Decision support method for an aircraft pilot, electronic decision support system and associated computer program products

Abstract

Decision support method for an aircraft pilot, electronic decision support system and associated computer program products. The present invention relates to a decision support method for an aircraft pilot (12). The method comprises a generation phase, implemented prior to the flight of the aircraft (12), and comprising obtaining a plurality of responses from pilots in the context of a change of responsibility. The generation phase further comprises extracting data from the responses to form a set of taxa, and generating the structured database by classifying the taxa and determining a set of quantified importance indicator(s). The method further comprises an interaction phase, implemented during the flight, and comprising determining and transmitting to the pilot specific data from the structured database according to a request from the pilot. Figure for the abstract: Figure 1

Description

Title of the invention: Decision-making assistance method for an aircraft pilot, electronic decision-making assistance system, and associated set of computer program products

[0001] The present invention relates to a decision-making assistance method for a pilot in an aircraft.

[0002] The present invention also relates to an electronic decision-making assistance system and an associated set of computer program products.

[0003] In the field of aviation, pilots are regularly required to make decisions within a short period of time. In order to assist them in decision-making, the cockpits of aircraft traditionally include different electronic systems, such as flight computers, also called FMS (from the English Flight Management System) as well as Human-Machine Interfaces (HMI) with which pilots are specific to interact.

[0004] These systems are designed to transmit various information to the aircraft pilot, for example by continuously displaying the value of different parameters, such as altitude, outside pressure or the amount of fuel.

[0005] However, when the pilot is faced with an unusual situation or one that takes him out of his usual routine, he must first determine what information he needs before consulting the associated parameters and deducing the actions to be taken.

[0006] This analysis requires valuable time from the aircraft pilot before any action can be taken. In some cases, this lost time poses a risk to the smooth running of the flight, or even to passenger safety in the most extreme cases.

[0007] Furthermore, in aircraft with at least two pilots, it is common practice for pilots to take turns during a long flight (e.g., over 6 hours). Therefore, in such cases, during the pilot change, the previous pilot must communicate the current flight situation to the incoming pilot. This task is also time-consuming, as the previous pilot must provide the incoming pilot with a comprehensive overview of the situation, including information relevant to the continuation of the flight. Moreover, omissions or unnecessary information may occur during this process.

[0008] The present invention aims to save this precious time for the pilot(s), in particular to limit the risks of flight accidents.

[0009] To this end, the present invention relates to a decision support method for an aircraft pilot, the method comprising a phase of generating a basis of structured data implemented by an electronic generation device, and an interaction phase between the structured database and the pilot, implemented by an electronic interaction device integrated into the aircraft cockpit,

[0010] the generation phase being implemented prior to the aircraft flight and comprising the following steps: - obtaining a plurality of texts corresponding to pilots' responses in the context of a transfer of responsibility from one pilot to another during an aircraft flight simulation, - data extraction from the obtained texts to form a dataset, called taxa, each taxon comprising a word, or set of words, relating to a characteristic of the cockpit, - generation of the structured database by classifying the taxa in a predetermined database structure, and by determining a set of quantified importance indicator(s) for each taxon by comparing each taxon to the texts obtained,

[0011] the interaction phase being implemented during the flight of the aircraft and comprising the following steps: - receipt of a request from the pilot, - determination of the data in the structured database, relevant to the query, by comparing the query to the data in the structured database, transmission of the determined data to the pilot.

[0012] According to particular embodiments, the process comprises the following characteristics, taken individually or in any technically feasible combination: - The obtaining stage includes the following sub-steps: - acquisition of audio tracks including the speech of pilots performing tasks on a simulator and being asked a list of questions; - conversion of each audio track into a respective text file; - the list of questions includes a first group of questions relating to a pilot taking over the cockpit, and a second group of questions relating to the transfer of the cockpit to another pilot; - the set of quantified importance indicator(s) includes at least one relevance score, one value of which is a function of the frequency of occurrence of said taxon in the texts, the relevance score being greater the more the corresponding taxon is presented as important in the texts obtained during the acquisition stage; - The set of quantified importance indicator(s) for each taxon includes: - the frequency of occurrence of the taxon in a portion of the texts obtained, - a TF-IDF score of said taxon obtained from a vectorization of the taxon, - the taxon relevance score, the taxon relevance score being calculated from the answers to questions relating to the transmission of the cockpit; - The second group of questions includes: - a first question asking the pilot what information he would have liked to transmit, - a second question asking the pilot what information he would not have liked to transmit,

