressions associated therewith. By means of the external server, update data can be generated to the speech model data by comparing the time data assigned to the event data with a current time, said update data comprising expressions assigned at least to the event data of the database. This update data can be transmitted from the data interface of the server to the data interface of the mobile unit. From the update data, the transmitted expressions are added to the expression set.In this specification, the word "expression set" denotes the vocabulary and phrase available to the speech recognition unit for performing speech recognition. The term "expression" refers in particular to the representation of individual words, word groups and linguistic constructions, but also statistical relationships of the individual words or word groups to one another. The whole of the expressions constitutes the expression set. This is included by the language model used for speech recognition, wherein the expression set can be supplemented by further data in the language model.In the invention, in addition to the mobile unit on which the actual speech recognition is carried out, the external server is provided by which the speech model data of the mobile unit can be updated. In this case, expressions are added to the expression set, which are taken from a database of the server. The selection of the terms thus added is made from time data associated with them. The database comprises in particular event data to which time data and expressions are assigned.The event data relate to "events", which can comprise a variety of data. In particular, this is to be understood as meaning processes which can be brought into a temporal context. For example, an event may be an event, such as a concert. In this case, the time data may include the beginning and the end of the event. The event may be associated with expressions relating to the event, such as names of persons involved, titles of pieces played, name of the game location, and terms relating to the description of the event, such as describing a musical style. In another example, the event may be a sports event associated with a start time and an end likely time. Furthermore, expressions can be assigned which relate to the sports event and the sport type, for example the names of the players and clubs involved, the name of the game site, terms from the rules of the game and terms typical for the description of the respective sport type.Events can also be taken from the messages, for example. For example, the associated terms may refer to political operations, such as the names of politicists or other persons and institutions involved, locations or technical terms called in reporting. The time data may correspond to the date of reporting, for example, in this case.Points in time and / or periods of time are thus assigned to an event, for example a date on which an event takes place or the duration of an event. Furthermore, expressions are assigned to the event data, which are relevant, for example, in connection with the respective event. These terms may include, for example, names and terms relevant to the event, but also terms and terms from the area to which the event belongs.By adding expressions which are in a temporal relationship to the update time, it can be ensured that the expression set provides expressions which are as relevant as possible in order to ensure reliable speech recognition. The comparison of the time data which are assigned to the event data with the current time can be carried out in various ways: if the time data comprises a time, the time interval between this time and the update time can be determined. In particular, it can be determined whether the event is associated with a time point lying before the update time point. For example, the event may have taken place in the past and thus may no longer be relevant. Further, the time data may include a time period, and it may be determined whether the update timing is before, after, or in the time period. This also allows a determination as to whether the respective event is relevant and is to be taken into account. In particular, a threshold value can define how far an event is in the future and can nevertheless be relevant, or whether past events are (still) relevant.The phrases associated with the events may include various words and language constructs that are likely to occur in the context of the particular event. For example, names of persons involved, location names, special destinations and further expressions relating to the event can be included. The added expressions therefore extend the expression set in a targeted manner such that particularly good and reliable speech recognition is possible. The available resources are used efficiently, since the expression set does not have to cover all possibleities from the outset and its scope only grows around the relevant expressions added.In one embodiment of the speech recognition system, a subset of expressions of the expression set can also be deleted on the basis of the update data. Thereby, the size of the print-out amount can be reduced advantageously. In particular, the size of the entire expression set can be kept substantially constant by adding one and deleting the other expressions.As the amount of expression available for speech recognition increases, the requirements for the computing capacity of the speech recognition system increase. This typically leads to the difficulty, especially in mobile applications, that the available computing capacity is not sufficient for a large amount of printout. The scope of the expression set must therefore be limited.In one embodiment of the speech recognition system, in particular fewer or non-relevant expressions can be removed from the expression set, wherein the expressions to be deleted are predefined by the update data. For example, it is thereby possible to remove those expressions which were relevant at an earlier point in time but which are associated, for example, with an event which is