Method for processing aeronautical information intended for aircrew

Abstract

The invention relates to a method (100) for processing aeronautical information intended for aircrew, implemented by computer, and designed to populate a digital database searchable by a user, said aeronautical information comprising textual messages represented by digital data, said method comprising a step (105) of receiving the messages, and for each message: a step (140) of determining named entities contained in the message and the relationships between said named entities, a step (150) of integrating said named entities and their relationships into a database in the form of knowledge graphs. Figure 2

Description

Title of the invention: Method for processing aeronautical information intended for air navigators

[0001] The invention relates to the field of aeronautical information intended for air navigators, and, in particular, to a method for processing such information.

[0002] Aeronautical information is known as NOTAMs (for "Notice to Airmen" in English). A NOTAM is a text message, generally in English, published by a civil aviation authority for the attention of flight personnel. Typically, a NOTAM includes information relating to air transport infrastructure, such as airports for example; airspace; services offered, such as hotels, fuel supply; or regulations, customs. This information may provide information on their status, availability, usage restrictions or procedural changes. A NOTAM may also include information on natural phenomena, for example on weather conditions or volcanic activity. A NOTAM amends official publications such as charts and procedures. It is generally temporary, in that it describes a current state that is likely to change.However, there are permanent NOTAMs.

[0003] [Fig.l] illustrates an example of a NOTAM. Aviation organizations have defined a structure that each NOTAM must present. NOTAMs are written with aviation-specific abbreviations or acronyms, an official list of which is maintained by TCAO (International Civil Aviation Organization). The NOTAM forms a text message structured using capital letters, generally at the beginning of each paragraph, which indicate the type of information mentioned. A second level of structuring is defined by the symbol " / " which acts as an information separator. Thus, the first line of the message represents the following information: NOTAM serial number / year. The first line ends with the indication that it is a NOTAM.

[0004] The following letters each indicate a type of information: i. Q): qualification line, which provides contextual elements; in particular the element located between the first and second “ / ” symbols provides the “Qcode” which provides information on the general subject of the message following predefined categories; ii. A): the place; iii. B): the start date of activity, i.e. the start date of application of the NOTAM; iv. C): the end date of activity, i.e. the end date of application of the NOTAM; v. D): the activity slot, i.e. the hours of application of the NOTAM; vi. E): the information which is the main object of the NOTAM; it is a text freely written by the editor; vii. F) lower limit of activity, i.e. the minimum altitude to which the NOTAM applies; viii. G) upper limit of activity, i.e. the maximum altitude to which the NOTAM applies.

[0005] These NOTAMs are intended in particular for pilots who must consult them to prepare their flight. To be relevant, such a review of the NOTAMs must be done in a period immediately preceding the flight, generally within two hours. This review is added to other activities of the pilot necessary before his flight, it is difficult for him to make an exhaustive analysis of the NOTAMS. Especially since a pilot may have to consult dozens of them.

[0006] Furthermore, it is difficult for the pilot to classify and prioritize them by criticality for his flight in an efficient manner. This is because the official abbreviations and acronyms must be decoded, and the writer of the NOTAM uses a telegraphic style with additional abbreviations and omissions of words. The Q-code, which is supposed to provide information on the main subject of the NOTAM, can take a large number of values, around 1200. This field is also not always well filled in by the person writing the NOTAM.

[0007] A technical solution to facilitate the extraction of relevant information from NOTAMs is therefore particularly advantageous for users. In particular, we know the publication “Interpretation and contextual visualization of NOTAMs (messages to air navigators)”, A. Arnold, G. Dupont, C. Kobus, F. Lancelot, P. Narayan, 26th Conference on Natural Language Processing, Toulouse, France, ATALA, p.639-643, 2019. This document describes a method for processing a NOTAM to visualize the information contained therein on an airport map. However, there is a lack of a NOTAM processing method that is generic and allows a database to be obtained that can be searched by a user.

[0008] Other aeronautical information is known as supplements to aeronautical information publications, known as SUP AIPs (for the English "supplement to aeronautical information publications"). Such aeronautical information amends aeronautical information publications or AIPs (for the English "aeronautical information publications"). Typically, like NOTAMS, SUP AIPs are of a temporary nature.

[0009] A method is therefore sought for processing aeronautical information, such as NOTAMs or SUP AIPs, making it possible to extract the information and make it available to a user.

[0010] To this end, the invention proposes a method for processing aeronautical information intended for air navigators, implemented by computer, and intended to supply a digital database that can be queried by a user, said aeronautical information comprising text messages represented by digital data, said method comprising a step of receiving the messages, and for each message: a step of determining named entities contained in the message and the relationships between these named entities, a step of integrating said named entities and their relationships into a database in the form of knowledge graphs.