[0013] the relevance score of each taxon depending positively on a frequency of occurrence of said taxon in the answer to the first question, and negatively on a frequency of occurrence of said taxon in the answer to the second question; - The term extraction step from the texts includes the following sub-steps: - division of each text into a plurality of word sets that are smaller than the original text, - parsing word sets into tokens comprising an even smaller number of words than the word sets, - for each token, reduction of said token to a lemma comprising a semantic root of the word(s) of the token, - selection, among the formed lemmas, according to the frequency of occurrence of this lemma among the texts from different pilots.

[0014] the selected lemmas forming the data set, called taxa; - The list of questions includes a question asking the pilot to pass on information they deem important for the handover to the other pilot.

[0015] during the selection substep of the extraction step, the taxon lemmas are further selected only from among the lemmas from the answer texts to said question; - the generation phase also includes a filtering step, during which taxa whose set of quantified importance indicator(s) does not meet a respective criterion are removed from the structured database.

[0016] The invention also relates to a set of computer program products comprising software instructions which, when executed by The computers implement a decision support process as described previously.

[0017] The invention also relates to an electronic system for piloting an aircraft, the system comprising a device for generating a structured database, and a device for interaction between the structured database and the pilot, the electronic interaction device being integrated into an aircraft cockpit,

[0018] the generation device being configured for, prior to the aircraft flight: - to obtain a plurality of texts corresponding to pilots' responses in the context of a transfer of responsibility from one pilot to another during an aircraft flight simulation, - extract data from the obtained texts to form a dataset, called taxa, each taxon comprising a word, or set of words, relating to a characteristic of the cockpit, - generate the structured database by classifying the taxa in a predetermined database structure, and by determining a set of quantified importance indicator(s) for each taxon by comparing each taxon to the texts obtained,

[0019] the interaction device being configured so that, during the flight of the aircraft: - to receive a request from the pilot, - determine the relevant data from the structured database relative to the query, by comparing the query to the data in the structured database, - transmit the determined data to the pilot.

[0020] The invention will become clearer upon reading the following description, given solely by way of non-limiting example, and made with reference to the drawings in which:

[0021] [Fig-1] [Fig.1] is a schematic representation of an aid system piloting according to the invention;

[0022] [Fig.2] [Fig.2] is a flowchart of a decision-making aid method according to the invention; and

[0023] [Fig.3] [Fig.3] is a graph representing an example of a lemma distribution among the pilot's speeches.

[0024] Fig. 1 illustrates an electronic decision support system 10 for a pilot of an aircraft 12.

[0025] The electronic system 10 includes an electronic device 14 for generating a structured database storing data said to be relevant to the pilot of the aircraft 12, and an electronic device 16 for interaction between the pilot of the aircraft 12 and the structured database.

[0026] The generation device 14 includes, for example, a first computing unit, such as a computer 18. In addition, the generation device 14 preferably includes a display screen 20 and a microphone 22.

[0027] The configurations of the generation device 14 will be detailed later with reference to a decision support method for an aircraft pilot, partly implemented by the generation device 14.

[0028] The interaction device 16 is preferably included in the aircraft 12. By way of example, the interaction device 16 includes a second computer 24, preferably coupled to a display screen 26 and an acquisition means 28 such as a keyboard, a touch surface or a microphone.

[0029] For example, the interaction device 16 is connected to an aircraft FMS, or included within that FMS.

[0030] Computers 18, 24 are suitable for implementing a decision support process for an aircraft pilot which will be described later.

