WO2022129760A2

WO2022129760A2 - Method for collecting data, method for using collected data, corresponding electronic device and computer program products and medium

Info

Publication number: WO2022129760A2
Application number: PCT/FR2021/052298
Authority: WO
Inventors: Oumaima ALAOUI ISMAILI; Lucie AIMAR; Jean-Philippe LUCAS; Philippe Legay
Original assignee: Orange
Priority date: 2020-12-14
Filing date: 2021-12-13
Publication date: 2022-06-23
Also published as: FR3117638A1; WO2022129760A3

Abstract

The invention relates to a method for using data structures comprising a text feature, the method comprising automatically selecting a first text feature taking into account an association between the first text feature and a research subject area; and extracting data from a first structure, taking into account said text feature of said first structure and said selected first text feature. The invention relates to a method, implemented in a communication network, comprising obtaining a contextual datum from an electronic message sent or received by a terminal of said network, said contextual datum comprising a descriptive label describing the content of said message and an identifier of a sender and/or receiver of said message, and obtaining a text feature from said descriptive label. The invention also relates to the corresponding electronic devices, computer program products and media.

Description

DESCRIPTION

Title of the invention: Method for collecting data, method for exploiting collected data, electronic device and corresponding computer program and support products

1. Technical area

This application relates to the field of the collection of data concerning at least one group of individuals, to constitute and/or enrich knowledge bases, as well as the exploitation of data concerning individuals in such knowledge bases. . This may in particular be a large group of individuals.

It relates in particular to a method for collecting data and a method for exploiting collected data as well as electronic devices, computer program products and corresponding media.

2, State of the art

Knowing the members of a group is an issue for the people in charge of managing this group, in particular when the size of the group grows.

Indeed, for a large group, it is very difficult for a person to know all the members of the group well and/or to remember their specificities in detail, a fine solution may consist in keeping a register of the members of the group (such as a customer file). Such a register can for example take the form of a database gathering information on the members, such as user profiles. Such registers can for example allow a statistical analysis of the group, or the implementation of collective actions on members of the group.

However, information relating to a member (such as a user's profile, or the address of a client company) is often updated manually or following a specific request from this member or from a registry manager. As a result, necessary updates are sometimes omitted or delayed. Several years can thus pass between an event and the corresponding update of the register (for example the change of telephone number of a contact person of a client company or supplier who is seldom contacted). Moreover, when the group is very large, the sharing of information within a group (for example between departments and/or branches of the same company) can prove to be difficult. As a result, it may not be possible for a registry manager to be informed of the need for an update. Thus, the more the size of the group increases, the more difficult it can be to maintain a register concerning its members and in particular to ensure the relevance of the information in the register.

The object of the present application is to propose improvements to at least some of the disadvantages of the state of the art.

3. Disclosure of Invention

The present application aims to improve the situation, according to a first aspect, using a method implemented in a communication network, comprising:

• obtaining at least one contextual data item of an electronic message sent or received by at least one terminal of said communication network, said contextual data item comprising at least one descriptive wording of the content of said electronic message and at least one identifier of at least one at least one sender and/or receiver of said message;

* Obtaining at least one textual characteristic from said descriptive wording. In at least one embodiment, the method comprises storing said characteristic obtained, in association with an identifier of said at least one transmitter and/or receiver.

In at least one embodiment, the method comprises an anonymization of said contextual data.

In at least one embodiment, said anonymization comprises a replacement in said at least one contextual datum of at least one of said identifiers of said sender and/or of said receiver of said message by an identifier of a group to which said sender and/or belongs. receiver.

In at least one embodiment, the method comprises an association of said characteristic obtained with an identifier of said group.

In at least one embodiment, the method includes obtaining said textual feature includes:

• a lemmatization of said wording taking into account at least one language of said descriptive wording;

• an extraction of said textual characteristic of said lemmatized wording.

The present application relates for example to a method implemented by at least one device of a communication network, comprising: obtaining at least one descriptive wording of an electronic message; a detection of the languages of the words of said descriptive wording according to a membership of said words of said descriptive wording in at least one electronic dictionary of a plurality of candidate electronic dictionaries containing words in at least one candidate language; obtaining the canonical forms of words of said descriptive wording taking into account the languages of said words of said descriptive wording; storage in at least one memory of said device of said canonical forms obtained, in association with at least one identifier relating to a sender and/or to a receiver of said electronic message.

In certain embodiments, the method comprises a test for the presence, in the descriptive wording, of accent and/or lexical elements specific to one of said candidate languages.

In certain embodiments, the method comprises a calculation of a probability of belonging of said descriptive wording to one of said candidate languages.

In certain embodiments, said calculation is implemented when said presence test is unsuccessful.

In some embodiments, the method includes filtering said descriptive labels.

In some embodiments, the method includes anonymizing said descriptive labels.

In certain embodiments, the method comprises a deletion of non-informative words from said descriptive labels taking into account that said words of said labels belong to at least one dictionary of words to be deleted in at least one of said candidate languages.

In certain embodiments, said identifier relating to said sender and/or to said receiver of said message is an identifier of a group to which said sender and/or receiver belongs.

In certain embodiments, said method is implemented locally by said device.

The present application aims to improve the situation, according to a second aspect, using a method for exploiting data structures comprising at least one textual characteristic, said structures being stored on at least one computer storage medium accessible from an electronic device, said method comprising: • an at least partially automatic selection of at least a first of said textual characteristics taking into account an association of said first textual characteristic with a research topic on said data structures;

• an extraction of at least part of the data from at least a first of said structures, taking into account said at least one textual characteristic of said first structure and said first textual characteristic selected.

Said method can be implemented at least partially in an electronic device.

In at least one embodiment, said sender and/or receiver is at least one user of at least one terminal of said communication network.

In at least one embodiment, the method includes:

• an at least partially automatic selection of at least a first of said textual characteristics taking into account an association of said first textual characteristic with a research topic on said data structures;

In at least one embodiment, said selection comprises:

• obtaining at least one candidate characteristic from among said textual characteristics taking into account a semantic proximity with at least one textual characteristic already associated with said theme;

• a selection of said first characteristic from said at least one candidate characteristic.

In at least one embodiment, said selection comprises:

• obtaining at least one textual definition of said textual and/or candidate characteristics from a knowledge base;

• a selection of said first characteristic taking into account the presence in said at least one definition obtained of at least one first word relating to the search.

The present application relates for example to a method implemented by at least one electronic device of a communication network and comprising: obtaining data structures, said data of a structure comprising at least one textual characteristic, said data structures being stored on at least one computer storage medium accessible from said electronic device; automatic obtaining of at least a first candidate characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one textual characteristic sought; at least one automatic obtaining of at least one second candidate textual characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one of said first and/or second candidate characteristics already obtained; an automatic exploitation of said data structures, taking into account the presence in said data structures of at least a third of said first and second candidate characteristics obtained.

