Method for managing the communication of semantic data

Abstract

Method for managing the communication of semantic data The invention relates to a method for managing the communication of semantic data (SD), called data to be communicated, by a first device (EQ1) to a second device (EQ2], characterized in that it comprises obtaining an index (I') corresponding to the data to be communicated (SD), said index (I') making it possible to find in a data structure (DS) another semantic data (SD'), called communicated data, the index (I') being chosen as a function of a distance (D) measured between the data to be communicated (SD) and the communicated data (SD'), followed by the transmission of the obtained index (I') through a communication channel (CNL) between the first device (EQ1) and the second device (EQ2). Figure 1

Description

Title of the invention: Method for managing the communication of semantic data Technical field

[0001] The technical field is that of telecommunications.

[0002] More specifically, the invention relates to a method for managing the communication of semantic data.

[0003] Semantic data will be communicated by the method according to the invention from one device to another device. Such data is a particular type of data that is used in the field of artificial intelligence. The capabilities of artificial intelligence, and in particular of generative artificial intelligence (translation of the English generative artificial intelligence), that is to say the branch of artificial intelligence that focuses on the automatic production of media, in particular texts, images or videos, appeared in the eyes of the general public with the launch of Chat-GPT (product of the company Open AI), the first version of which dates from November 2022. Other generative artificial intelligence programs are developed by the companies Google (product Gemini), Anthropic (product Claude) or Mistral AI (product Le Chat).The products mentioned here are conversational agents, but other generative artificial intelligence products can produce images (DALL-E products from the company Open AI, or Midjourney, from the research laboratory of the same name), videos, sounds, but also computer programs or specifications of chemical molecules. Other artificial intelligence tasks can concern, for example, pattern recognition, decision-making, classification, planning or any other type of activity imitating the capacities of human intelligence.

[0004] Generative artificial intelligence programs encode the data they manipulate for processing in a format that we call here semantic data or semantic vector. Semantic content, the meaning expected or interpreted by users of a text, image or video, is therefore associated with semantic data. Such semantic data is a high-dimensional vector in the space of real numbers. The typical number of dimensions for semantic data ranges from several hundred to several thousand. The MTEB project (acronym for Massive Text Embedding Benchmark) carries out performance comparisons between models used by generative artificial intelligence programs and, in March 2024, of the ten best-performing models, seven used vectors of dimension 4096 and three of dimension 1024.

[0005] The term most often used in English to talk about semantic data manipulated by generative artificial intelligence programs is embedding. However, this term confuses the function that will encode data such as text or an image for further processing, a mathematical function that is an embedding (translation of the English 'embedding') and the data obtained as a result of the application of this function. To avoid this confusion between a function and its result, we will subsequently use in this document only the term data or semantic vector, even if the term embedding is more often found in the literature, in English and French.

[0006] Semantic data are therefore characterized by the fact that they represent semantic content in the form of computer data. It is possible to see that semantic content is indeed associated with semantic data by looking at the transformations that can be applied to them. A classic example in artificial intelligence considers semantic data corresponding to the concept of "France" which can be transformed into another semantic data corresponding to the concept of "Paris" and the same transformation, applied to the semantic data associated with "Germany", will produce the semantic data associated with "Berlin" and the same for data "Spain"-"Madrid", "Italy"-"Rome" etc.This indicates that semantic content corresponding to a country or a city is indeed associated with the corresponding semantic data and that manipulations of the semantic data correspond to manipulations of the concepts associated with the semantic data.

[0007] Semantic data, initially appearing in the context of generative artificial intelligence, has found uses across the spectrum of artificial intelligence programs.

[0008] As indicated, the semantic data we are talking about are vectors of real numbers with dimensions ranging from several hundreds to several thousands. These several hundreds or thousands of real numbers which form a semantic data are generally represented for computer processing by a data type called a floating number (translation from the English float) which uses 32 bits to be represented, that is to say 4 bytes. A semantic data which is a vector of dimension 4096 will therefore have an approximate memory size of 16 KB (16 kilobytes, or 16 kilobytes in English, or 16 Kb).

[0009] It can be noted that the number of dimensions of semantic data tends to increase over time, and that the most efficient models tend to use higher dimensional vectors.

[0010] Some artificial intelligence researchers propose using vectors of complex numbers as semantic data. Such a proposal, if adopted and the number of dimensions remained constant, would involve multiplying by two the size of the semantic data by replacing a 32-bit floating point real number with two real numbers to represent a complex number.

[0011] Furthermore, we can talk about data communication in several cases. For example, there is communication when computer data is transmitted over a computer network, which is local to several computers, or over a telecommunications network which can be local or long distance, or when data is transferred from one medium to another, that is to say for example copied from a file system of a computer to a memory medium connected to the computer such as a USB key or an external hard drive, or vice versa.

[0012] In all these communication cases, transmission errors can occur. Computer data is conventionally defined as sequences of bits. When communicating data, one or more bits constituting this data can be modified by the transmission channel. In this case, there is a transmission error, and the data is no longer transmitted in its entirety and is therefore not communicated without error.

