WO2023030767A1 - Method and device for outputting offers - Google Patents

Abstract

The invention relates to a method for outputting offers to a user of a vehicle, comprising the following steps: determining a contact (7), in which the user is with teh vehicle; determining available offers (5) for outputting to the user; determining a degree of similarity between context and offers (11). According to the invention, the user is identified (12), a relevance value is allocated to the offers corresponding to a degree of similarity to the context, a personalized, stored acceptance evaluation of the identified user is read (13) and the relevance value of the offers is adjusted according to the associated acceptance evaluation. Offers having an adjusted relevance value, which is situated above a predetermined first threshold value, are outputted to the user.

Description

Mercedes-Benz Group AG

Method and device for issuing offers

The present invention relates to a method and a device for issuing offers to a vehicle user.

It is known from DE 10 2017 207 466 A1 to determine information offers depending on a thematic context on a travel route and to output them to a user. Advertising information provided by a service provider with an associated context description is compared with a thematic context of a travel route and is transmitted to the vehicle user when a predetermined degree of similarity is reached. The disadvantage of the method presented is that personal preferences of a user are not taken into account.

The object of the present invention is to provide a method that is improved over the prior art and an associated device for issuing offers.

The object is achieved by a method having the features of claim 1 and a device according to claim 10. The dependent claims define preferred and advantageous embodiments of the present invention.

In the method according to the invention, a user is identified, a relevance value is assigned to the offers according to a degree of similarity to the context, a personalized stored acceptance rating of the identified user is read out associated with the offers and the relevance value of the offers is adjusted according to the associated acceptance rating, with only the offers with a above a predetermined adapted relevance value lying at the first threshold value can be output to the user. The user is identified by manually entering the identity, for example using biometric identification methods fingerprint or retina sensing and/or carried mobile devices. The vehicle context and associated designation is determined using a vehicle sensor system that recognizes, for example, the traffic situation, the weather, the environment, objects in or around the vehicle, the location, the type of road, etc. The term offer refers to any type of content that is provided by radio, internet, databases, the vehicle itself or sensors for collecting information from the environment and is described or classifiable with a designation, ie information, songs, films, settings etc. A degree of similarity of the context to the offers is determined by methods known from the prior art for determining the semantic similarity of words, ie designations of offers and contexts. The degree of similarity is determined, for example, using the word2vec or GloVe method. A relevance value is assigned to the offers according to the ascertained degree of similarity to the context, ie the relevance value is high or low for a high or low degree of similarity. In order to ensure that an offer is also individually relevant for a vehicle user, a personalized database with acceptance ratings associated with the offers is read out. According to the acceptance rating, the relevance rating is adjusted and converted into an adjusted relevance rating. If the acceptance rating is high, the relevance rating is increased, and if the acceptance rating is low, it is correspondingly slightly increased or decreased. Offers that have an acceptance rating above a threshold value are issued to the user, the other offers are not taken into account. Advantageously, offers tailored to the user's preferences are output to the user depending on the situation.

In a further preferred embodiment, the acceptance rating is determined anew each time an offer is issued and stored in a personalized manner. Offers displayed for the first time are provided with an acceptance rating, offers displayed repeatedly are updated with a current acceptance rating, for example by averaging all previous acceptance ratings for this offer. The acceptance rating is determined from a user reaction, whereby the user's emotions and interaction are monitored. The emotions can be determined, for example, from a facial expression captured by a camera, or also via voice evaluation, the user reaction includes, for example, canceling a piece of music, rejecting an offer or accepting an offer without restrictions. The output of an undesired offer can advantageously be avoided.

According to one embodiment of the present invention, the adapted relevance ratings for the offers are compared with a second threshold value and a blocking time for repeating a display of the offers is increased or decreased if the acceptance rating is below or above the second threshold value. The blocking time specifies the time after which an offer can be issued again. A repetition rate is thus advantageously weighted in favor of offers with a high level of acceptance.