In certain embodiments, said method comprises: an automatic selection of said third candidate textual characteristic among said first and second candidate characteristics by taking into account the presence of at least one search word associated with said searched characteristic, in at least one definition obtained, from a knowledge base, for said first and/or second candidate textual characteristic.

In certain embodiments, said at least one textual characteristic sought is obtained from a user interface of said electronic device and/or a storage medium accessible from said electronic device.

In some embodiments, said third candidate text feature is selected from among said first and second candidate features via a user interface of said electronic device.

In certain embodiments, said searched word is obtained from a user interface of said device and/or from a storage medium accessible from said electronic device.

The characteristics presented separately in the present application in connection with certain embodiments of one of the methods of the present application can be combined together according to other embodiments of these methods. According to another aspect, the present application also relates to an electronic device adapted to implement at least one of the methods of the present application in any one of its embodiments. For example, the present application thus relates to an electronic device comprising at least one processor configured to:

• obtaining at least one textual characteristic from said descriptive wording. The present application also relates, for example, to an electronic device comprising at least one processor configured for: obtaining at least one descriptive wording of an electronic message; a detection of the languages of the words of said descriptive wording according to a membership of said words of said descriptive wording in at least one electronic dictionary of a plurality of candidate electronic dictionaries containing words in at least one candidate language; obtaining the canonical forms of words of said descriptive wording taking into account the languages of said words of said descriptive wording;

- storage in at least one memory of said device of said canonical forms obtained, in association with at least one identifier relating to a sender and/or to a receiver of said electronic message.

For example, the present application thus relates to an electronic device comprising at least one processor configured for an exploitation of data structures comprising at least one textual characteristic, said structures being stored on at least one computer storage medium accessible from said electronic device, said operation including:

• an at least partially automatic selection of at least a first of said textual characteristics taking into account an association of said first textual characteristic with a research topic on said data structures; • an extraction of at least part of the data from at least a first of said structures, taking into account said at least one textual characteristic of said first structure and said first textual characteristic selected.

The present application also relates, for example, to an electronic device comprising at least one processor configured to obtain data structures, said data of a structure comprising at least one textual characteristic, said data structures being stored on at least one medium computer storage accessible from said electronic device; automatic obtaining of at least a first candidate characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one textual characteristic sought; at least one automatic obtaining of at least one second candidate textual characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one of said first and/or second candidate characteristics already obtained; an automatic exploitation of said data structures, taking into account the presence in said data structures of at least a third of said first and second candidate characteristics obtained.

The present application also relates to a computer program comprising instructions for the implementation of the various embodiments of the above methods, when the program is executed by a processor and a recording medium readable by an electronic device and on which the computer program is recorded.

For example, the present application thus relates to a computer program comprising instructions for the implementation, when the program is executed by a processor of an electronic device, of a method comprising:

• obtaining at least one contextual data of an electronic message transmitted or received by at least one terminal of said communication network, said contextual data comprising at least one descriptive wording of the content of said electronic message and at least one identifier of at least one sender and/or receiver of said message;

* Obtaining at least one textual characteristic from said descriptive wording. The present application also relates, for example, to a computer program comprising instructions for the implementation, when the program is executed by a processor of an electronic device of a communication network, of a method comprising: a obtaining at least one descriptive label of an electronic message; a detection of the languages of the words of said descriptive wording according to a membership of said words of said descriptive wording in at least one electronic dictionary of a plurality of candidate electronic dictionaries containing words in at least one candidate language; obtaining the canonical forms of words of said descriptive wording taking into account the languages of said words of said descriptive wording; a storage in at least one memory of said device of said canonical forms obtained, in association with at least one identifier relating to a transmitter and/or to. a receiver of said electronic message.

For example, the present application also relates to a computer program comprising instructions for the implementation, when the program is executed by a processor of an electronic device, of a method for exploiting data structures comprising at least a textual characteristic, said structures being stored on at least one computer storage medium accessible from an electronic device, the method comprising:

* an at least partially automatic selection of at least a first of said textual characteristics taking into account an association of said first textual characteristic with a research topic on said data structures;

For example, the present application thus relates to a computer program comprising instructions for the implementation, when the program is executed by a processor of an electronic device, of a method comprising: obtaining data structures, said data of a structure comprising at least one textual characteristic, said data structures being stored on at least one computer storage medium accessible from said electronic device; automatic obtaining of at least a first candidate characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one textual characteristic sought; at least one automatic obtaining of at least one second candidate textual characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one of said first and/or second candidate characteristics already obtained; an automatic exploitation of said data structures, taking into account the presence in said data structures of at least a third of said first and second candidate characteristics obtained.

Furthermore, the present application relates for example to a recording medium readable by a processor of an electronic device and on which is recorded a computer program comprising instructions for the implementation, when the program is executed by the processor , of a method comprising:

* obtaining at least one contextual data of an electronic message sent or received by at least one terminal of said communication network, said contextual data comprising at least one descriptive wording of the content of said electronic message and at least one identifier of at least at least one sender and/or receiver of said message;

• obtaining at least one textual characteristic from said descriptive wording. The present application also relates, for example, to a recording medium readable by a processor of an electronic device of a communication network and on which is recorded a computer program comprising instructions for the implementation, when the program is executed by the processor, of a method comprising:

- Obtaining at least one descriptive wording of an electronic message:

- detection of the languages of the words of said descriptive wording according to a membership of said words of said descriptive wording in at least one electronic dictionary of a plurality of candidate electronic dictionaries containing words in at least one candidate language; obtaining the canonical forms of words of said descriptive wording taking into account the languages of said words of said descriptive wording; storage in at least one memory of said device of said canonical forms obtained, in association with at least one identifier relating to a sender and/or to a receiver of said electronic message.

For example, the present application also relates to a recording medium readable by a processor of an electronic device and on which is recorded a computer program comprising instructions for the implementation, when the program is executed by the processor, of a method for exploiting data structures comprising at least one textual characteristic, said structures being stored on at least one computer storage medium accessible from an electronic device, the method comprising:

• an extraction of at least part of the data from at least a first of said structures, taking into account said at least one textual characteristic of said first structure and said first textual characteristic selected. The programs mentioned above may use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in partially compiled form, or in n any other desirable shape.

The present application also relates, for example, to a recording medium readable by a processor of an electronic device and on which is recorded a computer program comprising instructions for the implementation, when the program is executed by the processor, of a method comprising: obtaining data structures, said data of a structure comprising at least one textual characteristic, said data structures being stored on at least one computer storage medium accessible from said electronic device; automatic obtaining of at least a first candidate characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one textual characteristic sought; at least one automatic obtaining of at least one second candidate textual characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one of said first and/or second candidate characteristics already obtained; an automatic exploitation of said data structures, taking into account the presence in said data structures of at least a third of said first and second candidate characteristics obtained.