[0013] We have just spoken of bit sequences to define computer data. However, in general, we should rather speak of symbol sequences. Indeed, bits are symbols that come in pairs (0 and 1) but the transmissions actually carried out at the physical level rarely concern such symbols that come in pairs. The symbols transmitted at the physical level will be, for example, electromagnetic waves to which phase-shift modulations are applied in order to transmit one value among 4, 8 or 16 values. These techniques are used, for example, in mobile telecommunications. Phase-shift modulations are called phase-shift keying in English and we find the acronyms 4-PSK, 8-PSK or 16-PSK to describe phase-shift modulations that allow one value among 4, 8 or 16 values ​​to be transmitted in a single transmission.In fiber optic transmissions, the use of wavelengths also makes it possible to transmit symbols that can take one of many possible values, and not just symbols such as bits that take their values ​​from a pair of possible values. Even though the transmitted symbols themselves are not bits and can take more than two values, there is still a direct link between a transmitted symbol and bits. For example, in the case of 4-PSK modulation, the transmitted symbols can take four values, and these four values ​​correspond respectively to the bit pairs '00', '01', '10', and '11'.

[0014] We will therefore speak in this document of the general concept of a sequence of symbols which can be understood in a more simplified reading as being a sequence of bits.

[0015] These points are well known to those skilled in the art and will not be developed further. State of the art

[0016] The upcoming development of artificial intelligence promises a future explosion in the transmission of semantic data. Indeed, one can easily imagine interesting applications in which semantic data would be communicated from a device that produces it to another device that will use it.

[0017] For example, during a telephone communication, a speech recognition system coupled with an interpretation system could interpret each sentence spoken by a speaker into semantic data. The meaning of the sentence spoken by the speaker would then be associated with the semantic data. This semantic data would be communicated to another speaker who would have a reverse interpretation system coupled with a speech synthesis system. This system would make it possible to carry out an automatic translation of the telephone conversation, regardless of the languages ​​used by the speakers. In a conference call with speakers each speaking their own language, the speakers would hear each speaker's interventions translated on the fly into their own language.

[0018] To give another example, an artificial intelligence system can detect events such as an attack or a burglary on a video filmed by a video surveillance camera. A user may wish to have such a video surveillance service without the video stream filming his house being transmitted outside. Rather than transmitting the video, an artificial intelligence video surveillance system can produce semantic data and communicate it. A system trained to detect intrusions will then only need this semantic data to detect possible intrusions and the confidentiality of the video surveillance users will be preserved.

[0019] In the field of telecommunications, research is beginning to explore the concept of a semantic network that automatically adapts to the needs of applications that transmit data over the network using artificial intelligence techniques. Such semantic networks must rely on general representations of information that lend themselves to automatic adaptations and optimizations. These representations can carry semantic data of various kinds, such as images, text, or even configuration elements of the semantic network itself. A semantic network therefore relies on semantic data as an elementary data format.

[0020] These possible application examples show the general interest in transmitting semantic data (often called embeddings in English) in computer and / or telecommunications networks or, more generally, in interconnecting artificial intelligence systems in order to perform complex tasks.

[0021] More generally, semantic data is a multimodal representation. It is possible to represent all kinds of data (texts, images, videos, etc.) in the form of semantic data, knowing that the semantic contents of these multimodal representations are associated with the semantic data. Semantic data could therefore be used as a universal data format within telecommunications networks instead of the bit sequences used today. Such data would then present a finer granularity of information and more advanced optimization possibilities.

[0022] The interconnection of artificial intelligence applications or the deployment of semantic networks will involve an explosion in the quantity of semantic data to be communicated between devices. These communications of semantic data will correspond to the effective transmission of symbols between devices. The problem to be solved is therefore to be able to communicate semantic data from a transmitter to a receiver while preserving the semantic contents associated with the semantic data to be communicated while saving the number of symbols actually transmitted.

[0023] The scientific field of semantic communications is concerned with the problem of communicating only the minimal content of the data necessary to carry out a task and in particular the content necessary to make artificial intelligence systems communicate with each other. The prevailing approach in this field is to communicate only the data relevant to a targeted task in order to limit the memory size of the elements transmitted in terms of bits or bytes transmitted. In the example given above of video surveillance, a pre-transmission processing will seek to detect images (or extracts of images) suspected of representing an intrusion and only these extracts will be communicated, by removing images or semantic data from elements deemed useless to the video surveillance task. For example, color may be removed before transmission in this approach.

[0024] A major drawback of this approach is that the processing carried out before transmission depends on the task that is expected at reception. However, this task at reception cannot be known in all cases. Knowing this task at reception also implies that the telecommunications operator (which transmits the data and will often prepare them for transmission) must be informed of the tasks carried out by the applications which use its network. However, telecommunications operators are generally agnostic about the applications that use their networks. And it would also be very impractical to develop network optimizations specific to each expected processing task. For these reasons, such an approach could only be applied in the very specific case of a dedicated application between a transmitter and a receiver that would be connected to each other by a private network in which specific optimizations would be implemented to optimize the targeted artificial intelligence task. Such an approach will not be able to "scale up" and generalize to the countless uses expected in artificial intelligence and more generally to all uses in which we seek to adapt the transmitted data according to the processing that will be carried out by the receiving application.

