Method and device for providing information on a vehicle malfunction

Abstract

The invention relates to a method and a device for providing information on a malfunction of a vehicle, said method comprising the steps of: Receiving (210) a generated text; Determining (220) a list of labels from said received text and a first predetermined model; Receiving (230) first data; Receiving (240) second data; Determining (250) a first list of words from said list of labels, said first and second data and a predetermined knowledge graph; Determining (260) a second list of words from the first list of words, said second data and a second predetermined model; Determining (270) a list of sentences from said second list of words and a third predetermined model; Emitting (280) said list of sentences to said audio system to provide information on said malfunction of said vehicle by speech synthesis.Figure to be published for the summary: Figure 2.

Description

Title of the invention: Method and device for providing information on a malfunction of a vehicle Technical field of the invention

[0001] The invention is in the field of autonomous vehicle driver assistance systems. In particular, the invention relates to a method and a device for providing information on a malfunction of a vehicle equipped with a speech recognition and synthesis audio system, in order to inform an occupant of said vehicle, said occupant issuing a verbal request. State of the art

[0002] The term "vehicle" means any type of vehicle such as a motor vehicle, moped, motorcycle, warehouse robot, etc. "Autonomous driving" of a "vehicle" means any process capable of assisting the driving of the vehicle; the "vehicle" is then also called an "autonomous vehicle." The process may thus consist of partially or fully controlling the vehicle or providing any type of assistance to a person driving the vehicle. The process thus covers all autonomous driving, from level 0 to level 5 in the OICA (International Organization of Motor Vehicle Manufacturers) scale.

[0003] Vehicles sold today include a wide range of features and are becoming increasingly complex to meet regulatory requirements such as emissions standards. Unexpected noises or malfunctions may lead a driver to believe that a problem exists with their vehicle. To verify the problem, the driver must schedule an appointment with the after-sales service and have their vehicle diagnosed. Some of these problems can be resolved or explained through remote communication (telephone, internet message, etc.) with the after-sales service.

[0004] After-sales service includes numerous diagnostic aid documents, referred to as Technical Documents. These documents may include repair methods, tool usage methods, methods for resolving identified anomalies (TSBs, from the English "Technical Service Bulletin"), diagnostic methods, diagrams, fault trees, etc. Based on this information, the after-sales service is able to answer customer questions. A "diagnostic guide" is defined as a set of diagnostic aid documents.

[0005] However, some anomalies appear only fleetingly or intermittently In specific life situations, the conditions under which the problem occurred during a service appointment are no longer reproduced. It is often very difficult to reproduce the problem and therefore to understand what is happening, explain the cause to the driver, and repair the issue if necessary. Also, sometimes what a driver perceives as a problem is actually normal operation.

[0006] Artificial intelligence, or AI, is defined as a machine or computer system designed using large amounts of data and capable of providing decisions or queries based on an expressed need. It is known that the design of these systems is based on learning a model from expressed needs, also called inputs, and expected decisions, also called outputs. The inputs and outputs of artificial intelligence are numerical and interpretable by computer. This is referred to as a machine learning model; the models are pre-trained (learning phase).

[0007] Natural Language Processing (NLP) is a field of artificial intelligence that focuses on the interaction between computers and human language. For example, NLP relies on machine learning techniques and models, such as neural networks, language models, classification and clustering algorithms, statistical methods, and so on. These techniques enable NLP systems to process and understand the structures, meanings, and nuances of natural language.

[0008] For example, version 3.5 of the artificial intelligence known as "ChatGPT®" is an NLP whose training data size contains 175 billion parameters to be trained. The training dataset contains 300 billion words, or approximately 570 gigabytes of textual data.

[0009] Currently, NLP models are far too large (memory) and consume too many computing resources (CPU time, processor power, etc.). These models cannot be embedded in a vehicle's computer (or ECU, for "Electronic Control Unit") while delivering near real-time services, as the computer must be compatible (with variations in temperature, humidity, etc.) for use in a "mass-market" vehicle (sold in tens of thousands of units).