The information carriers mentioned above can be any entity or device capable of storing the program. For example, a medium may comprise a storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or even a magnetic recording means.

Such a storage means can for example be a hard disk, a flash memory, etc. On the other hand, an information medium can be a transmissible medium such as an electrical or optical signal, which can be conveyed via an electrical or optical cable, by radio or by other means. A program according to the invention can in particular be downloaded from an Internet-type network.

Alternatively, an information medium may be an integrated circuit in which a program is incorporated, the circuit being adapted to execute or to be used in the execution of any one of the embodiments of the method which is the subject of the present application for patent.

4. Brief description of drawings

Other characteristics and advantages of the invention will emerge more clearly on reading the following description of particular embodiments, given by way of simple illustrative and non-limiting examples, and the appended drawings, among which: [Fig 1] presents a simplified view of a system, cited by way of example, in which at least certain embodiments of the method of the present application can be implemented,

[Fig 2] presents a simplified view of a device adapted to implement at least certain embodiments of one of the methods of the present application, [Fig 3] presents an overview of the method for collecting data from the present application, in some of its embodiments,

[Fig 4] presents an overview of the data mining method of the present application, in some of its embodiments, [Fig 5] presents an example of a cloud of words associated with the theme

"Artificial Intelligence" in certain embodiments of the data exploitation method of the present application,

[Fig 6] presents an example of thematic network obtained for certain embodiments of the data exploitation method of the present application,

[Fig 7] presents an example of technical skills resulting from an automatic selection for the "Artificial Intelligence" theme, in certain embodiments of the data exploitation method of the present application,

[FIG 8] presents an example of technical skills resulting from an automatic selection for the “Bigdata” theme in certain embodiments of the data exploitation method of the present application.

[FIG 9] presents an example of graphical restitution of an extraction in certain embodiments of the data exploitation method of the present application.

5. Description of embodiments

The present invention aims to collect data automatically (or at least partially automatically) concerning individuals of a group of individuals based on electronic exchanges by messages between and/or with these individuals. This data can for example be used for example to constitute or enrich a knowledge base concerning these individuals. These may be messages such as audio, video and/or textual electronic messages such as for example electronic mails (emails or “email” according to the English terminology) or instant messages (chat).

The digitization of exchanges between individuals indeed offers increased possibilities of at least partially automatic acquisition of information relating to these individuals, for example by means of automatic language processing techniques.

This information to be acquired is, for example, “properties” attached to individuals (such as information relating to the skills of these individuals).

By individual, we mean a sender or a receiver of a message. It can therefore be a single person or a plurality of persons acting collectively in the context of these exchanges (such as for example a plurality of persons accessible through the same email address). In the latter case, according to the embodiments, the identified properties can be attributed to a single “fictitious” individual associated with the plurality of individuals or with each of the persons. The group of individuals may in particular comprise a large number of individuals. It may be, for example, members of a sports federation, an association, a public or private company, and/or a community with a large number of employees, customers and/or suppliers ( or more generally interlocutors). They may also be members of an inter-project group, such as members of a collaborative communication platform such as “Slack © and/or a corporate social network.

The present invention is presented in more detail below in relation to embodiments, cited by way of non-limiting examples, which relate to a knowledge base of individuals of a company (for example a company considered as a Large Company (GE), with more than 5,000 employees). The individuals are for example the collaborators of the company or part of a company and the properties to be acquired relate to the skills of these individuals, such as skills in a particular technical field (such as a particular programming language, networks of neurons in artificial intelligence, etc.). Indeed, the identification of skills within a large company is a key factor for its organization and for its challenges. This identification can prove difficult in companies offering their employees a large number of professions, implying various skills for the employees, or when these skills are likely to change rapidly over time, due for example to numerous departures and /or arrivals of collaborators and/or a very evolving technological environment (as in high-tech sectors such as the telecommunications sector, aeronautics, the pharmaceutical or medical sector, etc.).

The present application will now be described in connection with FIG. 1 in more detail. Figure 1 depicts a telecommunications system 100 in which certain embodiments of the invention may be implemented. The system 100 comprises one or more electronic devices, at least some of which can communicate with each other via one or more communication networks 120, possibly interconnected, such as a local network or LAN (Local Area Network, 1) and/or a wide area type network , or WAN (Wide Area Network, 2). For example, the network may include a corporate or home LAN network and/or a WAN network of the internet or cellular type, GSM - Global System for Mobile Communications, UMTS - Universal Mobile Telecommunications System, Wifi - Wireless, etc.) . As illustrated in FIG. 1, the system 100 can also include several electronic devices, such as a terminal (such as a laptop computer 110, a smartphone 130, 132, 134, a tablet 136), and/or a server 140, for example an applications server, a storage device 150. The system may also include management and/or network interconnection elements (not shown). These electronic devices can be associated with at least one individual 160, 162, 164, 166 (for example through a user account accessible by login), some of the electronic devices 110, 130 being able to be associated with the same user 160. FIG. 2 illustrates a simplified structure of an electronic device 200 of system 100, for example device 110, 130 or 140 of FIG. 1, adapted to implement the principles of the present application. According to the embodiments, it may be a server, and/or a terminal.

The device 200 notably comprises at least one memory M 210. The device 200 can notably comprise a buffer memory, a volatile memory, for example of the RAM type (for "Random Access Memory" according to the English terminology), and/or a non-volatile memory. volatile (for example of the ROM type: (for "Read Only Memory" according to the English terminology). The device 200 can also comprise a processing unit UT 220, equipped for example with at least one processor P 222, and controlled by a computer program PG 212 stored in memory M 210. On initialization, the code instructions of the computer program PG are for example loaded into a RAM memory before being executed by the processor P. Said at least one processor P 222 of the processing unit UT 220 can in particular implement, individually or collectively, any of the embodiments of one and/or the other of the methods of the present application (described in particular in relation with the fig ure 3), according to the instructions of the computer program PG.

The device can also comprise, or be coupled to, at least one I/O input/output module 230, such as a communication module, allowing for example the device 200 to communicate with other devices of the system 100, via wired or wireless communication interfaces, and/or such as an interfacing module with a user of the device (also called more simply in this application “user interface”). an interface integrated into the device 200, or a part of a third-party device coupled to this device by wired or wireless communication means. For example, he can it may be a secondary screen of the device or a set of loudspeakers connected by wireless technology to the device.

A user interface can in particular be a so-called “output” user interface, suitable for rendering (or controlling a rendering) of an output element of a computer application used by the device 200, for example an application running at least partially on the device 200 or an "online" application running at least partially remotely, for example on the server 140 of the system 100. English terminology) on at least one user interface, in any form, for example comprising textual, audio and/or video components, or a combination of such components.

Examples of output user interface of the device include one or more screens, in particular at least one graphic screen (touchscreen for example), one or more loudspeakers, a connected helmet. The interface of the device 200 can for example be adapted to the renderings illustrated by FIGS. 5 to 9.