[0025] The invention improves the situation. Statement of the invention

[0026] According to a first functional aspect, the invention relates to a method for managing the communication of semantic data, called data to be communicated, by a first device to a second device, characterized in that it comprises obtaining an index corresponding to the data to be communicated, said index making it possible to find in a data structure another semantic data, called communicated data, the index being chosen as a function of a distance measured between the data to be communicated and the communicated data, followed by the transmission of the index obtained through a communication channel between the first device and the second device.

[0027] Thanks to the invention, the number of symbols to be transmitted to communicate semantic data is greatly reduced. Indeed, instead of transmitting the symbols or bits encoding the semantic data as they are, the method transmits an index corresponding to an element of a data structure. This index makes it possible to find semantic data in the data structure. If we assume that the semantic data is a vector of dimension 4096 of real numbers, and that these real numbers are represented by 32-bit floating points, its size is approximately 16 kilobytes. If the data structure comprises 65,536 elements, each of these elements can have a different index if they are sequences of 16 bits (indeed, 2 to the power of 16 is equal to 65,536). Instead of transmitting 16 kilobytes, the method will therefore transmit 16 bits, or 2 bytes. The reduction in the size of the transmitted data is very spectacular here.

[0028] The data structure can be of any type as long as it allows semantic data to be retrieved from an index. It can be, for example, a key-value table in which the indexes play the role of keys and the values ​​are specific semantic data. We also speak of a dictionary. But the structure can also be a database. Or the structure can organize the indexes according to a prefix tree and the semantic data stored in the structure will then be leaves of the tree in question.

[0029] The method comprises obtaining an index from the semantic data to be communicated. The data structure comprises a finite number of elements while the possible space of the semantic data is very large, even infinite. For example, if the semantic data are vectors of real numbers of dimension 4096, they belong to the vector space / ?4096. In reality, the space of the semantic data is finite since it is computer data which are sets of floating numbers coded on 32 bits. However, it is reasonable to consider the possible space of the semantic data as infinite as a first approximation. There is therefore very little chance that the data to be communicated, which belongs to an infinite space, is present in the data structure which is a finite set, of reasonable size.

[0030] Instead of the data to be communicated, the method transmits an index which makes it possible to find another data item, which will therefore be the data item actually communicated. In order for the semantic content associated with the data item to be communicated to be preserved, the index obtained must make it possible to find in the data structure a semantic data item whose associated semantic content is very close, or even as close as possible, to that of the data item to be communicated. The index is therefore chosen according to a distance measured between the distance to be communicated and the distance communicated. In general, the index will be chosen to minimize the distance measured between the data item to be communicated and the data item actually communicated.

[0031] Two cases are then possible: • Or the index obtained is exactly an index associated with a semantic data item in the data structure. In this case, if we assume that the index is transmitted without transmission error, it will allow us to directly find a semantic data item in the data structure, which is then the communicated data item. • Or the calculated index is not an index associated with semantic data in the data structure. In this case, the reception will have to search in the data structure for an index close to the calculated index which will give access to semantic data.

[0032] It may be noted that the index obtained is not always directly associated with a semantic data item in the data structure but makes it possible to find a semantic data item in the data structure. In addition, transmission errors added by the communication channel may cause an index obtained which was directly associated in the data structure with a semantic data item to be transformed into an index that no longer directly corresponds to semantic data. It is therefore necessary in all cases to foresee at the reception that it may be necessary to search in the data structure for an index close to the received index that directly corresponds to semantic data in the data structure. The way in which the index is obtained must seek to ensure that the index obtained, to which transmission errors are added, makes it possible to find semantic data in the data structure according to the distance measured with the data to be communicated.

[0033] The data structure performs a quantification of the vector space in which the semantic data take values. Instead of an infinite number of possible values, a finite and restricted set (let's say it includes 65,536 elements) is defined which will give representatives for the data to be communicated. These representatives are chosen according to a distance measured between semantic data. Two choices are possible concerning the size of the indexes with respect to the number of elements present in the data structure: • Or the size of the indexes is such that the number of possible indexes corresponds exactly to the number of elements in the data structure. In the example above of 65,356 elements, a 16-bit index will give exactly this number. Each index obtained from a semantic data to be communicated will therefore give an index present in the data structure, associated with a communicated data, the index being chosen according to the distance between the data to be communicated and that communicated. This index will be transmitted, possibly with transmission errors, which will give another index which will also be present in the data structure. Assuming that these errors are few, the received index will make it possible to find in the data structure a received semantic data which will also be at a suitable distance from the communicated data and the data to be communicated. • Or the size of the indexes is larger than the size that gives a number of possible indexes equal to the number of elements in the data structure. In the example of 65,356 elements, we could choose a 20-bit index that takes more than a million possible values. In this case, there are many more possible indexes than values ​​in the data structure. The index obtained could be exactly the index of a data item present in the structure, or a close index. Transmission errors could cause the received index to drift from the transmitted index. But since there are many more possible index values ​​than elements present in the data structure, it is likely that the received index will not does not correspond to any element in the data structure but that the index closest to the received index is the one transmitted initially.

[0034] In other words, the length of the indexes and the size of the data structure may be such that the number of possible values ​​of the indexes is equal to the number of elements in the data structure or that the number of possible values ​​of the indexes is greater than the number of elements in the data structure. The transmission of the index will allow, upon reception, to find in the structure a semantic data item which will then be considered as being the communicated semantic data item. When the number of possible values ​​of the indexes is greater than the number of elements in the data structure and the received index is outside the data structure, it is the index closest to the received index which will allow to find a semantic data item which will be used.