located in the past.Thus, the update data may indicate both expressions to be deleted from the expression set and expressions to be added. Thus, by simultaneously adding and deleting expressions upon updating, an excessive increase in the expression amount can be avoided. The reliability and response speed of the speech recognition system can thereby be optimized.In particular, the update can also comprise a general update of the expression set, wherein other expressions can also be added or removed in addition to the expressions determined in connection with the time data. In particular, the update is considered an event that is to be carried out at a specific point in time. This enables updates to the system to be performed efficiently.In a development, the transmitted expressions further comprise time data assigned to the event data. This advantageously makes it possible to determine which points in time or periods the respective expressions are assigned to.In particular, it can also be established without access to the server whether certain expressions are still current, i.e. at which time distance they are at the current time. For example, expressions relating to a past event may be removed.In a further embodiment, the mobile unit further comprises an updating unit, with which status data can be generated via the language model data and transmitted to the external server. The status data comprise information about previous changes of the printout set by update data of the server and / or about the current scope of the printout set. The update data can also be generated on the basis of the status data. As a result, it is advantageously possible to inform the server, during the generation of the update data, how the expression set is acquired before the update, for example which and how many expressions the expression set comprises.In particular, it is possible to determine how long the last update has been completed. This facilitates the determination of the expressions that must be added or removed to bring the system up to the current state. Further, it can be determined from the size of the current expression amount whether to make an enlargement by further adding expressions or whether to reduce the number of expressions.In one embodiment, context data can be captured by a context capture unit of the mobile unit and transmitted to the external server. The update data are then generated depending on the context data. This advantageously allows a determination of the expressions to be added or deleted with particularly high relevance.The context data can relate in particular to the user, the mobile unit and the environment of the mobile unit. For example, an identification of the user and / or of the mobile unit can be detected. This allows individualized updates of the speech recognition system, wherein particular settings, preferences and habits of the user can be taken into account. Furthermore, characteristics of the user, such as personal data such as age, sex and origin, may allow improved predictions of which expressions are relevant to the user. The context data about the mobile unit may include, for example, which devices and applications are activated or available. In this way, the update data can be matched to which further applications of the mobile unit are available.The context data about the environment of the mobile unit can be determined, for example, on the basis of the geographical position of the mobile unit, but also, for example, by a sensor system which supplies data about the environment. For example, the proximity of other mobile devices may be determined based on short range signal transmissions.In one embodiment, the context data can be acquired by further devices, in particular by a navigation device and / or an electronic appointment calendar. This advantageously makes it possible to determine with high probability whether specific expressions and events are relevant.The context data acquired by the navigation device can comprise, for example, the current position of the mobile unit, a planned destination of a route or geographical information about the environment of the mobile unit. Event data can be acquired by an electronic appointment calendar with time data, which is assigned to the user in particular. For example, the location and time of an upcoming deadline may be detected and the update data may then be generated such that expressions relevant to that deadline are added to the expression set. Conversely, expressions relevant to a past deadline may be deleted, for example.In one embodiment, geographical position data are also assigned to the event data of the database. A current position of the mobile unit can be determined by means of a position determination unit and transmitted to the external server. The update data is also generated by a comparison of the geographical position data assigned to the event data with the position.This also allows the relevance of the events to be determined more accurately based on the position of the mobile unit, in particular in conjunction with context data from a navigation device. However, the determination can also be made using other methods, for example by a radio connection with fixed stations at a known position or using inputs from the user.In particular, such events can be recognized as relevant, whose geographical position data do not exceed a predefined maximum distance from the position of the mobile unit. In particular, it can be assumed that events at a large spatial distance from the user are less relevant than near events. Conversely, on the basis of the spatial proximity between the mobile unit and an event, it can be concluded that expressions relevant for the event are very likely to be used and must be recognized. For example, a high relevance for a concert can be determined if this takes place in the immediate vicinity of the user, while a sports event at a great distance is not classified as relevant.In a further embodiment, instead of the current position of the mobile unit, a position lying in