In a further developed embodiment, only a limited number of offers that are above the first threshold value and are prioritized according to the adjusted relevance value are output. The number of offers to be output is limited to four, for example, i.e. only the four offers with the highest adjusted relevance values above the threshold value are output. This avoids an information overload for the user.

In a further additional embodiment, different users are clustered in the same context or at least in a context that has a predetermined degree of similarity, with the acceptance values for offers by a user that are above the threshold value being transmitted to other users in the same and/or similar context, preferably to users at for which no acceptance rating has yet been saved for the corresponding offers. The same offers are possible for users in the same context, so that a first starting value is also available for a user without an acceptance evaluation of an offer. The similarity of the contexts is in turn determined using context designations using a known method for determining the degree of similarity. Advantageously, the method enables acceptance evaluations of offers to be built up quickly in storage devices personally assigned to users. In a further embodiment, a degree of similarity of acceptance ratings of offers from different users is determined, users with at least a predetermined degree of similarity having acceptance ratings are clustered into groups, with acceptance ratings of offers being transferred to other users in the same group, for whom no corresponding offers with an acceptance rating have been saved so far are. It is assumed that users with similar acceptance ratings have similar interests and characteristics and will therefore evaluate offers with a similar acceptance rating. On this basis, within a group, an acceptance rating of an offer from one user is transferred to other users who have not saved a personalized acceptance rating for this offer.

In an advantageous manner, the method offers a further possibility for a rapid build-up of acceptance evaluations of offers in storage devices personally assigned to users.

According to a further refinement, offers without stored acceptance ratings determined from user reactions are output with a relevance value determined from the degree of similarity which is above a third threshold value. In the absence of an acceptance rating for an offer, only the criterion of the relevance value is used as an alternative, so that an initial display is made possible and a personalized acceptance rating can be determined, by means of which an adapted relevance rating can be generated when an offer is repeated. Alternatively, an acceptance rating can be preassigned to all offers with default values. In a further alternative, the relevance ratings are generated by means of user training. For this purpose, offers of different categories are issued to the user and the acceptance rating for the respective offers derived from the user reaction is saved.

The device according to the invention comprises: sensors for determining a context in which the vehicle is located means for determining available offers and for determining a degree of similarity between the context and offers, characterized by a device that identifies a user, which assigns a relevance value to the offer that corresponds to the degree of similarity to the context, which reads a personalized stored acceptance rating of the identified user from a memory associated with the offers and which adjusts the relevance value of the offers according to the associated acceptance rating, and output means which only display the offers with a value above a predetermined first Threshold lying adjusted relevance value issued to the user.

The device advantageously makes it possible to determine offers that match the context of the vehicle and are output in accordance with the preferences of a user.

The present invention is explained in more detail below using exemplary embodiments with reference to the attached drawing, in which the same or similar parts are denoted by the same reference symbols.

It shows:

1 shows a flowchart of a process sequence,

FIG. 2 clustering of users in a similar or the same context, FIG. 3 clustering of users with similar acceptance ratings, FIG. 4 user training to build up acceptance ratings, and FIG. 5 vehicle with a device according to the invention.

According to FIG. 1 , available offers and the context of the vehicle are recorded in step 1 . In step 3, the offers are processed and sorted, for example, by category, genre, content or format. Information and functions from all sources are available as an offer, ie information and offers from the Internet or radio, information from the vehicle itself, possible vehicle functions, apps, etc. In step 5, terms for the offers are defined. Features such as category, topic, keywords, headline, tags, etc. are extracted from the offers, for example using computer-aided linguistic processes. In step 7, the received context data are processed in the parallel branch. The context data includes data from vehicle sensors, for example from a rain sensor, from an outside or inside camera, data from the navigation device, etc. The context describes the situation in which the vehicle is located, i.e. weather, time of day, landscape, type of road, country, etc. In step 9, the context data are defined analogously to step 5 with terminology, ie provided with keywords.

Next, in the next step 11a, a measure of similarity between the context and the offers is formed. There are various methods for this, for example Word2Vec word embedding. Word2Vec assigns a vector to words; by comparing the vectors, the similarity between the words can be quantified.