[0035] In the first case, the advantage is to transmit an index of a minimal size compared to the number of elements present in the data structure. In the second case, the advantage is to be resistant to errors in transmitting the index.

[0036] The index is a sequence of symbols, belonging to a set of symbols. In some embodiments, these symbols are bits, that is to say they can take two possible values ​​generally represented by '0' and '1'. In other embodiments, these symbols correspond to quantities which will be those used in the physical transmission used by the method. For example, as explained above, the calculated index can be a sequence of symbols which correspond to phase change modulations among four possible phase changes, that is to say 4-PSK symbols. In this case, the indices would be sequences of values ​​chosen from the four possible 4-PSK values.A sequence of 8 4-PSK symbols will then take a value from among 65,536 possible (in fact, 4 to the power of 8 is worth 65,536) and a dictionary of 65,536 points can be indexed by indices of length 8 formed by 4-PSK symbols instead of indices of length 16 bits. Subsequently, we will most often choose examples in which the indices are sequences of bits but it is clear that these examples are only examples among others.

[0037] The transmission of the obtained index can be carried out by directly transmitting the sequence of symbols forming the index. In other embodiments, the sequence of symbols forming the obtained index is translated into a format suitable for transmission and then transmitted, which corresponds to the transmission of the index.

[0038] We have seen that the reduction in the size of the communicated data is very significant, since instead of transmitting the symbols or bits forming the data as they are, an index is transmitted which makes it possible to find semantic data. An advantage of the method is that this reduction is obtained without taking into account any subsequent processing which will be performed on the communicated semantic data. The transmission process is agnostic with respect to artificial intelligence or other applications using semantic data. This feature is particularly valuable for a telecommunications operator that wants to offer general data communication services to all kinds of possible customers.

[0039] The transmitted index makes it possible to find semantic data present in the data structure. The index is chosen according to a distance measured between this semantic data and the initial semantic data to be communicated. The distance between semantic data can be measured by a distance in the mathematical sense of the term on the set to which the communicated semantic data belongs. For example, the semantic data is a vector of 4096 real numbers and the distance will be measured by a distance on the vector set / j4096.If the semantic data is a vector of 1024 complexes, the distance will be measured by a distance on the vector set C1024- Since what is transmitted instead of the semantic data is an index corresponding to another semantic data, index chosen according to the distance between semantic data in the set to which the data belongs, the distortion caused by the transmission of the index and by the recovery of a data present in the data structure instead of the initial semantic data to be communicated itself, this distortion is minimized.

[0040] According to a first embodiment of this functional aspect of the invention, the distance measured between the data to be communicated and the communicated data is a function of the proximity between the semantic contents associated with the two semantic data to be communicated and communicated.

[0041] The choice of the distance used to measure the distance between semantic data is made in such a way as to correspond to the subsequent interpretation of the semantic data, that is to say that a minimum distance between semantic data corresponds to a minimum distance between meanings interpreted by users and applications which manipulate the semantic data. The measured distance is therefore a function of the proximity between the semantic contents associated with the semantic data or more precisely is calculated by a function preserving the proximity between the semantic contents associated with the semantic data. This function can be a distance in the mathematical sense of the term on the vector space in which the semantic data take values.The fact that the function which measures the distance between semantic data preserves the proximity between semantic content associated with the semantic data and that the index is chosen according to this distance measured between the semantic data to be communicated and that actually communicated will ensure that the semantic content associated with the semantic data to be communicated will not be too distorted by the distortion which consists of communicating. another semantic data, corresponding to the index obtained, in place of the data to be communicated.

[0042] The semantic content associated with semantic data is that which is present at the origin of the production of semantic data for later use. It can be the meaning of a text, an image, a video, a sound such that this content will be interpreted by users or artificial intelligence applications. Such semantic content can be said to be encoded in the semantic data, or present in the semantic data, or integrated in the semantic data, or represented by the semantic data. We use here the term associated with the semantic data.

[0043] According to a second embodiment of this functional aspect of the invention, which may be implemented alternatively or cumulatively with the previous mode, the index corresponding to the data to be communicated is obtained by searching in the data structure for the semantic data closest to the data to be transmitted according to the measured distance, and the transmitted index is the index of said closest semantic data.

[0044] With this embodiment, a simple way of calculating the index is provided. The semantic data present in the data structure has been chosen in advance to perform an efficient quantification of the space in which the semantic data to be transmitted is located. Indexes are associated with the semantic data present in the data structure and have been calculated beforehand. The semantic data present in the data structure closest to the semantic data to be transmitted is determined, the proximity between the semantic data being given by the measured distance between semantic data. In general, this measured distance is such that a minimal distance between semantic data corresponds to a minimal difference in subsequent semantic interpretations. In particular, the distance can be measured by a function preserving the proximity between the semantic contents associated with the semantic data.The index associated with the closest data is found in the data structure and it is this index that will be transmitted subsequently.

[0045] In this way, it is not necessary to calculate an index for each semantic data to be transmitted but simply to find in the data structure the semantic data closest to the data to be transmitted.