[0010] It is known to use a "smartphone," a mobile phone or multifunction device, connected to the "cloud" (remote computer server, or server). This smartphone may include an application connected, via an electromagnetic link (4G, 5G, etc.) or similar, to this cloud and to an AI of the ChatGPT® type. However, on the one hand, this requires a good quality connection and generally a paid subscription, and, on the other hand, these AIs are not suitable for providing in Vehicle operation training based on voice commands, including providing contextualized information based on a vehicle's make and / or model. Furthermore, this implies that voice commands are transmitted via the cloud and recorded by AI. A user may wish to prevent their conversations, made in the car, from being recorded by a third party outside the vehicle.

[0011] It is also known as speech recognition audio systems that convert human speech into written text. These systems work using signal processing algorithms that analyze the audio and identify the different components of language, such as phonemes, words, intonations, etc.

[0012] Some of these speech recognition systems are simpler and can be implemented in a vehicle. There is no conversion of human speech into written text. Based on the identification of intonations or a predetermined sound sequence, requests for actions such as "turn off" or "turn on" a car radio are recognized. These are simpler systems, but they are limited to a very small number of language component identifications. The expressed need is limited to a very small number (about ten) of functions. These systems are very sensitive to ambient noise, intonation, accents, etc., and therefore cannot provide a wide variety of operating information. Summary of the invention

[0013] An object of the present invention is to remedy the aforementioned problem, in particular to provide an occupant of a vehicle with information on a malfunction of a vehicle without having to contact the after-sales service and at a time when the malfunction is present and detectable on the vehicle.

[0014] To this end, a first aspect of the invention relates to a method for providing information on a malfunction of a vehicle equipped with a speech recognition and synthesis audio system, in order to inform an occupant of said vehicle, said occupant issuing a verbal request, said method being implemented by a processor (103) and comprising the steps of: • Receiving a text generated from said audio system processing said verbal request; • Determining a list of labels from the received text and a predetermined initial template, • said first model being a machine learning model in the field of natural language processing, • said first model being driven from • a predetermined set of texts representative of a verbal request for information of a malfunction in at least one vehicle, and • a predetermined set of said labels, • said labels being a group of characters, at least one word, used to identify and describe sets of words from said predetermined set of texts representative of a verbal request for the provision of information on malfunction of at least one vehicle; Reception of initial data, referred to as vehicle data, said initial data being related to said vehicle, said data including at least one identifier allowing the determination of a make and / or model of said vehicle, and including at least one piece of information on a state of said vehicle such as speed, temperature, ...; Reception of second data, called diagnostic data, said second data being related to diagnostic data of said vehicle such as a fault code, a list of fault codes, ...; Determining a first word list from said label list, said first and second data received, and a predetermined knowledge graph, • said graph comprising • nodes, said nodes being representative of entities, the entities being predetermined from • said labels, and • of a set of at least one word, said at least one word characterizing a predetermined set of the first and second data received; • arcs, said arcs being representative of a functional relationship between said entities, said arcs being described by at least one word; Determining a second word list from the first word list, the second set of received data, and a second predetermined model, • said second model being a machine learning model in the field of natural language processing, • said second model being driven from • from a predetermined set of initial word lists, and • a predetermined set of second data; Determining a list of sentences from said second list of words and a third predetermined model, said third model being a machine learning model in the field of natural language processing, • said third model being at least trained from a set of second word lists; • Transmission of said list of phrases to said audio system to provide information on said malfunction of said vehicle by speech synthesis.

[0015] Thus, the first, second, and third models are defined to the bare minimum, and are therefore of minimal size, making them suitable for execution on an automotive ECU (electronic control unit) (a vehicle marketed to the general public). The limited memory size of these models allows for downloads within reasonable timeframes, without requiring the vehicle to be idle for several hours. These models can then be updated in a reasonable time, more quickly than models not defined to the bare minimum. The voice request remains internal to the vehicle's computer system; it is not transmitted to a remote server. The confidentiality of the "conversations" is thus preserved. Furthermore, relevant information about the anomaly is provided at the appropriate time when the anomaly occurs.