[0011] The representation in the form of knowledge graphs facilitates the analysis of information extracted from aeronautical information messages and their ability to be found and prioritized by a user.

[0012] According to one embodiment, the method comprises, after the reception step, a step of translating the messages into natural language.

[0013] According to a variant, said translation step comprises, in a message: the replacement of abbreviations with their meaning, or the addition of missing verbs, or the addition of missing conjunctions, or replacing regular expressions with their meaning.

[0014] According to a variant, the method comprises, before the translation step, a classification step which includes: - classification of the message according to a database of message types, and - the determination of the subjects covered in said message from a database keywords, said translation step being a function of the classification of the message and the subjects covered in said message.

[0015] According to one embodiment, the method comprises, after the translation step: a segmentation step involving the segmentation of the message into a list of words for comparison with a vocabulary, a grammatical analysis step involving the attribution to each word of a label representative of its grammatical position, a step of extracting named entities and their relationships, a step of forming a knowledge graph from the named entities and their relationships.

[0016] According to one embodiment, the method comprises a step of detecting inconsistencies between the knowledge graphs of the database.

[0017] According to one embodiment, the method comprises a merging of knowledge graphs which are at least partially identical.

[0018] According to one embodiment, the method comprises, before said step of receiving messages: - a request for aeronautical information by an air navigator on an aeronautical database, in particular aeronautical information relating to one or more geographical areas; - sending said aeronautical information to a computer.

[0019] According to one embodiment, the text messages of the aeronautical information comprise messages to air navigators, called “NOTAMs” and / or supplements to the aeronautical information publications, called SUP AIPs.

[0020] Other characteristics and advantages of the present invention will appear better on reading the description which follows in relation to the following appended figures:

[0021] [Fig-1]: [Fig.l], already described, represents an example of NOTAM;

[0022] [Fig.2]: [Fig.2] represents the steps of an example of a method according to the invention;

[0023] [Fig.3]: [Fig.3] represents the preprocessing of a NOTAM in the example of process of [Fig.2];

[0024] [Fig.4]: [Fig.4] represents an example of a concept ontology usable in the process example illustrated in [Fig.2].

[0025] [Fig. 2] illustrates an example of a method 100 for processing aeronautical information according to the invention. The method is implemented by a computer to feed a digital database that can be searched by a user, in particular a pilot. The aeronautical information comprises text messages represented by digital data.

[0026] In particular, an air navigator such as a pilot issues a request for aeronautical information on an aeronautical database. In particular, such a request aims to obtain NOTAMs or SUP AIPs in relation to one or more geographical areas concerned by a flight plan of the air navigator. In a step 105, the aeronautical information is then received on a computer.

[0027] Aeronautical information, in particular NOTAMs and / or SUP AIPS are processed together to be integrated into a database in the form of knowledge graphs. In the following, an example of processing aeronautical information of the NOTAM type is explained. However, the explanations apply similarly to other types of aeronautical information, such as SUP AIPs. In the context of the present application, NOTAM designates the received text message, but also its various intermediate states obtained during the implementation of the example of method according to the invention.

[0028] Primarily, for each NOTAM received, the named entities are determined, as well as the relationships between them. These entities and relationships between them are formalized in the form of knowledge graphs in a database. The NOTAMs processed by the method 100 are all integrated into the same database. For example, all the NOTAMs to be reviewed by a pilot are integrated into the same database in the form of knowledge graphs. The method comprises automatic processing of several NOTAMs to obtain a semantic representation of the world that can be used for the pilot's mission. The representation in the form of a knowledge graph makes it possible to sort and prioritize the knowledge contained in the NOTAMs for use by the pilot. This sorting and prioritization can be updated during the pilot's mission as a function of changes in the mission parameters in order to produce recommendations to the pilot.

[0029] In order to facilitate the formalization into a knowledge graph, the method may comprise an initial preprocessing step 110 of the NOTAM, which consists in particular of a translation or conversion of the raw NOTAM NI into natural, syntactically correct language. By the expression “natural language” is meant a normal language commonly used by a human being, as opposed to a formal language, such as a computer language.

[0030] The preprocessing step 110 is illustrated in [Fig.3]. The raw NOTAM NI is received as input to the method and is in particular as issued by an air authority. The example of raw NOTAM NI illustrated includes the following text:

[0031] “RWY 16 / 34 CLSD AVBL TWY”

[0032] Here, only the text following the letter E) of the NOTAM is represented. During this preprocessing 110, in a natural language conversion step 114, the raw text of the NOTAM is modified so as to replace the abbreviations with their meaning, namely a word or expression designated by this abbreviation. Missing verbs or conjunctions are added. Regular expressions (sequence of known characters) can also be recognized and replaced with their meaning, namely a word or expression belonging to the natural language. Thus, at the end of this translation step, the NOTAM N2 includes the following text, the added words being underlined.