[0046] According to another embodiment of this functional aspect of the invention, which may be implemented alternatively or cumulatively with the preceding modes, the index is obtained by using an indexing function chosen in such a way that, for two distinct semantic data, the closer these semantic data are according to the distance measured between them, the closer the indexes calculated for these two semantic data by the indexing function are to each other.

[0047] Thanks to this embodiment, the method for communicating semantic data according to the invention resists possible transmission errors of the symbols which form the calculated index. Indeed, if one or more symbols, in small quantity, which form the index transmitted in place of the semantic data, are changed during transmission, the index finally received will still be close to the index initially transmitted because the number of transmission errors is generally low. Indeed, if the transmission errors become too large, the transmission stops. The point in the dictionary which corresponds to the received index will therefore be close to the point which corresponds to the index initially transmitted by virtue of the choice of the indexing function.The distortion due to the transmission error of one or more symbols of the index, which is added to the distortion of the replacement of the semantic data to be communicated by another semantic data present in the data structure, will therefore also be minimized.

[0048] It can be noted that if a received index, due to transmission errors, does not correspond to an index present in the data structure, the receiver of the index will look for an index close to the received index which corresponds to an element of the structure. This search can make it possible to return to the index which was initially transmitted provided that the number of transmission errors is low enough. And if the index found in this way is still different from the initial index, it still remains close to the initial index and the choice of the indexing function then ensures proximity between the semantic data corresponding to the received index and the semantic data corresponding to the initial index.

[0049] According to another embodiment, which may be implemented alternatively or cumulatively with the previous modes, the distance between semantic data is measured using cosine similarity.

[0050] Cosine similarity is a known quantity in the field of artificial intelligence to measure the similarity of semantic data between them. The cosine similarity of two vectors is calculated by dividing the scalar product of the two vectors by the product of the norms of the two vectors. This quantity is between -1 and 1, and will be -1 for opposite vectors, 0 for orthogonal vectors and 1 for vectors which have the same direction, and which are therefore equal to within a proportional factor.

[0051] It is possible to calculate a distance between vectors from the cosine similarity, for example by removing the cosine similarity from 1. Such a distance varies between 0 (for vectors having the same direction) and 2 (for opposite vectors). Other formulas for obtaining a distance from the cosine similarity are possible.

[0052] Such a distance between vectors, and therefore between semantic data, has the advantage of not taking into account the magnitude of the vectors but only their direction. Such a distance is suitable for calculating semantic proximity between semantic data. It preserves the proximity between the semantic contents associated with the semantic data whose distance is measured with a cosine similarity.

[0053] In other embodiments, the distance between two semantic data is measured with the Euclidean distance.

[0054] The well-known Euclidean distance in vector spaces provides an alternative implementation compared to distances using cosine similarity. Proximity according to Euclidean distance corresponds to a certain proximity between semantic contents associated with semantic data.

[0055] According to another embodiment, which may be implemented alternatively or cumulatively with the previous modes, the proximity between two indexes is defined by the Hamming distance.

[0056] The Hamming distance is a defined distance between bit sequences that is calculated by counting the number of distinct bits at a given position between two bit sequences. For example, the Hamming distance between byte 00000000 and byte 00010001 is two, because, at the fourth and eighth positions, the bits are different. The notion of Hamming distance can be extended to apply to sequences of symbols as are the calculated indices. This distance provides a practical and efficient way to calculate the proximity between the calculated indices used in the method. The proximity between indices measured according to the Hamming distance corresponds well to the notion of seeking to minimize the distortion caused by transmission errors. Indeed, it is directly associated with the number of transmission errors suffered by a sequence of bits or symbols in a communication channel.

[0057] According to another embodiment, which may be implemented alternatively or cumulatively with the preceding modes, the indexing function is chosen in such a way that, for two distinct semantic data, if the distance measured between these two semantic data is less than a given threshold, then the probability that the indexes calculated by the indexing function for these two semantic data are the same is greater than a given probability.

[0058] Thanks to this embodiment, the link between the distance measured between two semantic data and the proximity between the indexes calculated for these two data is refined. In a bubble around a semantic data, that is to say for the set of semantic data which are at a distance lower than a threshold from the initial data, the probability that these data have the same calculated index is greater than a given probability. In this way, the closer the data are to each other, the closer their indexes are. The distances between semantic data and proximities between Indexes can be calculated using cosine similarities on the one hand, Hamming distances on the other hand, or any other suitable method of calculating distances between semantic data and indexes.

[0059] According to another embodiment, which may be implemented alternatively or cumulatively with the preceding embodiments, the indexing function is a locality-sensitive hash function.

[0060] A locality sensitive hash function (for which the acronym LSH can be used) is a function, which belongs to a family of functions, which calculates indexes for points of a given vector space and which verifies the following two properties, namely that, on the one hand, if the distance between two points is less than a threshold, the probability that the indexes calculated by this function for the two points are equal is greater than a first given probability and that, on the other hand, if the distance between two points is greater than another threshold, a multiple of the first, the probability that the indexes calculated for the two points are equal is less than a second given probability, much lower than the first.Functions of this type which index the points of a given vector space respond well to what is desirable for the invention, namely that close semantic data will have close, or even confused, indexes, and that distant semantic data will on the contrary have distant indexes, knowing that the semantic data are points of a given vector space.