Furthermore, a user interface can be a so-called “input” user interface, suitable for acquiring information from a user of the device 200. It can be in particular information intended for an application computer accessible via the device 200, for example an application running at least partially on the device 200 or an "online" application running at least partially remotely, for example on the server 140 of the system 100. Examples of The input user interface of the device 200 includes a sensor, an audio and/or video acquisition means (microphone, camera (webcam) for example), a keyboard, a mouse.

In certain embodiments, said at least one microprocessor of device 200 can for example be adapted to:

• obtaining at least one textual characteristic from said descriptive wording. In certain embodiments, said at least one microprocessor of device 200 can for example be adapted to: obtaining at least one descriptive label of an electronic message; a detection of the languages of the words of said descriptive wording according to a membership of said words of said descriptive wording in at least one electronic dictionary of a plurality of candidate electronic dictionaries containing words in at least one candidate language; obtaining the canonical forms of words of said descriptive wording taking into account the languages of said words of said descriptive wording; storage in at least one memory of said device of said canonical forms obtained, in association with at least one identifier relating to a sender and/or to a receiver of said electronic message.

In certain embodiments, said at least one microprocessor of device 200 can for example be adapted for use of data structures comprising at least one textual characteristic, comprising:

• obtaining data structures, said data of a structure comprising at least one textual characteristic, said data structures being stored on at least one computer storage medium accessible from said electronic device;

• an automatic obtaining of at least a first candidate characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one textual characteristic sought;

• at least one automatic obtaining of at least one second candidate textual characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said characteristics texts of said data structures with at least one of said first and/or second candidate features already obtained;

• an automatic exploitation of said data structures, taking into account the presence in said data structures of at least a third of said first and second candidate characteristics obtained.

Some of the above input-output modules are optional and may therefore be absent from device 200 in some embodiments. In particular, if the present application is sometimes detailed in connection with a device communicating with at least a second device of the system 100, the method can also be implemented locally by a device, when it is used successively by several users exchanging messages asynchronous for example.

On the contrary, in some of its embodiments, the method can be implemented in a distributed manner between at least two devices 110, 130, 132, 134, 136, 140 and/or 150 of the system 100.

The term "module" or the term "component" or "element" of the device here means a hardware element, in particular wired, or a software element, or a combination of at least one hardware element and at least one software item. The method according to the invention can therefore be implemented in various ways, in particular in wired form and/or in software form.

Figure 3 illustrates some embodiments of the data collection method 300 of the present application. The method 300 can for example be implemented by the electronic device 200 illustrated in FIG. 2.

As illustrated in FIG. 3, the method 300 for collecting data can comprise obtaining 310 (or collecting) data from electronic messages exchanged between several individuals, using for example at least one of the electronic devices of the system 100.

In the detailed embodiments, the data (or traces) obtained from an electronic message include, for example, a date and/or time of transmission and/or reception of the message, an identifier of at least one sender and/or receiver of the message, and/or a descriptive wording of the message (such as a title, or subject, of the message in the case of an e-mail for example, or a textual content of an e-mail or instant message). In the context of an audio message or a video message having an audio component, the descriptive wording may correspond, for example, to a textual transcription of at least part of the message (obtained for example by voice to text conversion techniques known as STT (Speech To Text) according to the English Terminology).

As explained in more detail below, the method can also include obtaining 330 at least one textual characteristic, from the contextual data of a message. Various treatments can be carried out on the data collected to obtain these characteristics.

Thus, in some embodiments, the method 300 may include filtering

331 of the data collected. For example, some data considered personal or confidential may be deleted. This filtering can take into account the descriptive wording of the message. Thus, this filtering can be carried out at least partially automatically, by applying filtering rules, such as the deletion of data from a message, in the presence of a keyword (e.g., "personal",

"confidential", "private", etc.) indicative of a personal or confidential nature of a message, in a descriptive wording obtained.

In Figure 3, the data collection method also includes a detection step

332 of at least one language (for example French and/or English) used in a descriptive wording of a message, among a plurality of candidate languages.

According to the embodiments, and in particular according to the data collected (their size, the "candidate" languages supported, and the security constraints relating to this data, particularly in terms of access, for example), different methods can be used to detect a language. For example, in the illustrated embodiment, detection 332 can take advantage of lexical differences between at least two candidate languages, such as French and English. Indeed, with the exception of the mention of entities and/or individuals (name of persons, company, etc.), an English sentence does not contain accents or certain special characters (such as é, è, ù , ô, û, ....), the opposite of a sentence in French. In addition, there are lexical connectors, such as "linking words", (such as "for", "des", "la", "le", "then", "etc" for French and (by, the, for, ...). for English), which are specific to certain languages, and can therefore help to differentiate them. In certain embodiments, the method can therefore comprise a test for the presence, in the descriptive wording, of accent and/or of lexical elements specific to one of the candidate languages.

In certain embodiments, for example when the previous test was unsuccessful, the method can comprise a calculation of a probability of belonging of the descriptive wording to one of the candidate languages. Such a probability can for example be defined taking into account an intersection between the words of the label and the existing words in an electronic dictionary of a candidate language. The algorithm below can for example be implemented on the label (which is in this example the subject of an email) in connection with a FR dictionary in French language and with an EN dictionary in French English language.

Algorithm input: Subject of an email

Initialization: fr_special_characters = ["à","â","ç","é","è","œ","ê","ù","ô",....] fr relative words = ["the", "the", "not", "here", "then", "also", "again", "by", "in", "on", "for”, ....] in relative words = ["by", "the", "in", "under", "for", ....]

First test: intersection_elt = fr_caracteres_speciaux ∩ the alphabets constituting the object

If intersection_elt != 0 then Return fr End of algorithm Otherwise: Second test: fr element = fr_mots_relatifs ∩ the list of words constituting the object

If fr_element != Φ then Return fr End of algorithm

Otherwise in element = in relative words ∩ the list of words constituting the object

If in element != Φ then Return to End of algorithm Otherwise: Third test: fr element = the list of words constituting the object ∩ list of words constituting the dictionary fr en element = the list of words constituting the object ∩ list words making up the dictionary en If the size of fr element > en element then Return fr End of algorithm

Else Return to End of Algorithm

Experimental results

The example of detection detailed above can present, in certain embodiments, advantages in terms of relevance of language detection with respect to. certain existing algorithms, for example algorithms using the Python programming language.

In particular, it may for example, in certain embodiments, not be sensitive (or very weakly) to the length of the input sentences, unlike certain other algorithms whose results degrade as the length of input sentences increases.

The embodiment detailed above also offers the advantage of being able to be implemented locally without access to. Internet (hors-ligne - or "offline" according to the English terminology). Thus, it can be adapted, at least in certain embodiments, to use of data stored in a secure environment without Internet access.