[0031] The calculators 18, 24 are respectively electronic circuits designed to manipulate and / or transform data represented by electronic or physical quantities in registers of the calculator and / or memories into other similar data corresponding to physical data in register memories or other types of display devices, transmission devices or storage devices.

[0032] As specific examples, the calculators 18, 24 are each implemented as a programmable logic component, such as an FPGA (Field Programmable Gate Array), or as an integrated circuit, such as an ASIC (Application-Specific Integrated Circuit).

[0033] Alternatively, when the process is carried out in the form of one or more software programs, i.e. in the form of a computer program, also called a computer program product, it is further able to be recorded on a set of computer program products, not shown, readable by computer.

[0034] Computer program products are, for example, respectively suitable for storage on a medium capable of storing electronic instructions and for being coupled to a bus of a computer system. By way of example, said medium is an optical disc, a magneto-optical disc, a ROM memory, a RAM memory, any type of non-volatile memory (for example FLASH or NVRAM) or a magnetic card.

[0035] The decision support method 1000 for a pilot of aircraft 12 will now be described with reference to [Fig.2].

[0036] The method 1000 includes a phase 1100 of generating a structured database and an interaction phase 1500 between the pilot and the structured database.

[0037] Preferably, initially, during the acquisition phase 1110, a plurality of pilots are performing tasks on a flight simulator.

[0038] The generation phase 1100 includes a step of obtaining 1110 a plurality of texts corresponding to pilot responses in the context of a transfer of responsibility, from one pilot to another, during a flight simulation of the aircraft.

[0039] Preferably, the acquisition step 1110 comprises, for each of said pilots, a sub-acquisition step 1111 of audio tracks corresponding to the speech of pilots performing tasks on a simulator and to whom a list of questions is posed.

[0040] For this purpose, the computer 18 communicates, for example, the list of questions via the display screen 20 to each pilot.

[0041] Preferably, the list of questions includes a first group of questions relating to the taking of the cockpit by a pilot, and a second group of questions relating to the transfer of the cockpit to another pilot.

[0042] According to an example, the first group of questions includes the following questions: - El: What information was available to you, via the simulator, that was useful to you in carrying out your task? - E2: What additional information, via the simulator, would you have liked to have at your disposal to complete your task? - E3: What information available to you via the simulator seemed useless?

[0043] Preferably, questions E2 and E3 are asked when the task is handed over to another pilot.

[0044] As an optional supplement or alternative in this example, the second group of questions includes the following questions: - S0: What information do you transmit to a pilot who must continue to carry out your task? - IF: What additional information would you have liked to communicate during your handover? - S2: What information could you have refrained from transmitting during the handover?

[0045] Preferably, question S0 is asked just before the previous pilot makes his handover of the task to the next pilot, questions SI and S2 being preferably asked a posteriori, that is to say a few minutes or hours after this handover.

[0046] These questions allow us to evaluate the information that is most relevant and useful jointly to both pilots during the handover.

[0047] During the acquisition substep 1111, the pilot responses are for example acquired by the generation device 14 via the microphone 22.

[0048] Metadata is preferentially added to each audio track. The metadata includes the reference of the question corresponding to that audio track, i.e. E1, E2, E3, S0, SI, S2. This is referred to as labeling.

[0049] The obtaining step 1110 preferably includes a substep 1112 of converting each audio track into a respective text file.

[0050] For this purpose, the calculator 18 preferably applies an autonomous transcription model to each audio track, such as the "Whisper-timestamped" model as described in the article Computing and visualizing dynamic time warping alignments in R: The dtw package by Giorgino T. This model presents a very good transcription quality and is able to provide, for each transcribed word of the text file, a confidence quantifier in the transcription.

[0051] Optionally, during the conversion substep 1112, the generation device 14 displays each text file and receives validation and / or correction instructions from a third-party operator. The generation device 18 then performs the required corrections, if necessary.

[0052] The generation phase 1100 then includes a step 1120 of data extraction, called relevant data, from the text files.