the past can also be transmitted to the external server. In this way, event data can be recognized as relevant, which are relevant in the rear view, for example in order to allow the user an input with respect to a previous event.According to the invention, instead of the current position of the mobile unit, a position to be reached in the future is transmitted to the external server, namely the destination of a route which is currently active in the navigation system. This allows in particular the recognition of speech inputs with respect to an event to which the user is currently on the route.In a further development, statistical data for the acquired voice input data can also be acquired by the voice recognition unit and transmitted to the server. The update data can then be generated by the server as a function of the statistical data. This advantageously allows a high level of personalization of the speech recognition system.In particular, it is possible to determine the frequency with which a user has used certain expressions in past voice inputs. Frequently used terms may indicate that the term is highly likely to be used in the future. Conversely, less used terms may mean low relevance. It can thereby be ensured that seldom or unused expressions are removed from the expression set, while frequently used expressions are not removed and / or other, for example, related expressions can be added.In addition to statistical analysis, machine learning techniques may also be used to identify key words and determine non-key words for deletion.In one embodiment, the mobile unit is arranged in a vehicle. The user can be, in particular, the driver of the vehicle. This advantageously allows devices of the vehicle to be operated by means of voice control.The speech recognition system can have access to the data of other devices arranged in the vehicle. The reliable and rapid functioning of a speech recognition system is of particular importance in vehicles because the driver of the vehicle is to be distracted as little as possible from the traffic situation. Moreover, because of the movement of the vehicle, it must be taken into account that the relevance of individual expressions can change as a function of the geographical position and the time. With the limited performance of the vehicle embedded computing systems, this means that the expression set must be updated again and again if the amount of the expression set is not to become too large.In the method according to the invention of the above-mentioned type, the external server comprises a database with event data to which time data and expressions are assigned. By means of the external server, by comparing the time data assigned to the event data with the current time, update data to the language model data, which comprises at least expressions assigned to the event data of the database, are generated and transmitted to the mobile unit. Based on the update data, the transmitted expressions are then added to the expression set. The method is used in particular for operating the speech recognition system. It thus has the same advantages as the speech recognition system.In one development, a subset of expressions of the expression set is furthermore deleted on the basis of the update data. Thereby, the amount of expression can be reduced advantageously.The invention will now be explained on the basis of exemplary embodiments with reference to the drawings. FIG. 1 shows an embodiment of the speech recognition system; and FIG. 2 shows an embodiment of the method for operating the speech recognition system.Referring to Fig. 1, an embodiment of the speech recognition system will be explained.The speech recognition system comprises a mobile unit 1 and an external server 10. The mobile unit 1 comprises a speech detection unit 2 with a microphone 2a. In the exemplary embodiment shown, the user can signal the start of a voice input by pressing a push-to-talk switch (PTT), whereupon the voice detection unit 2 detects and records the voice input data spoken by the user. Intermediate storage is carried out, so that the voice input data is available for further use.A speech recognition unit 3 is coupled to the speech detection unit 2. By means of this, a recognized machine-readable text can be generated on the basis of the acquired voice input data. For this purpose, speech model data SM stored on a storage unit 4 are used. The speech model data SM comprise at least one expression set A i with expressions. The terms of the expression set A i describe in particular a vocabulary and phrase used for recognizing the text. The whole of the speech model data SM is used as a speech model for speech recognition.The mobile unit 1 further comprises an update unit 5 coupled to a data interface 6 aof the mobile unit. The mobile unit 1 further comprises a context acquisition unit 8, which in the illustrated case is coupled to the speech recognition unit 3, the update unit 5 and four further devices 7 ato 7 d. One of the devices 7a to 7d is a navigation device, another device is an electronic appointment calendar.A data connection can be established at least temporarily via the data interface 6 aof the mobile unit 1 to the data interface 6 bof the external server 10. the external server 10 comprises a database 9. Furthermore, the connection can be established via a telecommunication network, such as a telephone network, or a wireless local area network (WLAN). Furthermore, the data connection can be effected by the connection of a data cable. The connection can also be established via another unit which can itself establish a connection to the external server 10. For example, a data connection can exist between the mobile unit 1 and a mobile telephone connected to the Internet, for example by a data cable or a radio connection, for example by Bluetooth. In particular, the connection to the external server 10 is established via the Internet.With reference to FIGS. 1 and 2, an exemplary embodiment of the method is explained.As explained above, the storage unit 4 stores speech model