[0016] Advantageously, said list of phrases is further determined from said vehicle data and diagnostic data, said second model being further trained from metadata relating to a predetermined set of vehicle data and diagnostic data.

[0017] Thus, the phrase list is contextualized according to the vehicle and the needs. Said second model is even more optimized.

[0018] Advantageously, said vehicle data also include data relating to said occupant.

[0019] Thus, the list of phrases is also contextualized according to the occupant. The information provided to the occupant is personalized, resulting in an improved perceived quality of the vehicle. The information provided can also be more precise.

[0020] Advantageously, said first model and said third model are neural network type models.

[0021] Advantageously, said second model is a type model, from English, "encoder / decoder with head classification".

[0022] Thus, the model is particularly suited to input data of different categories (data, vehicle, first list of words, diagnostic data ...).

[0023] Advantageously, the first, second and third models are, in addition, trained from text extractions from diagnostic guides.

[0024] Thus, the models are simpler to predetermine and are optimized for the problem. The information is also more precise.

[0025] A second aspect of the invention relates to a device comprising an audio speech recognition and synthesis system, a memory associated with at least one processor configured to implement the method according to the first aspect of the invention.

[0026] The invention also relates to a vehicle comprising the device.

[0027] The invention also relates to a computer program comprising instructions which, when the program is executed by the device according to the second aspect of the invention, lead the latter to implement the process according to the first aspect of the invention. Brief description of the figures

[0028] Other features and advantages of the invention will become apparent from the description of the following non-limiting embodiments of the invention, with reference to the accompanying figures, in which:

[0029] [Fig.1] schematically illustrates a device, according to a particular embodiment of the present invention.

[0030] [Fig.2] schematically illustrates a method of providing information on a malfunction of a vehicle, according to a particular embodiment of the present invention. Detailed description of the invention

[0031] The invention is described below in its non-limiting application to the case of an autonomous motor vehicle traveling on a road or traffic lane. Other applications such as a robot in a warehouse or a motorcycle on a country road are also conceivable.

[0032] Figure 1 shows an example of a device 101 included in the vehicle, in a network (“cloud”), or in a server. This device 101 can be used as a centralized device responsible for at least some steps of the process described below with reference to Figure 2. In one embodiment, it corresponds to an autonomous driving computer.

[0033] In the present invention, the device 101 is included in the vehicle.

[0034] This device 101 can take the form of a housing comprising circuits printed materials, from any type of computer or even a mobile phone (“smartphone”).

[0035] The device 101 includes a random access memory 102 for storing instructions for the implementation by a processor 103 of at least one step of the process as described below. The device also includes a mass storage 104 for storing data intended to be retained after the implementation of the process.

[0036] Device 101 may further include a digital signal processor (DSP) 105. This DSP 105 receives data to shape, demodulate and amplify, in a way known in itself, this data.

[0037] The device 101 also includes an input interface 106 for receiving the data implemented by the method according to the invention and an output interface 107 for transmitting the data implemented by the method according to the invention.

[0038] For example, the input interface 106 can receive the following data: vehicle position or geographical location, vehicle speed and / or acceleration, setpoint or predetermined positions / speeds / accelerations, engine speed, position and / or travel of the clutch, brake, and / or accelerator pedals, detection of other vehicles or objects, position or geographical location of other detected vehicles or objects, speed and / or acceleration of other detected vehicles or objects, sensor operating states, confidence level of data from or processed by sensors and / or devices similar to device 101. For example, sensors capable of providing data include: GPS with or without mapping, tachometers, accelerometers, RADAR, LIDAR, lasers, ultrasound, cameras...