[0053] To this end, the extraction step 1120 preferably includes a splitting substep 1121, in which each text file, corresponding to the answers to question S0, is split into smaller word sets, such as sentences or groups of sentences forming a paragraph. This substep facilitates subsequent processing while preserving the overall semantics of the discourse.

[0054] Optionally, the extraction step 1120 includes a sorting substep 1122 during which the calculator 18 removes, in each set of words, stop words, also called "function words" or "stop-words", such as definite or indefinite pronouns, adverbs or linking words.

[0055] The extraction step 1120 then preferably includes a substep 1123 of parsing the resulting word sets into tokens, also called cyber-tokens. This substep is also known by the anglicism "tokenization".

[0056] During the parsing substep 1123, the word sets are divided into smaller word groups. For example, the calculator 18 applies the N-Gram method, which consists of segmenting the word sets into sequences of consecutive words.

[0057] For example, with a value N equal to 2, i.e. the bi-Gram method, the following set of words: "right engine temperature above threshold" is converted into the following sequence of tokens: "engine temperature", "right engine", "upper right", "upper threshold".

[0058] This technique, as is known, uses, for example, UPOS (Universal Parts of Speech), which are labels used to categorize words in a text according to their grammatical role. These labels form a standard in the field of natural language processing.

[0059] The extraction step 1120 further preferably includes a substep 1124 of reducing each token into a lemma. Each lemma comprises a semantic root of the word or words of the associated token. This substep 1124 is also known by the anglicisms "stemming and m-matization".

[0060] To this end, the calculator 18 removes the prefix and suffix from each word of each token (stemming) and uses a dictionary to determine a root of each word deprived of its prefix / suffix(es) (lemmatization).

[0061] According to one example, the token “re-action shutter” becomes “action shutter”.

[0062] It is then understood that, among the set of determined lemmas, some are present several times. This is in particular linked to the reduction substep 1124 which makes it possible to form identical lemmas from tokens having similar semantics.

[0063] The extraction step 1120 preferably includes a vectorization substep 1125. During this substep 1125, the computer 18 converts each lemma into a vector with a numerical value. For example, the computer uses Bag of Words and TF-IDF (Terni Frequency-Invert Document Frequency) techniques.

[0064] Preferably, the extraction step 1120 further includes a substep 1126 for selecting a reduced number of lemmas, during which the calculator 18 determines, for each lemma or at least a portion of the lemmas, a number of occurrences among the speeches of the different drivers from the audio tracks, based on the metadata. In other words, for each portion of the lemmas, the calculator 18 determines the number of drivers for whom an audio track provided a text whose substeps 1121 to 1125 led to the formation of an identical lemma.

[0065] Preferably, the lemmas considered in the selection substep 1126 are only the lemmas from the audio tracks corresponding to the answers to question S0.

[0066] Lemmas from the speeches of several pilots in response to question S0 are hereinafter referred to as "common lemmas".

[0067] Fig. 3 represents an example of the number of common lemmas as a function of the number of drivers they are common to.

[0068] On [Fig.3], it is visible that the majority of lemmas come from speeches given by a single pilot and decrease with the number of pilots.

[0069] Then, during the selection substep 1126, the calculator 18 determines a breakpoint aimed at reducing the considered lemmas to only the most relevant ones, that is, those from the discourse of a majority of pilots, while maintaining sufficiently high representativeness. According to this technique, the lemmas that satisfy this dual constraint of relevance and representativeness are the lemmas from the discourse of a number of pilots greater than this breakpoint. The breakpoint is determined, for example, using the Changepoint library of the R software.

[0070] In the example of [Fig.3], the breaking point is determined for example at the value 2.5. Also, in this example, the lemmas selected during the selection substep 1126 are the lemmas from the discourse of at least 3 pilots, and preferably from the answer to question S0, that is to say from the handover of tasks between the two pilots.

[0071] These selected lemmas correspond to the data selected by the extraction step and are hereinafter referred to as: taxa.

[0072] The generation phase 1100 then includes a step 1130 of generating the structured database from the data extracted during the extraction step 1120, i.e. preferentially from taxa.