The table below represents the correct classification rate obtained, by several algorithms, including the algorithm presented above, during a classification test, according to their language, of data from a test database containing 109 sentences (long and short). As the table shows, in this test, the algorithm presented above is the most efficient algorithm, compared to the other algorithms tested, and is 99% successful in assigning the correct language to each sentence.

In certain embodiments, the detection 332 can allow detection of the presence, in a wording classified as relating to a first language, of words of a second language. Indeed, an email title in French may for example contain technical words in English such as "data", "machine learning").

In certain embodiments, for example when an automatic detection has not made it possible to determine with certainty (or with a sufficient probability, for example a probability greater than a first “so-called threshold” value) a language used, the method can include obtaining or verifying the language used in at least one descriptive wording via a user interface (such as the user interface of the device 200 for example).

The data collection method of the present application may also comprise a step 333 of lemmatizing at least part of the descriptive wording, for example certain words or set of words of the wording. A lemmatization designates a lexical treatment given to a word in such a way as to associate it with its canonical neutral form, and to thus be able to group the words of the same family in the same canonical form, For example, a lemmatization makes it possible to group together verbs, nouns, and/or adjectives associated with the same canonical form (ior lemrne) in a common coding.

According to the embodiments of the invention, several lemmatization methods can be implemented. Thus, in embodiments where the lemmatization is performed on a label after a detection 332 (or determination) of at least one language used in the label, different lemmatization methods can be used depending on the language used, For example, in certain embodiments, the “treetagger” lemmatization method can be used for a label (or part of a label) in French and while the “nltk” lemmatization method can be used for English. According to another example, it can be used, for the English language for example, a lemmatization method based on the “nltk” method but making certain modifications thereto. For example, a lemmatization method based on the “nltk” method but integrating a syntactic function (verb, noun, etc.), such as that used by the treetagger method, can be used. Such integration can indeed allow, in certain embodiments, a better identification of the syntax of a word than with the “nltk” method, thus resulting with more probability in obtaining the canonical form of the word. (For example, for the sentence “the best moment be when / saw my baby walk on their foof\ the lemmatizer of “nltk” cannot consider “saw” as the verb “to see” in the past tense and therefore it remains unchanged ). Thus the integration of a syntactic function can make it possible, in certain embodiments, to obtain better performance than with the “nltk” method alone.

In certain embodiments, for example when the detection 332 has concluded that words from a second language are present in a label using a first language, a first lemmatization can be applied to the parts of the label in the first language of the label, a second lemmatization different from the first lemmatization being applied to the words in the second language. Indeed, if an email title in French contains, for example, technical words in English such as "data", "machine learning", a lemmatizer adapted to French could transform these English technical words into "dater" and "machine learning" respectively , which would not make sense, In certain embodiments, after the lemmatization(s) step, the wording only contains verbs, nouns, adjectives and lexical connectors (such as adverbs).

The method of the present application may also comprise, in certain embodiments, a step 334 of deleting non-informative words (“stop words” according to English terminology), such as lexical connectors (such as “before”, “for “, “on”, “the”, etc.). The detection of these non-informative words to be deleted can for example implement a comparison of the words of the label with those of at least one list (or dictionary) grouping non-informative words in at least one of the candidate languages. For example, the method can implement a consultation of at least one list corresponding to the language(s) detected in the label from among at least a plurality of lists, each being specific to one of the languages candidates.

In some embodiments, the data collection method may include a detection 335 of compound words, which should be processed together (and not separately). These may include compound words related to the field of application concerned and/or the resources (quality, skills) sought (such as "machine learning", "artificial intelligence", Edge computing, UX design in the field telecommunications or “grand cru” in the wine sector).

In certain embodiments, the detection 335 of compound words can for example take account of the presence in a technical dictionary and/or a knowledge base of an association of at least two successive words of a description.

In certain embodiments, the detection 335 of compound words can for example take into account a number of occurrences of a succession of at least a first and a second word in a plurality of labels and a comparison of this number of occurrences "in association", with the number of occurrences of these same words (first word, second word, etc.) taken separately. For example, in certain embodiments, the detection 335 of compound words can use a model such as “Phrases” from the Gensim library of Python to detect compound words in the set of labels of the messages obtained.

The obtaining step 330 thus results in labels containing only verbs, nouns and/or adjectives (words or groups of words such as compound words). These terms (or characteristics) relate to the senders and/or receivers of the messages (indeed, a topic addressed in an exchange can involve both the sender and the receiver of the message). As illustrated in FIG. 3, the method can also comprise a step 340 for storing the characteristic(s) obtained. For example, the characteristics can enrich a knowledge base comprising at least some of the senders and/or receivers of the messages. In certain embodiments (in particular when the identifier of the sender and/or of the receiver of the message has not previously been "anonymized" (ie made anonymous) as explained below), these characteristics can be associated with sender or to the receiver of the message (for example, they can be added to his personal data, so as to enrich his "profile"), or alternatively to at least one terminal of the sender or receiver of the message. Alternatively or cumulatively, in certain embodiments, these characteristics can be associated with an entity of several individuals to which the sender or receiver of the message belongs (or as a variant an entity of several terminals of which at least one belongs to the sender or of the receiver of the message), for example a substitution entity identified, during an anonymization 321 of an identifier of the sender and/or of the receiver of the message as explained below.

In some of the embodiments of the data collection method 300, at least certain stored characteristics can be used subsequently by the exploitation method 400 described below in connection with FIG. 4, in certain of its embodiments.

In some embodiments, the data collection method may include a step 320 of anonymizing the data obtained during step 310. This step may be optional or systematic depending on the embodiments, for example depending on the data collected , and/or securing the equipment or network through which the collected data will pass, and/or depending on the regulations in force in a country in which the process for collecting this application is implemented and/or another internal regulation of a company implementing the process for collecting this request (in application, for example, of the General Data Protection Regulation (RGPD) (or GDPR, from the English "General Data Protection Regulation ") of the European Union).

In the example illustrated, the anonymization 320 may comprise a substitution 321, in the data obtained from a message, of the identifier of the sender and/or of the receiver of the message by another identifier, for example a relative identifier to an entity (hereinafter referred to as a substitute entity) to which belongs (is attached) to the at least one sender and/or receiver of the message (or as a variant a terminal of this sender and/or this receiver). It may thus be an identifier such as an alphanumeric reference or a visual acronym (logo, particular color, etc.) relating to an entity such as a sports team and/or a hierarchical or functional entity at the within a company.

According to the embodiments, the substitution entity can be chosen (or selected) taking into account various criteria, such as a number of individuals (or terminals) belonging to the substitution entity, and/or that a proximity (logical, functional and/or geographic) between a substitute entity and the sender or receiver of the message concerned. For example, in a company, the substitution entity can be the hierarchical entity, the closest to the individual to which belongs a number of employees greater than a certain value (for example a number of employees at least equal to 1 , 5, 10 or 15 employees).