data SM including a print-out quantity A i therein. A data connection is established via the data interfaces 6 aand 6 bof the mobile unit 1 and the server 10, so that the language model data SM can be updated. For this purpose, the update unit 5 generates status data SD via the language model data SM. The status data SD comprise in particular information about the time at which the last update of the speech model data SM was made, how large the expression set A is i and which is the maximum desired size of the expression set A i.Furthermore, context data KD are acquired by context acquisition unit 8, data of the other devices 7 ato 7 dbeing acquired in particular. The context data KD comprise, among other things, the geographical position of the mobile unit 1, a route currently output by the navigation system and geographical information about the environment. Furthermore, the context data KD comprise data from a deadline calendar, in particular times, geographical positions and description texts of pending deadlines. The context data KD further comprise statistical data on how frequently individual expressions of the expression set A i were used in past voice inputs. In particular, the terms from the term set A are comprised of i which have been used particularly frequently or particularly rarely.The status data SD and the context data KD are transmitted to the server 10. This includes the database 9 with event data. In particular, data sets correspond to individual events. Time data are assigned to the event data, for example the time of a future event and / or the duration of the event. The event data is also assigned expressions. These are, in particular, expressions which are particularly relevant for voice inputs in connection with the respective result.In the example shown, one of the events is a concert named in the user's calendar and to which additional information is available on the server. In this case, the time data includes the beginning and the end of the concert. The concert will be assigned expressions such as the name of the interpreter, the titles of the played and other pieces, the place of play, and terms typically used in a description of music, particularly for the musical style.Furthermore, a sports event is included by way of example, with which a time for the start and a time for the prospective end are assigned. Furthermore, expressions are assigned which relate to the sports event and the sport type, such as the names of players and clubs, playgrounds, rules of play and terms which are typical for the description of the respective sport type.Events are also extracted from the messages. The associated terms refer to, for example, political processes, such as the names of politicists, locations or technical terms mentioned in the report. The time data corresponds to the date of reporting and a "expiration date" is determined, after the expiration of which the event is interpreted as less relevant or no longer relevant.Update data AD is now generated. These include at least expressions A j, which are associated with events in the database. In the example shown, for this purpose, on the basis of the current point in time, i.e. the point in time of the update, each event is considered to be relevant, the time data of which make it possible to identify that the event is still current, and in which the geographical position data is not too far away from the current position of the mobile unit 1 or from the target position at the end of the route determined by the navigation device. In particular, those events are considered which are in the future or whose duration has not yet ended. For events for which such a duration is not clearly defined, such as for the information from the messages, a defined time interval, such as beginning at the time of reporting the event, is determined as the duration.The update data AD further includes information on which expressions A k are to be deleted from the expression set A i. For example, this may be the set of expressions that were added at a previous update and that are no longer up-to-date. Further, these may be, for example, rare or unused terms. The update data AD is transmitted from the server 10 to the mobile unit 1. The update unit 5 performs changes to the language model data SM on the basis of the update data AD, wherein in particular a subset AK of the expression set A i is deleted while the expressions A j are newly added. Specifically, the scope of the expression set A i remains substantially constant.In a further exemplary embodiment, the mobile unit 1 is arranged in a vehicle. The devices 7 ato 7 dare in this case further devices of the vehicle, including various driver assistance systems, the navigation system, an infotainment system and sensors of the vehicle. As a result, in particular the context acquisition unit 8 can determine the context data KD in a more targeted manner, with the result that the update data AD is generated with higher accuracy for the respective situation.In particular, the speech recognition system can be used in the vehicle to capture speech inputs of the user for controlling devices 7 ato 7 dof the vehicle, for example the navigation device. In particular, the update data AD are determined in such a way that the expression set A i enables reliable and fast speech recognition in connection with special destinations in the environment of a planned route.List of reference characters1 Mobile unit 2 Voice acquisition unit 2 a Mikrofon 3 Voice recognition unit 4 Storage unit 5 Update unit 6 aData interface of mobile unit 6 bData interface of external server 7 a, 7 b, 7 c, 7 d Einrichtungen 8 Context acquisition unit 9 Database 10 External server A i Expression set Aj Expressions of database A k Expressions to be deleted AD Update data KD Context data SD Status data SM Voice model data

Claims