[0073] To this end, generation step 1130 includes a substep 1131 of classifying taxa into a predetermined taxonomy structure.

[0074] A taxonomy is a hierarchical classification of different entities of interest (e.g. of a company, organization or administration), used to classify documents, digital assets and other information.

[0075] The taxonomy structure is for example derived from business expertise and integrated into the calculator 18 prior to the implementation of process 1000.

[0076] The taxonomy structure includes, for example, the following classes: What, Why, How and ActRel which characterizes relief activities.

[0077] The classification of taxa in the taxonomic structure is carried out, for example, in a manner known per se, using the Protégé software as described in the article The Protégé project: A look back and a look forward by Musen, MA (2015).

[0078] By way of example, each taxon is classified into different classes based on the similarity between the taxon and the classes. For this purpose, a natural language processing tool, also called an NLP (Natural Language Programming) tool, is used to assign each taxon to a respective class. Such a tool is preferably trained beforehand on classification examples already performed and related to the domain concerned.

[0079] It is then understood that the taxonomy data are the taxa from extraction step 1120.

[0080] Generation step 1130 preferably further includes a substep 1132 of determination, for each taxon, of indicators of importance in the taxonomy.

[0081] For example, the importance indicators include an SP relevance score and more preferably also: an FSO occurrence frequency of taxa in response to question S0, and a TF-IDF score.

[0082] To this end, the calculator 18 determines, for example, for each respective taxon, the FSO occurrence frequency using, for example, a large language model, or LLM model (from the English Large Language Model), such as the MistralAl model.

[0083] For this purpose, the model preferentially uses the numerically valued vectors from the vectorization substep 1125 to compare them with each other.

[0084] The model then generates, for each taxon, a counter that is incremented when the text in response to question S0 includes words with semantics comparable to those of the taxon. This counter is then normalized by the number of occurrences of each taxon in the responses to question S0, thus forming the FSO frequency of occurrence of the taxon in question.

[0085] As an optional extra, the calculator determines the TF-IDF score of each taxon.

[0086] The TF-IDF score is a quantifier of the frequency of occurrence of the taxon in the responses to question S0. In particular, the TF-IDF score is, for example, the product of the FSO frequency of occurrence of the taxon in the responses to question S0, and the logarithm of the quotient between the number of responses to question S0 collected and the number of responses to question S0 in which the taxon appears at least once.

[0087] According to a particular embodiment, the TF-IDF score of each taxon is determined during vectorization substep 1125, since this vectorization is performed using Bag of Words and TF-IDF techniques known per se. Indeed, these techniques allow, in addition to the production of vectors with numerical values ​​for each taxon, the provision, for each taxon, of a quantifier of the importance of each taxon relative to the others in the set of texts resulting from the responses to question S0.

[0088] To calculate the relevance score, according to an example, a query including the SI question, the S2 question, the pilots' answers to the SI and S2 questions, as well as contextual elements, is passed to the LLM model.

[0089] In particular, the request is formulated, for example, as follows: - "You are a conversation assistant between two pilots during a Changeover activity. The two pilots, respectively referred to as the incoming pilot and the outgoing pilot, exchange information during this handover. At the end of the exchange, the incoming pilot asks the outgoing pilot the following questions: - question S1 - question S2 - "To which the outgoing pilot replies:" - answer to the question IF, - answer to question S2, - "As a conversation support assistant, base your work solely on the Given the context, which of the taxa present in the taxonomy should be transmitted to the incoming pilot, and which should not, so that he can resume operations?

[0090] To perform such processing, the LLM model uses, for example, the numerically valued vectors from the vectorization substep 1125 to perform its processing.

[0091] The LLM model is then suitable for providing for each taxon, a frequency of occurrence FS1 in the answers to questions SI, and a frequency of occurrence FS2 in the answers to questions S2.

[0092] Calculator 18 then preferentially determines, for each taxon, the relevance score according to the following formula:

[0093] [Math.l] SP = FS0 + a(FS1-FS2)

[0094] where FS0 is the frequency of occurrence of the taxon in the response to question S0, and

[0095] a is the TF-IDF score of the taxon.