The degree of similarity can be mapped on a numerical scale. Offers with a high degree of similarity to the context are assigned a high relevance in step 11b, offers with a lower degree of relevance are given a low relevance value, i.e. the smaller the degree of similarity, the smaller the relevance value. In one embodiment, the degree of similarity can also be identical to the relevance value, so that an assignment is also to be understood as taking over the degree of similarity as a relevance value.

In step 13, a database is accessed in which user-specific acceptance ratings for offers are stored. The identity of the user is determined beforehand in step 10. The respective acceptance is derived from the reaction of the user during or after the issue of an offer, in that his emotions are determined using a facial expression determined by a camera and/or an interaction, for example acceptance or rejection of the offer. Acceptance is preferably mapped to an acceptance scale.

If the offers have already been evaluated according to this database, an adjusted relevance value is formed, i.e. if the value on the acceptance scale is high, the relevance is increased, if the evaluation is low, it is increased less, or if the value is negative, it is even reduced.

A first threshold value is now queried in step 15 , with only offers whose adjusted relevance value is greater than this threshold value being output to a user in step 19 and otherwise discarded in step 17 . In step 21, the database is updated each time an offer is issued by determining an acceptance value. The offer and/or features of the offer are stored together with the acceptance value in the vehicle or outside of the vehicle in the personalized database 23. The database is therefore always updated during use, so that new offers that have not yet been evaluated are also stored with an acceptance value.

In a further step 25, the acceptance value is compared with a second threshold value. If the acceptance value is greater than the second threshold value, a blocking time delta_t is reduced in step 27, otherwise increased in step 29. Each offer, i.e. also the offer associated with the acceptance value determined in step 21, is assigned a blocking time within which repeated display within the blocking time is prevented. The blocking time delta_t is taken into account in recording step 1, so that an associated blocking time can be assigned to possible offers from the Internet, radio, etc. that have already been issued to the user in the past and this is retained until the blocking time has expired, i.e. not processed further according to the following steps become.

This ensures that the user is rarely confronted with offers rated as low acceptance, but is frequently confronted with high rated offers.

In a specific, greatly simplified example in the method according to FIG. 1, a soccer stadium is recorded as the context using a route stored in the navigation device. Recorded possible offers from the Internet are “Goal football team A”, “Goal football team B” and “Goal handball team C”. The method now forms a measure of similarity and now recognizes the similarity from the context of soccer stadium to soccer and thus forms a high relevance value of 8 for soccer or soccer teams and a low relevance value of 4 for handball or handball teams. In the database, a positive acceptance value of 4 is stored for "Goal Soccer Team A" and a negative acceptance value of -4 for "Goal Soccer Team B". This results in the following adjusted relevance: "Goal soccer team A" 8+4 =12, "Goal soccer team B": relevance 8-4 = 4 and for "Goal handball team C" 4, ie the relevance value without adjustment. After comparison with the first threshold with a value of 8, the user is only offered the "Goal football team A" with an adjusted relevance value greater than 8. If the user reacts positively, the acceptance value in the database is updated with a value of 4. Since the adjusted relevance value is also greater than the second threshold value, the blocking time delta_t is reduced by a specified value, for example by 10 s.

2 illustrates a clustering of users in the same or at least a similar context. In the present example, the users 100, 102 and 104 are in the same context and are driving on the same street, for example. The users and the personalized database 106 comprising the acceptance ratings associated with each of the users are stored together in a memory according to the context in step 108 . The acceptance ratings for offers relevant to this context are transferred to the personalized databases of other users 110 located in the same context, so that the method according to FIG. 1 can be used accordingly for offers in this context.

According to FIG. 3, a measure of similarity between databases 106, 110 containing acceptance ratings from different users is determined in a functional step S112. Similar databases 106 with a filling level above a minimum value are combined and the acceptance rating, in particular for non-rated offers in databases 110 that are sparsely populated, is transmitted in step 114, so that in the following step 116 the method according to FIG. 1 can be carried out for the user of database 110 with increased accuracy in terms of the best possible personalized context-related output.