[0039] Interface 106 can also receive: text or an alphanumeric string of characters, vehicle data such as a make, model, an identifier allowing the determination of a make and / or model of the vehicle, data relating to a state of said vehicle such as speed, temperature, or at least one piece of data from the preceding paragraph, data relating to diagnostic data of said vehicle such as a fault code, a list of fault codes, ..., data relating to an occupant, ...

[0040] For example, the output interface 107 can transmit data similar to the data received by the input interface 106. For example, the output interface 107 can also transmit: a list of labels, a first list of words, a second list of words, a list of sentences, metadata relating to a predetermined set of received data, ...

[0041] [Fig-2] schematically illustrates a method for providing information on a malfunction of a vehicle, according to a particular embodiment of the present invention.

[0042] The method 200 is a method for providing information on a malfunction of a vehicle equipped with an audio system for speech recognition and synthesis, to inform an occupant of said vehicle, said occupant issuing a verbal request, said method being implemented by the processor 103 and comprising several steps.

[0043] Audio speech recognition and synthesis systems are known. They are capable of converting a verbal request from an occupant into text. By text, we mean a sequence of characters, for example alphanumeric characters, interpretable by computer. ticically. For example, these characters are coded, encoded, in ASCII, UTF-8, UTF-18 or other encodings known to the Man Art of character encoding.

[0044] The occupant can make various verbal requests. For example, a verbal request could be the verbal expression uttered by the occupant, "What does the flashing red light on the dashboard mean?", when the occupant wishes to understand the meaning of an alert.

[0045] Step 210, RxTxt, is a text reception step generated from said audio system processing said verbal request. Following the above verbal expression, said audio system is capable of transmitting to device 101 a text such as "what does the flashing red light on the dashboard mean?". In one operating mode, the audio system is capable of transmitting metadata, such as metadata relating to a "Question" through intonation analysis.

[0046] Step 220, LList, is a step for determining a list of labels from the received text and a predetermined first model. This first model is a machine learning model from the field of natural language processing (NLP), such as neural networks, language models, classification and clustering algorithms, statistical methods, etc. In one operating mode, this first model is a neural network model whose design method (number of nodes, number of layers, choice of activation functions, etc.) and training method (parameterization of activation functions, etc.) are well known; the remaining difficulty is to choose relevant inputs and outputs for the model in relation to the problem to be solved.

[0047] Said first model is trained from a predetermined set of texts representative of a verbal request for information on a malfunction of at least one vehicle, and from a predetermined set of said labels. Said labels are a group of characters, at least one word, used to identify and describe sets of words from said predetermined set of texts representative of a verbal request for information on a malfunction of at least one vehicle.

[0048] For example, a predetermined set of texts representing a verbal request for information on a vehicle malfunction may consist of several subsets. To illustrate, a subset may be linked to an expression introducing a question, such as the set {"that", "how", "why", "means", "signify", ...}. In another example, a subset may be linked to a vehicle part or device, such as the set {"engine", "dashboard", "warning light", "turn signal", "red", ...}.

[0049] For example, to illustrate, a label could be the expression "question" to identify an expression introducing a question. Another label could be The expression "signal" to identify an element "flashing", "light", ... Thus, words or expressions having similar meanings or which appear in similar contexts will have the same label.

[0050] Advantageously, the construction of the labels is carried out from Verbatim of customers reporting an anomaly and / or from text extractions from diagnostic guides.

[0051] Since the first model is predetermined, machine learning is performed offline, before implementation in RAM 102 or mass storage 104 of the device 101. The set of labels and label lists is available. All or part of the set can be used for training the models described below or for constructing the knowledge graph described below.

[0052] Step 230, RxDataV, is a step for receiving initial data, referred to as vehicle data, said initial data being related to said vehicle, said data comprising at least one identifier allowing the determination of a make and / or model of said vehicle, and comprising at least one piece of information on a state of said vehicle such as a speed, a temperature, ...