[0096] It is then understood that the taxa with the highest relevance score are those concerning the information that the greatest number of pilots would have liked to transmit during their assessment (question S1). Conversely, the taxa with the lowest relevance score are those concerning information that they could have refrained from transmitting during the assessment (question S2).

[0097] Optionally, the query also includes the answers to questions E1, E2 and E3.

[0098] The LLM model is then also able to provide, for each taxon, the frequency of occurrence FE1 in the responses to questions El, and the frequency of occurrence FE2 in the responses to questions E2.

[0099] The aforementioned formula for the SP relevance score would then, for example, be modified as follows:

[0100] [Math.2] SP = S0+a(FSl +FE1 + FE2 - FS2 - FE3)

[0101] According to this optional addition, it is understood that the taxa with the highest SP relevance score are the taxa concerning information that the greatest number of pilots also found useful or would also have liked to have (question E2), Conversely, the taxa with the lowest relevance score are those concerning information also absent from the answers to questions E1, E2, SI but concerning information that the pilots also judged to be useless (question E3).

[0102] Optionally, the generation phase 1100 further includes a filtering step 1140 of the structured database, during which the computer 18 removes from the structured database taxa with a relevance score SP below a first threshold or with a TF-IDF score below a second threshold.

[0103] Indeed, this step makes it possible to reduce the size of the structured database by removing, for example, data corresponding to information that the pilots considered superfluous during the handover, for example in response to question S2.

[0104] Also, this step makes it possible to further limit the size of the structured database to the most relevant data.

[0105] Following generation phase 1100, generation device 14 generated the structured database including the most relevant data for information transfer between two pilots. Such a structured database is therefore called a CROP, from the English Common Relevant Operating Picture.

[0106] It will be understood that the structured database is an ontology insofar as it presents a taxonomic structure, while being enriched by respective indicators of the data it contains. In particular, these respective indicators reflect semantic characteristics of the data, transcribing the semantics of the texts from which the structured database is generated. It is then also referred to as a knowledge base.

[0107] The generated structured database is then integrated into the interaction device 16, preferably before the flight of the aircraft 12.

[0108] During the flight of the aircraft, an aircraft pilot wishes to obtain information on the current situation he is facing. This is done, for example, during a pilot change on a long-haul flight, or when the pilot is faced with an unusual situation.

[0109] The process then includes an interaction phase 1200, implemented by the interaction device 16.

[0110] The interaction phase 1200 includes a step 1210 of receiving a request from the pilot. The request is preferably received via the acquisition means 28

[0111] The interaction phase 1200 then includes a step 1220 of determining the data in the structured database, relevant to the query, by comparing the query with the data in the structured database.

[0112] The interaction phase 1200 then includes a step 1230 of transmitting the determined data to the pilot, preferably via the display screen 26

[0113] Thus, at the end of the interaction phase, the pilot has access to the relevant data for the situation he is facing and can deduce the actions to take to successfully complete his mission. Furthermore, since only relevant data is transmitted to the pilot, he does not have to sift through the information provided to him to deduce the most useful data. This time saving is essential because it allows the pilot to act quickly and reduces the cognitive load that could have led him to make poor decisions regarding the actions to take.

Claims

Demands

1. A method (1000) for decision support for a pilot of an aircraft (12), the method (1000) comprising a phase (1100) of generating a structured database implemented by an electronic generation device (14), and an interaction phase (1200) between the structured database and the pilot, implemented by an electronic interaction device (16) integrated into a cockpit of the aircraft (12), the generation phase (1100) being implemented prior to the flight of the aircraft (12) and comprising the following steps: - obtaining (1110) a plurality of texts corresponding to pilots' responses in the context of a transfer of responsibility from one pilot to another during an aircraft flight simulation, - extraction (1120) of data from the obtained texts to form a dataset, called taxa, each taxon comprising a word, or a set of words, relating to a characteristic of the cockpit, - generation (1130) of the structured database by classifying the taxa in a predetermined database structure, and by determining a set of quantified importance indicator(s) for each taxon by comparing each taxon to the texts obtained, the interaction phase (1200) being implemented during the flight of the aircraft (12) and comprising the following steps: - receiving (1210) a request from the pilot, - determining (1220) the data in the structured database that are relevant to the request, by comparing the request to the data in the structured database, - transmission (1230), to the pilot, of the determined data.