A criterion relating to a minimum size of a surrogate entity can indeed limit the risks of identification of an individual in the surrogate entity. This minimum size, and therefore the degree of precision desired during a subsequent analysis of the data, can vary according to the embodiments. In particular, it may depend on the potential number of message senders and/or receivers, regulatory constraints (linked to the GDPR for example) and/or the sensitive nature of a theme that may be associated with it (see below). -after), For example, a larger minimum size can be chosen for embodiments in the medical field than in the sports field.

In some embodiments, the anonymization 320 may comprise an anonymization 322 of the descriptive wording, in order to remove, where appropriate from a wording, at least one indication that can help identify at least one sender and/or at least one at least one receiver of the message whose data is collected or other individuals or entities implied by the data collected (for example to remove the name and/or first name of at least one person or entity named in a textual data). According to the embodiments, the anonymization 321 of at least one of the identifiers and/or the at least partial anonymization 322 of the wording can be carried out before, during or the obtaining and/or the storage of the characteristic. For example, the anonymization 321 of the identifier can be performed in parallel with the obtaining 330 of at least one characteristic from the wording. The anonymization 322 of the label, if it is carried out before the detection of the language of the label can offer the advantage of potentially deleting certain words (such as names of persons or entities) whose presence in a label could disturb the detection the language of the label (such as a first name with accent in an English label). In addition, it may be more efficient, in terms of processing load, to anonymize a label upstream of language detection, lemmatization, deletion of non-informative words and/or detection of compound words) in order to possibly shorten a label and therefore to limit the input data of these steps.

A method 300 for collecting data relating to a group of individuals has been detailed above, in order to obtain characteristics intended to be stored for subsequent use in relation to a group of individuals or entities comprising these individuals ( for example to enrich a knowledge base of a group of individuals or of entities comprising these individuals).

A method 400 for exploiting (or using) data relating to a group of individuals, and in particular textual characteristics of a group of individuals, or of entities comprising these individuals, is now detailed. The group of individuals can correspond for example to the employees of a company, or of a part of a company. According to the embodiments, at least some of the characteristics of at least one of the individuals or entities used by the exploitation method 400 may have been obtained by the collection method 300 described above and/or in a different way. For example, characteristics relating to at least one individual may have been entered via a user interface (for example, in a group of employees of a company, during hiring or following collective training). Thus, according to the embodiments, the collection method and the exploitation method can be implemented independently, or jointly.

As illustrated in FIG. 4, the method 400 of data exploitation of the present application may comprise an extraction 440 of data relating to a group of individuals (for example senders and/or receivers of electronic messages), according to at least some of the characteristics (from, for example, the wordings of the messages) associated with these individuals or entities, for example according to a particular characteristic or a plurality of particular characteristics, associated with a "theme" targeted for data exploitation. Thus, when the operating method 400 is applied for a group of employees of a company, a thematic can correspond to a technical field (for example quite general) of the company for which one wishes to identify skills available to the company, and/or for which one wishes to locate, within the hierarchical and/or functional organization of the company, employees with skills related to this technical field. For example, a plurality of themes related to the technical fields of the company can be identified (For example: Artificial intelligence, 5G, banking, insurance, .. ,.). Each theme can be associated with one or more words or sets of words characterizing it (and hereinafter called "characteristics"). For example, to an “Artificial Intelligence” theme, the characteristic words “machine learning” and “neural networks” can be associated.

A thematic and/or associated characteristics can be obtained 410 in various ways depending on the embodiments. For example, they can be obtained at least partially via a communication interface or a user interface of the device 200 of FIG. 2, and/or from a storage means accessible from the device 200 (for example from one or more configuration file(s), which can be local to the device or remote).

In the example of FIG. 4, obtaining 410 a theme and associated characteristics can for example comprise obtaining 411 a theme and/or first skills associated with this theme via a user interface, and be followed by automatically obtaining second characteristics associated with this theme. More precisely, the method can comprise an at least partially automatic obtaining 412 (determination) of a cloud of words (or semantic field) characterizing the topic obtained. Such a cloud of words can be obtained by calculating, or estimating, proximity relations between words from at least one electronic glossary with characteristics already associated with the topic (such as the first characteristics obtained at step 411) . Thus, words that are semantically close to the characteristics already associated with a topic can in turn be associated with the topic, and subsequently used for proximity calculations with other words of the glossary.

Gradually, by iterating the calculation, we can increase the cloud of characteristics associated with a theme,

The glossary can be dedicated to the group considered and include and/or correspond, in certain embodiments, to all the textual characteristics of all the individuals or entities of the group considered. For example, in connection with process 300 collection previously presented, the glossary can include all of the verbs, nouns and/or adjectives obtained following step 330 from the labels of the messages exchanged. These words (or characteristics) are indeed likely to translate the interests, or work subjects of the senders/receivers of the messages and therefore to correspond, at least for some, to the skills of these senders/receivers. The glossary can for example contain terms specific to the group of individuals concerned (for example terms linked to a context particular to the group considered), Such an embodiment can indeed make it possible to take into account, in connection with a theme, of terms which would be absent from a general dictionary (for example a dictionary on the market), or with a meaning quite different from that conferred on them by a message from which they would be extracted, for example.

In a variant, the electronic glossary may for example be a specialized dictionary, for example a technical dictionary. The determination 412 can for example use several models for representing textual elements, such as models for representing words in a mathematical repository (“word embedding” according to English terminology) (such as Word2vec© and Fasttext©). For example, after a learning phase of several models on the whole of the glossary (for example the set of characteristics of all the individuals or entities of the group considered), we can identify N characteristics that are semantically the closest to the same characteristic (already identified) of a theme using at least two of the learned models (with N integer, strictly positive). This identification can then be repeated incrementally starting from the characteristics already or newly identified. Of course, in certain embodiments, the number N of “closest identified characteristics” can vary according to the models used, for example the reliability of the models.

The learning can for example be carried out for a determined period (such as a period of a few months or tens of months, for example 12 months).

A criterion for stopping the increments can for example be the reception of a stop command via a user interface, the running of a number of iterations, a minimum and/or maximum number of characteristics associated with a theme. These numbers can be constant or vary depending on the size of the group of individuals or entities and/or the total number of characteristics of the individuals or entities in the group.

Thus, according to an exemplary implementation, the number N can be between 5 and 15 (for example 10) and 3 to 7 iterations (for example 5) can be performed. The method may include an optional step 413 of rendering (audio and/or visual) the word cloud on a user interface. Thus, a rendering of a word cloud comprising terms close to the theme "Artificial Intelligence" is illustrated by way of example in Figure 5.

In the embodiment of FIG. 4, the data exploitation method can comprise an extraction 440 of data related to a theme and at least some of its associated characteristics (for example at least part of the word cloud associated with the theme). This extraction may differ according to the embodiments.