[0053] It is known that a vehicle is identified by its Vehicle Identification Number, or VIN, a standardized alphanumeric string of characters. The first characters identify a make or manufacturer, such as a Peugeot® vehicle, a Citroën® vehicle, etc., a car model, such as a 208®, C3®, etc., an engine type, the model's year of manufacture, etc. Additionally, the vehicle's electronic control units (ECUs), or devices similar to device 101, contain other data that allow the vehicle's make and model to be identified. These ECUs are capable of transmitting data relating to that vehicle to device 101.

[0054] It is known that the vehicle includes numerous devices that are controlled, activated, deactivated, and monitored by sensors and actuators. These devices are capable of providing information on the vehicle's state. A state corresponds, for example, to data received by the input interface 106, described previously.

[0055] Advantageously, said vehicle data also includes data relating to said occupant. For example, by means of driver or passenger recognition devices, the name and surname of the occupants and / or the driver, a driver's driving style, etc., can be received.

[0056] For a given set of vehicles, for example vehicles of the same make and model, it is possible to construct, offline, a set of initial data that can be received. This set can be used for training the models described below or for constructing the knowledge graph described below.

[0057] Step 240, RxDataD, is a second data reception step, referred to as diagnostic data, said second data relating to diagnostic data of said vehicle such as a fault code, a list of fault codes, ...

[0058] It is known that the vehicle has numerous devices that are controlled, activated, deactivated, and monitored by sensors and actuators. These devices are capable of providing information about their operation. Typically, if a malfunction is detected, a fault code, or DTC (Diagnostic Trouble Code), is generated. Generally, a control unit stores all the DTCs along with a predetermined list of vehicle statuses. Input interface 106 is capable of receiving these DTCs, fault code lists, and predetermined vehicle status lists.

[0059] For a given set of vehicles, for example vehicles of the same make and model, it is possible to construct, offline, a set of secondary data that can be received. This set can be used for training the models described below or for constructing the knowledge graph described below.

[0060] Step 250, WListl, is a step for determining a first word list from said label list, said first and second data received, and a predetermined knowledge graph, said graph comprising nodes, said nodes being representative of entities, the entities being predetermined from said labels, and a set of at least one word, said at least one word characterizing a predetermined set of the first and second data received. Said graph also comprises arcs, said arcs being representative of a functional relationship between said entities, said arcs being described by at least one word.

[0061] Knowledge graphs (KGs) are well-known. A KG is a data structure that organizes knowledge in the form of graphs by linking entities and relationships. The challenge lies in obtaining a list of entities. In the present invention, this list is determined from the set of possible labels of the first model and the set of said at least one word, said at least one word characterizing a predetermined set of the first and second data received. Said at least one word can be considered similarly to other labels or additional labels that may have the value, for example, "Vehicle", "Peugeot®", "Citroën®", "208®", "C3®", "engine temperature 200°C", "engine speed 18000 rpm", "flashing light #12 activated", ...

[0062] Once the list of entities is obtained, a person skilled in the art can construct the KG and determine the relationships between these entities. The KG is predetermined, so it is implemented "offline". Only the model is implemented in RAM 102 or mass memory 104. The model is interpretable by computer in a known way, allowing the capture of complex relationships between the entities and providing semantics to the data. Thus, the output of this KG model, based on the list of labels determined in step 220, and the data from steps 230 and 240, yields a list of words. For example, this list of words could be a set of possible explanations for the state of a warning light such as "engine overheating," "oil too hot," "insufficient engine oil," "insufficient fuel injection," "incomplete combustion," etc.

[0063] The use of a KG simplifies all models, particularly the second model described below. The KG identifies various possible explanations for an anomaly only vaguely described by the occupant. The KG acts as a filter on all possible explanations for all possible anomalies. The second model identifies and diagnoses the most probable explanation for a given vehicle operating condition.

[0064] For a given set of vehicles, for example vehicles of the same make and model, it is possible to construct, offline, a set of initial word lists. This set can be used for learning the models described below.