Speech recognition system for detecting speech inputs of a user for controlling devices (7a to 7d) of a vehicle, having a mobile unit (1) arranged in the vehicle and an external server (10), wherein the mobile unit (1) comprises: a speech detection unit (2) having a microphone (2a) for detecting speech input data, a storage unit (4) on which speech model data (SM) comprising at least one expression set (A i) having expressions are stored, a speech recognition unit (3) by means of which a recognized text can be generated on the basis of the speech model data (SM) for the detected speech input data, a position determination unit by means of which the destination of a route which is currently active in a navigation system is generated, as a position of the mobile unit (1) to be reached in the future, and a data interface (6a) by means of which a mobile radio connection can be established at least temporarily as a data connection with a data interface (6b) of the external server (10), wherein it is not ensured at all times in areas with intermittent coverage by mobile radio networks that the speech recognition system can exchange data with the server at all times in order to transmit the position of the mobile unit (1) to be reached in the future to the external server (10), wherein the external server (10) comprises a database (9) with event data to which time data, geographical position data and expressions are assigned, the external server (10) is designed, by comparing the time data associated with the event data with a current time and by comparing the geographical position data associated with the event data with the destination of the route, generating update data (AD) to the language model data (SM) comprising expressions associated with at least the event data of the database (9) and transmitting them from the data interface (6b) of the server (10) to the data interface (6a) of the mobile unit (1), wherein a relevance of the event data is determined on the basis of the destination of the route, wherein such event data are considered to be relevant in which the geographical position data are not too far from the destination of the route, and based on the update data (AD), the transmitted expressions (A j) can be added to the expression set (A i).Speech recognition system according to claim 1, characterised in that a subset (A k) of expressions of the expression set (A i) can be deleted on the basis of the update data (AD).A speech recognition system according to any preceding claim, characterized in that the transmitted expressions (A j) further comprise time data associated with the event data.Speech recognition system according to one of the preceding claims, characterized in that the mobile unit (1) further comprises an updating unit (5) with which status data (SD) can be generated via the speech model data (SM) and transmitted to the external server (10), the status data (SD) comprise information about previous change of the expression set (A i) by updating data (AD) of the server (10) and / or about the current scope of the expression set (A i) and the updating data (AD) can furthermore be generated on the basis of the status data (SD).Speech recognition system according to one of the preceding claims, characterized in that context data (KD) can be captured by a context capture unit (8) of the mobile unit (1) and transmitted to the external server (10), and the update data (AD) are generated further as a function of the context data (KD).Speech recognition system according to claim 5, characterised in that the context data (KD) can be captured by further devices (7a to 7d), in particular by a navigation device and / or an electronic appointment calendar.Speech recognition system according to one of the preceding claims, characterized in that statistical data relating to the captured speech input data can also be captured by the speech recognition unit (3) and transmitted to the server (10), and the update data (AD) can be generated by the server (10) as a function of the statistical data.Method for operating a speech recognition system for detecting speech inputs of a user for controlling devices (7a to 7d) of a vehicle, having a mobile unit (1) arranged in the vehicle and an external server (10), in which a speech detection unit (2) detects speech input data using a microphone (2a), a position determination unit determines the destination of a route which is currently active in a navigation system as a position of the mobile unit (1) to be reached in the future, speech model data (SM) which comprise at least one expression set (A i) with expressions are stored on a storage unit (4) of the mobile unit (1), A recognized text can be generated on the basis of the speech model data (SM) with respect to the acquired speech input data and a mobile radio connection is established at least temporarily as a data connection to a data interface of the external server (10), wherein it is not ensured at all times in areas with gapous coverage by mobile radio networks that the speech recognition system can exchange data with the server at all times in order to transmit the position of the mobile unit (1) to be reached in the future to the external server (10), wherein the external server (10) comprises a database with event data to which time data, geographical position data and expressions are assigned, by the external server (10), by comparing the time data associated with the event data with a current time and by comparing the geographical position data associated with the event data with the destination of the route, update data (AD) are generated and transmitted to the mobile unit to the language model data (SM) comprising expressions (A j) associated with at least the event data of the database, wherein a relevance of the event data is determined on the basis of the destination of the route, wherein such event data are considered to be relevant in which the geographical position data are not too far from the destination of the route, and on the basis of the update data (AD), the transmitted expressions (A j) are added to the expression set (A i).Method according to Claim 8, characterized in that the update data (AD) is used to further delete a subset (A k) of expressions of the expression set (A i).

Images