[0061] According to another embodiment, which may be implemented alternatively or cumulatively with the preceding embodiments, the indexing function is a locality-sensitive hash function for cosine similarity.

[0062] In this embodiment, the family of locality-sensitive hash functions is determined for a given distance between semantic data and this distance uses cosine similarity. The advantage of this mode is to perform indexing adapted to this distance which is adapted to the proximity of the semantic contents associated with the semantic data.

[0063] According to another embodiment, which may be implemented alternatively or cumulatively with the preceding embodiments, the method comprises a prior step of fragmenting the semantic data to be communicated into several fragments and the subsequent steps of the method apply to said several fragments.

[0064] Until now, the method according to the invention processed a semantic data item which is a point in a high-dimensional vector space (up to 4096, or even more) in a single block. But it may be advantageous to carry out a fragmentation of the semantic data item to be communicated and to apply the different steps of the method to fragments of data. For example, a semantic data item which is a vector of 4096 real could be divided into 1024 fragments of dimension 4 or into 512 fragments of dimension 8.

[0065] A first advantage is to facilitate the exchange of data structures in certain embodiments of the method. Indeed, the data structure must be known in transmission and reception and, in certain cases, it will be transmitted before the implementation of the method. If we take our example of a data structure comprising 65,536 elements, and the elements are semantic data in the R4096 vector space, then each element has an approximate size of 16 kilobytes, and the data structure has an approximate size of 1 gigabyte. The exchange of such a data structure between transmitter and receiver becomes prohibitive. If we fragment the semantic data into fragments of dimension 4, and keep the same size for the data structure, the 65,536 elements will only weigh 4 times 32 bits (i.e. 4 reals encoded in float32\ i.e. the structure will have an approximate size of 128 bits times 65,536, or 1 megabyte.The data structure that has to be stored and possibly exchanged has therefore decreased in size by a factor of 1000.

[0066] Furthermore, fragmentation makes it possible to improve the precision of the communicated data, although it is true at the expense of the number of symbols or bits transmitted forming the indexes. If we take the example of a semantic data item belonging to the R4096 vector space fragmented into 1024 fragments of dimension 4, and the data structure used has 65,536 elements of dimension 4, these elements will be used to quantify fragments of dimension 4 and no longer a complete data item of dimension 4096. The communicated data item is then much more precise, at the expense of the memory size required for the transmission of the indexes, since it will be necessary to transmit 1024 indexes of size 16 bits and no longer a single index of size 16 bits. The transmission gain of a 16-bit index instead of a 4-dimensional semantic data (i.e. 4 reals encoded on 32 bits, therefore 128 bits) is therefore only a factor of 8.Fragmentation therefore allows a choice between very economical communication (a 16-bit index instead of data of approximately 16 kilobytes, i.e. a gain factor of 1024) or much more precise communication (65,536 elements to cover a space of dimension 4 instead of a space of dimension 4096) with a gain factor of only 8.

[0067] It is possible to mix these effects. While a data structure of size 65,536 elements can be seen as appropriate for data in the R4096 vector space, fragmentation in a 4-dimensional space can allow data structures of size 1024. The indexes for this structure size will then be 10 bits long. The gain factor compared to 4 reals encoded on 32 bits is then greater than 12. The precision given by 1024 elements can be quite sufficient to represent 4-dimensional semantic data fragments.

[0068] These digital examples are of course only embodiments among others.

[0069] According to an embodiment which can be implemented cumulatively with the previous embodiment, the size of the fragments into which the semantic data to be communicated is fragmented is chosen as a function of the expected size of the data packets which can be transmitted through the communication channel between the first device and the second device.

[0070] In this embodiment, fragmentation makes it possible to adapt the size of the transmitted indexes to the expected size of the packets that can be transmitted by the network. It is indeed advantageous not to divide a piece of data into several packets. The size of the packets that will be transmitted through the communication channel is called the maximum transmission unit (MTU). Depending on the expected MTU on the transmission, a given fragmentation can be defined which will prevent the transmitted indexes from being divided into several packets.

[0071] According to another embodiment, which may be implemented alternatively or cumulatively with the previous embodiments, the data structure is defined beforehand according to the format of the semantic data to be communicated.

[0072] The operation of the method implies that the data structure is available at reception to find semantic data corresponding to the received index, which will be close to the data to be communicated by virtue of the distance measured between semantic data and the chosen indexing function. It is possible to define in advance a data structure comprising semantic data representative of the semantic data to be communicated. These representative semantic data will have the same format as the semantic data to be communicated and will for example be of the same dimension.

[0073] One way to proceed is to define in advance a grid of points (the semantic data present in the data structure) which appropriately covers the vector space in which the semantic data to be communicated will take values. This grid may be denser at the locations of the vector space in which semantic data to be communicated are expected in order to ensure a finer representation in the locations of interest. It can be noted that, in general, the semantic data used or produced by artificial intelligence applications have a reduced magnitude; the value of the real numbers composing a semantic data is most often between -100 and 100. It will therefore be advantageous for the semantic data composing the data structure to be predominantly around 0, with possibly larger values ​​to capture certain exceptional data.