[0033] “the runway 16 / 34 is closed and / but available as taxiway”

[0034] Such a step 114 of conversion into natural language facilitates the subsequent steps of processing the NOTAM which will be described later in relation to [Fig.2]. There are several tools for translation into natural language. One can use, among others, deep learning models, such as for example the so-called “transformer” models GPT2, GPT3.

[0035] Prior to translation, the preprocessing step 110 may comprise a classification step 112 of the NOTAM.

[0036] In particular, the NOTAM is then classified according to a NOTAM type base so as to know the general subject of the NOTAM. Such a classification can be based on the content of the Q-code, in particular on a reconstruction of the Q-code using a natural language processing tool. Such a reconstruction makes it possible to compensate for poor writing by the issuer of the NOTAM. Knowing the general subject of the message can facilitate the subsequent step 114 of conversion into natural language. The NOTAM can also be classified according to its language level. For example, from zero for a text comprising mainly abbreviations, to five for a text already in natural language. The subjects covered in the NOTAM can also be determined from a keyword base.

[0037] At the end of this preprocessing step 110, the NOTAM N2 in natural language undergoes a series of steps, illustrated in [Fig.2], to be represented in the form of knowledge graphs. In particular, the NOTAM N2 undergoes a segmentation step 120 comprising the segmentation of the NOTAM into a list of words for comparison with a vocabulary. Then a grammatical analysis 130 is applied to the NOTAM N3, comprising the attribution to each word of a label representative of its grammatical position. Annotated sentences are then obtained. During a step 140, the named entities and their relationships are then extracted from the NOTAM N4 in the form of N-tuples N5. From which, in a step 150, the knowledge graph is then formed.

[0038] The step 140 of extracting entities and relationships and the step 150 of forming a knowledge graph can be aided by an ontology of concepts 170 specific to aeronautical information, for example NOTAMs and / or SUP AIPs.

[0039] [Fig.4] presents an example of a concept ontology, in which the attributes related to the mission concept are represented: the attributes include the following items: ATC (Air Traffic Control) authorization, aircraft fuel, aircraft status, aircraft, airport, operational runway.

[0040] Knowledge graph representation makes it possible to detect inconsistencies between information from aeronautical information, and provides the possibility of merging knowledge graphs that provide the same information. Such graph merging can be achieved by a maximal subgraph isomorphism search method.

Claims

Claims

1. Method (100) for processing aeronautical information intended for air navigators, implemented by computer, and intended to feed a digital database that can be queried by a user, said aeronautical information comprising text messages represented by digital data, said method comprising a step (105) of receiving the messages, a step (114) of translating the messages into natural language, and for each message: - a step (140) of determining named entities contained in the message and the relationships between these named entities, - a step (150) of integrating said named entities and their relationships into a database in the form of knowledge graphs.

2. Method (100) according to the preceding claim, wherein said translation step (114) comprises, in a message: - the replacement of abbreviations by their meaning, or - the addition of missing verbs, or - the addition of missing conjunctions, or - the replacement of regular expressions by their meaning.

3. Method (100) according to the preceding claim comprising, before the translation step (114), a classification step (112) which comprises: - the classification of the message according to a base of message types, and - the determination of the subjects dealt with in said message from a base of keywords, said translation step (114) being a function of the classification of the message and of the subjects dealt with in said message.

4. Method (100) according to one of the preceding claims comprising, after the translation step (114): - a segmentation step (120) comprising the segmentation of the message into a list of words for comparison with a vocabulary, - a grammatical analysis step (130) comprising the allocation to each word of a label representative of its grammatical position, - a step (140) of extraction of named entities and their relationships, - a step (150) of formation of a knowledge graph from the named entities and their relationships.

5. Method (100) according to one of the preceding claims, comprising a step of detecting inconsistencies between the knowledge graphs of the database.

6. Method (100) according to one of the preceding claims, comprising a merging of the knowledge graphs which are at least partially identical.

7. Method (100) according to one of the preceding claims, comprising, before said step of receiving (105) messages: - a request for aeronautical information by an air navigator on an aeronautical database, in particular aeronautical information relating to one or more geographical areas; - sending said aeronautical information to a computer.

8. Method (100) according to one of the preceding claims, in which the text messages of the aeronautical information comprise messages to air navigators, called "NOTAMs" and / or supplements to aeronautical information publications, called SUP AIPs.

Images