FIG. 4 illustrates a possibility for filling a database 106 that is not or not very full of acceptance ratings for offers, in which in method step S120 a user is presented with various offers in a training phase and associated acceptance ratings are determined and stored in a personalized manner.

FIG. 5 shows a highly abstracted device for carrying out the above method for the personalized output of offers and information to a driver of a vehicle. Sensors 200, for example a camera, determine this or a GPS sensor, a context in which the vehicle is located. The vehicle further comprises means 202 for determining available offers of information and for determining a degree of similarity between context and offers. The means a telecommunications module for connection to the Internet and a module for determining the degree of similarity. The module determines terms such as category, topic, keywords, headline, tags for the determined context and offers, for example using linguistic processes. Using the terms, the degree of similarity between the context and the offers is determined using known methods.

A device 204 identifies a user so that data can be retrieved and stored in a personalized manner. The device 204 assigns a relevance value corresponding to the determined degree of similarity to the context to the determined offers. Furthermore, the facility reads the personalized, stored acceptance ratings of the identified user associated with the offers from a memory 206 and adjusts the relevance value of the offers according to the associated acceptance rating. Output means 208 only output offers with an adapted relevance value above a predetermined first threshold value to the user.

Claims

Claims Method for issuing offers to a user of a vehicle with the following steps: - Determination of a context (S9) in which the user is located with the vehicle, -Determination of available offers (S5) for output to the user - Determination of a degree of similarity between context and offers (S11a), characterized in that the user is identified (S10), a relevance value is assigned to the offers according to a degree of similarity to the context (S11b), a personalized stored acceptance rating of the identified user is read out associated with the offers ( S13) and the relevance value of the offers is adjusted according to the associated acceptance rating, and only the offers with an adjusted relevance value above a predetermined first threshold value are output to the user (S19). Method according to Claim 1, characterized in that the acceptance rating is determined with each issue of an offer based on the reaction of a user and is stored (S21) in a personalized manner with the offer. Method according to claim 2, characterized in that the acceptance rating includes an emotion rating and/or an interaction rating. Method according to one of Claims 1 to 3, characterized in that the adapted relevance ratings for the offers are compared with a second threshold value (S25) and a blocking time for a repetition of a display of the offers is increased (S29) or decreased (S27) if the acceptance rating is below or above the second threshold. Method according to one of Claims 1 to 4, characterized in that only a limited number of offers which are above the first threshold value and which are prioritized according to the adjusted relevance value is output. Method according to one of Claims 1 to 5, characterized in that different users are clustered (S108) in the same context or at least in a context having a predetermined degree of similarity, with the acceptance values for offers above the threshold value from a user to other users in the same and /or be transmitted in a similar context, preferably to users for whom no acceptance rating is stored for corresponding offers. Method according to one of Claims 1 to 6, characterized in that a degree of similarity of acceptance ratings of offers from different users is determined, users with acceptance ratings having at least a predetermined degree of similarity are clustered into groups (S112), acceptance ratings of offers being based on non-rated same and/or or similar offers from other users in the same group. Method according to one of claims 1 to 7, characterized in that Offers without a stored acceptance rating determined from user reactions are output with a relevance value determined from the degree of similarity above a third threshold value. The method according to any one of claims 1 to 8, characterized in that Acceptance rating for offers by means of a user training (S120) are determined and saved personalized. Device for carrying out the above method for issuing offers to a driver of a vehicle with: - Sensors (200) for determining a context in which the vehicle is located, - Means (202) for determining available offers and for determining a degree of similarity between the context and offers, characterized by a device (204) that identifies a user, which assigns a relevance value corresponding to the degree of similarity to the context to the offers, which the offers associated reads a personalized stored acceptance rating of the identified user from a memory (206) and adjusts the relevance value of the offers according to the associated acceptance rating and Output means (208) which only outputs the offers with an adapted relevance value above a predetermined first threshold value to the user.