[0074] Such a data structure can therefore be defined in advance for a given semantic data format. Applications handling semantic data of a given format can then have in memory the data structure corresponding to this precise format. The method can then be used for any semantic data communication task, using the data structure associated with the format, and the receiving applications will be able to have information regarding the semantic data format which will allow them to find the similar semantic data associated with a received index, these similar semantic data being present in the data structure.

[0075] This data structure defined for a precise format of semantic data can also be used to achieve efficient storage of semantic data. An index is calculated for a semantic data to be stored, which corresponds in the data structure to a semantic data close to the semantic data to be stored. It is this index which is saved and which, upon subsequent reading, will give access to the semantic data close to the semantic data to be stored. Possible errors in saving the index will be minimized, as for transmission errors, by the nature of the chosen indexing function. And the distortion between the semantic data to be stored and that corresponding to the index actually stored will also be minimized by the choice of the distance measured between semantic data which preserves the proximity between semantic contents associated with the semantic data.

[0076] According to another embodiment, which may be implemented alternatively or cumulatively with the previous embodiments, the data structure is obtained by automatic learning based on the semantic data already communicated and / or intended to be communicated.

[0077] In this embodiment, the data structure is learned, by a machine learning method, in order to best adapt to the set of semantic data already communicated or intended to be communicated. Several learning techniques can be used, for example supervised or not, or using or not using a reward function in the case of reinforcement learning. A machine learning technique will make it possible to have a greater density of representative semantic data in the portion of the vector space in which a greater number of semantic data to be communicated is expected.For the same size of the data structure, this embodiment will therefore have the advantage of greater proximity of the data communicated with the initial data to be communicated since the grid of communicated data is denser where it will be most needed, whereas a generic data structure, defined in advance, will be less precise.

[0078] According to another embodiment, which may be implemented alternatively or cumulatively with the previous embodiments, the automatic learning of the data structure uses the k-means method where k is the number of semantic data present in the data structure.

[0079] When the embodiment of the method comprises automatic learning of the data structure, the objective of the automatic learning is to carry out a clustering of a set of data communicated or to be communicated, called a training set, around a reduced number of representative semantic data which will be recorded in the data structure. Such a clustering can be carried out using the k-means method, where k is the size of the data structure. Thanks to this embodiment, the clustering is uniform, in the sense that the data belonging to the training set are well distributed in an egalitarian manner into groups whose center is a representative semantic data which will be found in the data structure to represent all the semantic data of the same clustering.

[0080] According to another embodiment, which may be implemented alternatively or cumulatively with the previous modes, the indexing function is defined beforehand according to the format of the semantic data to be communicated.

[0081] In this embodiment, similarly to what was seen previously for the data structure, it is the indexing function which is defined beforehand according to the format of the semantic data to be communicated. For a given vector space in which the semantic data to be communicated will take values, an indexing function having the expected properties can be defined once and for all. This function can then be used by default for the semantic data of the expected format. This embodiment has the advantage, when used in common with the similar mode relating to the data structure, of allowing efficient communications, without prior processing, of the semantic data of a given format. These communications can also correspond to backups as we have already seen.

[0082] According to another embodiment, which may be implemented alternatively or cumulatively with the previous embodiments, the indexing function is obtained by automatic learning based on the semantic data already communicated and / or intended to be communicated.

[0083] This embodiment has the advantage of allowing the use of an indexing function particularly suited to the semantic data to be communicated. For example, as for learning the data structure, the learned indexing function may be more discriminating in the areas of the vector space where the density of semantic data to be communicated is higher.

[0084] Some functions are particularly suited to machine learning. This is the case of locality-sensitive hash functions (LSH functions). These functions are defined according to hyperplanes that divide the vector space in which the semantic data take values ​​into two parts. One part, located on one side of the hyperplane, will correspond to the value 0 for a bit of given rank in the calculated index while the other part, located on the other side, will correspond to the value 1 for the bit of the same rank. For a given index length L, an equal number of hyperplanes is fixed that will divide the vector space into 2 to the power of L, if we consider that the index is a sequence of bits. Similar calculations can be done when the index is a sequence of other symbols, by reducing these other symbols to bits.

[0085] The placement of these hyperplanes, which define an LSH indexing function, can be obtained by learning according to the density of the data to be transmitted in the vector space, in order to produce the most suitable cuts.

[0086] According to another embodiment, which may be implemented alternatively or cumulatively with the preceding embodiments, the method comprises the transmission of parameter values ​​useful for the communication of semantic data according to the method.

[0087] The method according to the invention operates by assuming that the entity receiving the transmitted index knows at least the data structure to find semantic data corresponding to the received index. Other parameter values ​​useful for the operation of the method may need to be known by the receiving entity, for example the size of the fragments if the semantic data has been divided into several fragments before applying the method.

[0088] For example, when the data structure and the indexing function are defined beforehand, as has been seen in several embodiments, the transmitter simply needs to place in its transmission of the index obtained references to the data structure and to the function that it will use. These references can be placed for example in a preamble packet to the transmission or in a data packet header. The receiver will receive these parameter values ​​which refer to known indexing functions or data structures. It will then know which data structure to use to find the semantic data close to the data to be transmitted thanks to the index received. Such data structures and sets of parameters could be defined beforehand and standardized for semantic data formats used in a generalized manner, such as those presented in the MTEB comparison.The communication of semantic data according to the method would then include in this embodiment the transmission of information of. reference allowing the equipment receiving the index associated with the communicated semantic data to know which parameters are used for the communication.