According to FIG. 4, the method can also include a rendering 440 (or restitution), for example visual, of the result of this extraction 440. FIG. 6 illustrates an example of rendering, graphic, of a thematic mesh (or thematic network) entities bringing together employees of a company and concerned by the characteristic (here a skill) “process mining”. The branches of the mesh are organized by descending hierarchical levels in this example.

In certain embodiments, the extraction 440 of data can be preceded by obtaining at least one selection criterion to be used during the extraction 440 of data, in particular a criterion relating to the textual characteristics of the data.

For example, in an embodiment where the word cloud is rendered 413 on a user interface, one or more characteristics can be selected 430 via the user interface (for example by clicking on one or more word(s) of the cloud) . Their selection can allow for example an extraction of the identifiers of the individuals and/or of the entities associated with at least one of the characteristics selected (or with all the characteristics selected in other embodiments). Thus, if the word cloud contains the name of a platform (or a tool) related to the theme, the selection 430 of the name of this platform can launch an extraction making it possible to display during rendering 460 (for example graphic) of the entities or individuals associated with this platform in the data collected and therefore a priori concerned by this platform.

Alternatively, one or more features may be selected 430 via a user interface using a search engine limited to the perimeter of the word cloud. Thus in certain embodiments, the method may comprise the acquisition of at least one word from a user interface and a selection of at least one word from the word cloud as a function of said acquired word or words. In another embodiment, a selection 430 can be made automatically from among the characteristics of the cloud of characteristics associated with the theme, thanks to filtering based on information relating to these characteristics and obtained via a knowledge base (such as Wikipedia ©).

More precisely, the method can comprise obtaining 420 from a knowledge base such as Wikipedia, adapted for example to the language of the words or characteristics present in the glossary, of at least one definition of at least one of the characteristics of the glossary. In some embodiments, for example when done before selecting a topic (as shown in Figure 4), getting definitions 420 can be done for all of the features in the glossary. Such an embodiment can make it possible to limit the cost, in time and in computing resources, of a possible subsequent automatic selection 430 . In certain embodiments, for example when it is carried out after selection of a theme and determination 412 of the cloud of associated characteristics (systematically and/or upon activation of an automatic selection command of characteristics of the cloud), the getting definitions 420 can be done for features in the feature cloud only. Such an embodiment can help to limit the number of searches in the knowledge base.

When searching for a word on a knowledge base such as Wikipedia, several paragraphs may be returned, including a definition of the word as well as examples of contextual applications of the word. In certain embodiments, the method can therefore comprise a (manual or automatic) selection of a paragraph from among the paragraphs obtained (received) to select a paragraph corresponding (probably) to the definition of the word. An automatic selection of a paragraph can be based on grammatical and/or syntactic rules. For example, an automatic selection can for example include:

• a verification of the presence of the word from the glossary/word cloud sought in at least one of the first two sentences of the paragraphs received,

• if the word is indeed present in a sentence of a paragraph, a calculation of the position of the word in the sentence, and

• a selection of the paragraph, if the word is located in the first or second position in the sentence.

Often, the explanation of an acronym appears, in a sentence, before the acronym itself, which shifts the position of the word sought in the sentence when it is a acronym. Therefore, in certain embodiments, the method can for example comprise, before calculating the position of a word:

• a test for the presence of parentheses around the word in the paragraph;

• when parentheses are present, a comparison of the first letters of the words preceding the searched word (or more exactly the opening parenthesis) with the letters composing the word in parentheses, and

• deletion of these words which precede the searched word (so as to bring the acronym closer to the beginning of the sentence before calculating its position in the sentence as described above).

The content of the selected paragraph can for example be considered as a definition of the relevant word of the glossary/cloud.

In some embodiments, the automatic selection 430 of characteristics from the cloud of characteristics associated with the thematic (to be used as an extraction criterion), can comprise a filtering 432 of the characteristics of the cloud on the basis of the definitions obtained for these characteristics. For this, the method can comprise for example obtaining 431 (from a user interface, a storage medium and/or a communication interface) a plurality of words likely to be mentioned in the definition of a word in connection with the theme. For example, when the theme relates to a particular technical field of software development (Artificial Intelligence), and the purpose of the extraction is to identify skills in this particular technical field, the plurality of words ("method", " algorithm”, “set of approaches”, etc.) can be chosen in such a way as to retain only definitions relating to a technical competence.

The filtering 432 can comprise a determination of the presence of at least one word of this plurality of words in the definitions obtained. It may also include a verification of at least one positioning criterion of at least one word of this plurality of words present in a definition obtained (such as positioning in the first and/or second sentence of the definition, and/or just before a verb for example).

The filtering can for example retain only words from the cloud having at least one definition for which at least one of the positioning criteria is verified.

Figure 7 and Figure 8 respectively represent examples of textual characteristics (technical skills in this example) resulting from an automatic selection for the “Artificial Intelligence” theme and for the “Bigdata” theme. The mining method can allow different extractions according to its embodiments. The examples of extraction cited above can help a user to explore more or less finely, data relating to a group of individuals or entities, in connection with certain characteristics (skills for example) that 'he is looking for (such as a name of the technical skill sought or a support tool name for the skill sought).

In particular, a user can obtain, at least in certain embodiments, more precise results than those which he could obtain by existing solutions. Indeed, in the field of collecting skills of individuals for example, the use of a register often leads to standardizing, for the purpose of classification, the skills within a group of individuals. As a result, a register is often limited to the identification of very general skills (eg specialist in artificial intelligence). According to the present application, it may on the contrary be possible to vary the degree of precision of the information returned, and in particular to detail very specific technical skills (such as very "specific" skills, or in a very narrow field of expertise, or even skills that are not very widespread (eg: specialist in word2vec).

In some embodiments, the method may include pre-rendering filtering of the extracted data. For example, after extraction 440 of information representative of at least some of the entities of the group concerned by at least one characteristic associated with the theme (for example a skill of the “Artificial Intelligence” theme), the method can comprise filtering the information representative data obtained to keep, for example, only the most significant entities, such as the I (I integer greater than 0) entities most often associated with a selected characteristic, and/or the J (J integer greater than 0) entities using most often the characteristic selected in the labels of messages having for a sender an individual of the entity or for a substitution entity this entity, and/or the K (K integer greater than 0) entities which most often use a selected characteristic and/or the L (L integer greater than 0) words calculated as being the closest to this selected characteristic (as explained in connection with the eta pe 410 for obtaining the thematic glossary).