2. A process (1000) according to claim 1, wherein the production step (1110) comprises the following substeps: - Acquisition (1111) of an audio track comprising the speech of pilots performing tasks on a simulator and being asked a list of questions. - Conversion (112) of each audio track into a respective text file.

3. Method (1000) according to the preceding claim, wherein the list of questions comprises a first group of questions (E1, E2, E3) relating to the taking of the cockpit by a pilot, and a second group (S0, SI, S2) of questions relating to the transfer of the cockpit to another pilot.

4. A method (1000) according to any one of the preceding claims, wherein the set of quantified importance indicator(s) includes at least one relevance score, the value of which is a function of the frequency of occurrence of said taxon in the texts, the relevance score being greater the more the corresponding taxon is presented as important in the texts obtained during the obtaining step (1110).

5. Method (1000) according to claims 3 and 4, wherein the set of quantified importance indicator(s) of each taxon comprises: - a frequency of occurrence of the taxon in a part of the texts obtained, - a TF-IDF score of said taxon resulting from a vectorization of the taxon, - the relevance score of the taxon, the relevance score of the taxon being calculated from the answers to the questions (S0, SI, S2) relating to the transmission from the cockpit.

6. A method (1000), according to the preceding claim, wherein the second group of questions (S0, SI, S2) comprises: - a first question (SI) asking the pilot what information he would have liked to transmit, - a second question (S2) asking the pilot what information he would not have liked to transmit, the relevance score of each taxon depends positively on a frequency of occurrence of said taxon in the answer to the first question (S1), and negatively on a frequency of occurrence of said taxon in the answer to the second question (S2).

7. A method (1000) according to any one of the claims, wherein the step of extracting terms from the texts comprises the following substeps: - dividing (1121) each text into a plurality of word sets of reduced size compared to the text, - parsing (1123) the word sets into tokens comprising respectively a number of words even smaller than the word sets, - for each token, reducing (1124) said token to a lemma comprising a semantic root of the word or words of the token, - selecting (1126), from among the lemmas formed, according to the frequency of occurrence of this lemma among the texts from different pilots, the selected lemmas forming the data set, called taxa.

8. Method (1000) according to any one of the preceding claims, wherein, the list of questions includes a question (S0) asking the pilot to transmit information that he or she considers important for the handover to the other pilot, during the selection substep of the extraction step, the taxon lemmas are further selected only from the lemmas from the answer texts to said question.

9. A method (1000) according to any one of the preceding claims, wherein the generation phase further comprises a filtering step, in which taxa whose set of quantified importance indicator(s) does not meet a respective criterion are removed from the structured database.

10. A set of computer program products comprising software instructions which, when executed by computers, implement a decision support process according to any one of the preceding claims.

11. A decision support system (10) for a pilot of an aircraft (12), the system (10) comprising a device (14) for generating a structured database, and an interaction device (16) between the structured database and the pilot, the electronic interaction device (16) being integrated into an aircraft cockpit (12), the generation device (14) being configured to, prior to the flight of the aircraft (12): - to obtain a plurality of texts corresponding to pilots' responses in the context of a transfer of responsibility from one pilot to another during an aircraft flight simulation, - extract data from the obtained texts to form a dataset, called taxa, each taxon comprising a word, or set of words, relating to a characteristic of the cockpit, - generate the structured database by classifying the taxa in a predetermined database structure, and by determining a set of quantified importance indicator(s) for each taxon by comparing each taxon to the texts obtained, the interaction device (16) being configured for, during the flight of the aircraft (12): - to receive a request from the pilot, - determine the data for the structured database, relevant to the query, by comparing the query to the data in the structured database, - transmit the determined data to the pilot.

Images