[0065] Step 260, WList2, is the step of determining a second word list from the first word list, said second data received and a second predetermined model, said second model being a machine learning model from the domain of natural language processing, said second model being trained from a predetermined set of first word lists, and a predetermined set of second data.

[0066] Said second model is a machine learning model from the field of natural language processing (NLP), such as neural networks, language models, classification and clustering algorithms, statistical methods, etc. Advantageously, said second model is an "encoder / decoder with head classification" type model. This allows it to take into account different types of input data, such as a word list and diagnostic data. The construction and parameterization of "encoder / decoder with head classification" type models are known.

[0067] In the case of the invention, we have found that the "encoder / decoder with head classification" model architecture makes it possible to isolate and identify the anomaly from a list of possible anomalies. The classification head is an additional layer added to the encoder / decoder architecture. The classification head allows the input data of the second model to be classified. It establishes relationships between a list of words and diagnostic data.

[0068] An encoder and a decoder can each be based on a network model of neurons, for example recursively, or on a transformer. The encoder outputs are the decoder inputs. This architecture allows us to understand and represent the input data, and to generate a list of words interpretable by the third model.

[0069] For example, at the output of the second model, we have the following second list of words: “Pierre”, “Citroën C3 car”, “meaning of red warning light”, “high engine temperature detection”, “low engine oil anomaly”, “check engine oil level”

[0070] For a given set of vehicles, for example vehicles of the same make and model, it is possible to construct, offline, a set of second word lists. This set can be used for learning the model described below.

[0071] Step 270, SList, is a step for determining a list of sentences from said second word list and a third predetermined model, said third model being a machine learning model from the domain of natural language processing, said third model being at least trained from a set of second word lists.

[0072] Said third model is a machine learning model from the field of natural language processing, NLP, such as neural networks, language models, classification and clustering algorithms, statistical methods, etc. In an operational mode, said third model is a neural network type model whose design method (number of nodes, number of layers, choice of activation functions, ...) and training method (parameterization of activation functions, ...) are well known; the remaining difficulty is to choose relevant inputs and outputs of the model in relation to the problem to be solved.

[0073] Said third model being trained at least on a set of second word lists. In order to train said third model, it is possible to extract all or part of the set of second word lists from the second data model. For example, this extraction can be performed by simulation by feeding said second model with a set of first word lists and a set of received data, including the second data.

[0074] Advantageously, said phrase list is further determined from said vehicle and diagnostic data, said second model being further trained from metadata relating to a predetermined set of vehicle and diagnostic data. Thus, the phrase list is contextualized according to the make and / or model of the vehicle. Said third model is even more optimized. For example, metadata is "Citroën C3®". The third model is not trained This only covers known malfunctions for a Citroën C3® vehicle. It does not contain information on any other manufacturer or vehicle model. The third implemented model will only include information related to that specific vehicle.

[0075] Advantageously, the third model is further trained using text extracts from diagnostic guides. Thus, the third model can generate a more relevant and complete phrase list with respect to a detected anomaly. This simplifies the training of the third model, as the phrase list is adapted to respond to a request for information on a vehicle malfunction. If the first and second models are also trained using text extracts from diagnostic guides, the determination of the phrase list is more robust to variations in verbal queries for the same operational request.

[0076] In an operating procedure, if the said received data also includes data relating to said occupant, the list of phrases can be contextualized and personalized with respect to the occupant.

[0077] For example, at the output of the third model, we have the following list of phrases: "Dear Pierre.", "for your Citroën C3", "Here is the meaning of the flashing red light.", "We have detected an abnormally high engine temperature.", "It seems that the engine oil level is insufficient.", "Please check the engine oil level with the engine off.", "If oil needs to be added, we remind you that the recommended oil type is 15W40."

[0078] There are NLGs, from the English "Natural Language Generation", sub-models of NLP specifically designed to generate sentences from a list of instructions. Advantageously, said third model is obtained from an existing NLG after specialization, additional training and simplifications, from a set of initial word lists and, if necessary, from text extractions from diagnostic guides and / or from metadata relating to a predetermined set of vehicle and diagnostic data.