In certain embodiments of the exploitation method implementing several themes, the method can comprise obtaining 410 a cloud of characteristics for several themes. The method may also include obtaining statistics concerning these themes and/or the distribution of textual characteristics of the data between these themes, such as for example a distribution (in number of characteristics and/or messages) of the themes in the group (or a sub-group such as one or more entities). For example, as illustrated in FIG. 9, it is possible in certain embodiments to obtain a characterization of each entity of the group in terms of theme(s), a portion of the glossary of words associated with a theme in at least one entity, and /or an extraction of the characteristics associated with a theme for an entity. Thus, figure 9 shows a rendering of the different themes addressed by an entity, the surface of the bubbles being representative of the importance of the theme in this entity, of the cloud of characteristics associated with a theme ("artificial intelligence") and of the characteristics associated with this theme in this entity.

The operating method, in at least some embodiments, may be useful for group steering. Thus, in certain embodiments, it can help a user to obtain (for example to visualize) a distribution of characteristics associated with at least one theme in the different entities of a group. He can thus, for example, help in the constitution of a new team, dedicated to at least one theme", transverse to entities of attachment of individuals (such as hierarchical entities) (for example a transverse project team in a company) .

An iterative use of the exploitation method, in at least certain embodiments, can also help to study the diffusion of a theme, for example how it spreads within a group

Example of application: Automatic determination and characterization of project teams in a company and the themes addressed by these projects.

With the assumption that the people who most often send emails whose title includes the name of a project are often those who belong to or are close to a project team, we can reconstitute a project team within a group by selecting the individuals associated with emails that mention the name of the project (or its acronym) in their title, with optional filtering on the entities, depending for example on the number of emails with such a title per entity. This entity can for example be a substitution entity added during the anonymization of the data, so as to keep only the entities which add up the greatest number of “project” emails. This approach can in particular be adapted to the reconstitution of a team from a cross-functional project to teams using different, even incompatible, project management tools.

The operating method can be implemented iteratively on a first group to obtain a second group then on the second group to obtain a third group. etc..., with identical or different themes.

In at least certain embodiments, the method of collecting and/or the method of exploitation can help to locate themes and technical skills in a company (hierarchical entities, projects), thanks to the electronic messages exchanged in the company. These electronic messages, issued within the framework of professional exchanges, can be considered as a reliable source of information. In addition, it is not distorted or modified before it is used (due, for example, to a standardization constraint). It can be precise and require neither the support of a manager nor a personal statement from employees. Where data (such as e-mail message titles) is collected frequently (eg, daily or weekly), the methods of this application may therefore help to maintain the distribution of skills within a company and obtain maps of skills relating to new or already identified themes more easily. For example, updates can be carried out automatically, at least in certain embodiments, without requiring an effort from the individuals concerned or their management, on the contrary of certain solutions of the prior art.

In at least certain embodiments, certain steps (such as the association of electronic messages with at least one topic) can be done without training data, which can facilitate the implementation of at least one of the methods of the this request.

Certain embodiments have been described with desired characteristics corresponding to technical skills. In other embodiments, the desired characteristics may be different (for example at least certain embodiments described above may be integrated into a management/audit tool of a company intended for Human Resources services (HR) Indeed, determining where the skills are in a company, evaluating the emergence of new skills and watching how they evolve temporally in the company are issues shared by many Human Resources managers in a company.

At least some embodiments described above can also be used in a management tool for a company's strategy (in order, for example, to visualize how an existing, new, or future theme is applied in the entities of a company)

At least some embodiments described above may also be implemented by enterprise production pilots (such as Project Management Offices "PMOs"). Thus, certain embodiments described above can also give a transverse view of at least one project team and/or provide an indicator of its life cycle (based for example on the number of occurrences of particular words in the exchanges (such as growth, stability, decrease), or the satisfaction of a team (based, for example, on the number of occurrences of particular words (with a positive sounding or conversely negative) in the exchanges).

Claims

CLAIMS Method implemented by at least one electronic device of a communication network and comprising: obtaining data structures, said data of a structure comprising at least one textual characteristic, said data structures being stored on at least one computer storage medium accessible from said electronic device; automatic obtaining of at least a first candidate characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one textual characteristic sought; at least one automatic obtaining of at least one second candidate textual characteristic among said textual characteristics of said data structures taking into account a semantic proximity of said textual characteristics of said data structures with at least one of said first and/or second candidate characteristics already obtained; an automatic exploitation of said data structures, taking into account the presence in said data structures of at least a third of said first and second candidate characteristics obtained. Method according to Claim 1, in which the said at least one textual characteristic sought is obtained from a user interface of the said electronic device and/or a storage medium accessible from the said electronic device. A method according to claim 1 or 2, wherein said third candidate text feature is selected from said first and second candidate features via a user interface of said electronic device. A method according to claim 1 or 2, wherein said method comprises: automatically selecting said third candidate textual feature from among said first and second candidate features taking into account the presence of at least one search word associated with said search feature, in at least one definition obtained, from a knowledge base, for said first and/or second candidate textual characteristic.

5. Method according to claim 4, wherein said searched word is obtained from a user interface of said device and/or from a storage medium accessible from said electronic device.

6. Method implemented by at least one device of a communication network, comprising:

- Obtaining at least one descriptive wording of an electronic message; a detection of the languages of the words of said descriptive wording according to a membership of said words of said descriptive wording in at least one electronic dictionary of a plurality of candidate electronic dictionaries containing words in at least one candidate language; obtaining the canonical forms of words of said descriptive wording taking into account the languages of said words of said descriptive wording; storage in at least one memory of said device of said canonical forms obtained, in association with at least one identifier relating to a sender and/or to a receiver of said electronic message.

7. Method according to claim 6 comprising a test for the presence, in the descriptive wording, of accent and/or lexical elements specific to one of said candidate languages.

8. Method according to claim 7 comprising a calculation of a probability of belonging of said descriptive wording to one of said candidate languages.

9. Method according to claim 8 wherein said calculation is implemented when said presence test is unsuccessful.

10. Method according to one of claims 6 to 9 comprising a filtering of said descriptive wordings.

11. Method according to one of claims 6 to 10 comprising an anonymization of said descriptive wordings.

12. Method according to one of claims 6 to. 11 comprising a deletion of non-informative words from said descriptive labels taking into account that said words of said labels belong to at least one dictionary of words to be deleted in at least one of said candidate languages.

13. Method according to one of claims 6 to 12 wherein said identifier relating to said sender and/or said receiver of said message is an identifier of a group to which said sender and/or receiver belongs.

14. Method according to one of claims 6 to 13 wherein said method is implemented locally by said device. Electronic device comprising at least one processor, said processor being configured to implement the method according to one of Claims 1 to 5 or the method according to one of Claims 6 to 14. Computer program comprising instructions for the implementation , when the program is executed by a processor of an electronic device, of a method according to one of Claims 1 to 5 or the method according to one of Claims 6 to 14 Recording medium readable by an electronic device and on which is recorded a computer program comprising instructions for the implementation, when the program is executed by a processor of said electronic device, of a method according to one of claims 1 to 5 or the method according to one of claims 6 to 14.