[0079] Step 280, TxSList, is a step in transmitting said list of phrases to said audio system to provide information about said malfunction of said vehicle by means of speech synthesis. The audio system is then capable, through its speech synthesis system, of transmitting the operating information to assist and guide the occupant.

[0080] The present invention is not limited to the embodiments described above by way of example: it extends to other variants.

[0081] For example, the process has been described as a sequence of steps. Some steps can be carried out in parallel or in a different sequence.

[0082] Thus, an example of an embodiment has been described above in which the following is involved: Words, phrases, groups of words... A phrase can also be a group of words. Advantageously, words, phrases, and groups of words can be encoded to be easily interpreted by a computer system or a 101 device.

[0083] Advantageously, the training can be carried out on a particular make or model of vehicle. The device 101 will then include in its RAM 102 and mass storage 104 only models adapted and optimized for the vehicle containing the device 101.

Claims

Claims

1. Method (200) for providing information on an operating anomaly of a vehicle having an audio voice recognition and synthesis system, to inform an occupant of said vehicle, said occupant issuing a verbal request, said method being implemented by a processor (103) and comprising the steps of: • Receiving (210) a text generated from said audio system processing said verbal request; • Determination (220) of a list of labels from said received text and a first predetermined model, • said first model being a machine learning model from the natural language processing domain, said first model being trained from • a predetermined set of texts representative of a verbal request for the provision of information on an operating anomaly of at least one vehicle, and of a predetermined set of said labels; said labels being a group of characters, at least one word, used to identify and describe sets of words from said predetermined set of texts representative of a verbal request for the provision of information on an operating anomaly of at least one vehicle; Reception (230) of first data, called vehicle data, said first data relating to said vehicle, said data comprising at least one identifier making it possible to determine a make and / or a model of said vehicle, and comprising at least one item of information on a state of said vehicle such as a speed, a temperature, etc.; Reception (240) of second data, called diagnostic data, said second data relating to diagnostic data of said vehicle such as a fault code, a fault code list, etc.; Determination (250) of a first list of words from said list of labels, said first and second received data and a predetermined knowledge graph, • said graph comprising • nodes, said nodes being representative of entities, the entities being predetermined from • said labels, and • a set of at least one word, said at least one word characterizing a predetermined set of the first and second data received; • arcs, said arcs being representative of a functional relationship between said entities, said arcs being described by at least one word; Determination (260) of a second list of words from the first list of words, said second received data and a second predetermined model, • said second model being a machine learning model from the natural language processing domain, • said second model being trained from • a predetermined set of initial word lists, and • a predetermined set of second data; Determining (270) a list of sentences from said second list of words and a third predetermined model, said third model being a machine learning model from the natural language processing domain, • said third model being at least trained from a set of second lists of words; Transmission (280) of said list of phrases to said audio system to provide information on said operating anomaly. operation of said vehicle by voice synthesis.

2. The method of claim 1, wherein said list of sentences is further determined from said vehicle data and diagnostic data, said second model being further trained from metadata relating to a predetermined set of vehicle data and diagnostic data.

3. Method according to the preceding claim, said vehicle data further comprises data relating to said occupant.

4. Method according to one of the preceding claims, wherein said first model and said third model are neural network type models.

5. Method according to one of the preceding claims, in which said second model is a model of the type, from the English, “encoder / decoder with head classification”.

6. Method according to one of the preceding claims, in which the first, second and third models are, in addition, trained from text extractions of diagnostic guides.

7. Device (101) comprising an audio system for voice recognition and synthesis, a memory (102) associated with at least one processor (103) configured to implement the method according to one of the preceding claims.

8.

9. Vehicle comprising the device according to the preceding claim. Computer program comprising instructions which, when the program is executed by the device (101) according to claim 8, cause the latter to implement the method according to one of claims 1 to 7.

Images