[0089] According to another embodiment, which may be implemented alternatively or cumulatively with the preceding embodiments, the method comprises a preliminary phase of negotiating parameter values ​​useful for the communication of semantic data according to the method, said parameters being able to be chosen from the following non-exhaustive list: 1. Dimension of semantic data; 2. Dimension of a fragment of semantic data; 3. Data type used as the base type by a semantic data; 4. Format of the semantic data; 5. Identifier of the model producing the semantic data; 6. Dictionary size; 7. Contents of the dictionary; 8. Size of dictionary indexes; 9. Symbols used by the indexes; 10. Indexing function used.

[0090] Thanks to this embodiment, more advanced possibilities of communication between the transmitter and the receiver are implemented. We saw in the previous embodiment that values ​​useful for the execution of the parameter could be transmitted, for example in a packet header, and make it possible to find elements already defined, such as the data structure used in reception to find the semantic data close to the data to be communicated. But this previous mode is not sufficient when, for example, the data structure and / or the indexing function are learned by machine learning in a specialized way for data to be communicated. The receiver cannot in this case know a priori the elements necessary for the operation of the method.

[0091] In this embodiment, a negotiation phase prior to communication allows the transmitter and the receiver to agree on all the parameters useful for the communication of semantic data according to the method. This mode can be combined with the previous one: for example, a first transmission can give the references to a codec that will be used, then a negotiation allows the transmitter and the receiver to agree on a fragmentation, which can depend on the MTU expected on the communication channel between the first device and the second device. This embodiment proposes a non-exhaustive list of the elements that can be negotiated during this preliminary negotiation step.

[0092] According to a second functional aspect, the invention relates to a method for managing the reception of semantic data, characterized in that it comprises the reception of an index, called the received index, followed by reading semantic data in a data structure based on the received index.

[0093] This second functional aspect of the invention corresponds to the reception of the transmitted index in place of the semantic data to be communicated. We speak here of the received index, because it is possible that transmission errors have modified the index during transmission and that the received index is not equal to the transmitted index. Once the index is received, it is possible to find in the data structure a semantic data from the received index. This read semantic data will be considered as the semantic data to be communicated initially.

[0094] According to a first embodiment of this second functional aspect, the method comprises, if the received index does not correspond to any index present in the data structure, a search in the data structure for the index closest to the received index and the semantic data read in the data structure is that which corresponds to said closest index.

[0095] According to another embodiment of this second functional aspect, which may be implemented alternatively or cumulatively with the previous modes, the proximity between the received index and the indexes present in the data structure is defined by the Hamming distance.

[0096] As already indicated previously, transmission errors may cause the received index to be different from the transmitted index and in particular, the received index to not be present in the data structure. In this case, an advantageous embodiment is to search for the closest index that is present in the data structure in order to finally obtain semantic data.

[0097] To summarize, the method according to the invention can accumulate the following distortions in the transmission of semantic data: • In general, the semantic data to be communicated is not present in the data structure. The calculated index will correspond directly or indirectly to a similar semantic data present in the data structure, called communicated semantic data. • In some cases, there are as many semantic data in the data structure as there are possible index values. The calculated index will in this case be directly associated with a semantic data. • In other cases, there is less semantic data in the data structure than possible index values. Then, the calculated index is not necessarily directly associated with a semantic data present in the data structure. But it is possible to find in the data structure the index closest to the calculated index. A semantic data associated with this closest index can then be found in the data structure. • In all cases, the calculated index therefore corresponds directly or indirectly to a semantic data present in the data structure. The indexing function and the proximity between indexes are defined so that this transmitted semantic data is close to the semantic data to be transmitted according to a measured distance which preserves the semantic contents associated with the semantic data, but this is still a first distortion. • The calculated index is transmitted. It may be modified by the transmission channel and the received index may be different from the transmitted index. • In some cases, there is as much semantic data in the data structure as there are possible index values. The index received will in this case be directly associated with a semantic data item, which may be different from the communicated semantic data item. This is a second distortion. • In other cases, there is less semantic data in the data structure than possible index values. Then, the received index is not necessarily directly associated with a semantic data present in the data structure. But it is possible to find in the data structure the index closest to the received index. This closest index can be the transmitted index: in this case, we fall back on the index that we wanted to transmit and the distortion due to transmission errors is canceled. Or the closest index is different from the transmitted index: in this case, another semantic data than the communicated data is associated, and here we have a second distortion.

[0098] The method according to the invention must therefore make it possible to carefully choose the indexing function, the proximity between indexes, the fact that the distance measured between semantic data preserves the semantic contents associated with the semantic data, the number of representative data present in the data structure, the concentration of these representative data in the areas richest in data to be communicated of the vector space in which the semantic data take values ​​and any other useful parameter in order to minimize these distortions. As indicated, the choice of cosine similarity, Hamming distance and locality-sensitive hash functions for cosine similarity are examples of operating parameters of the method according to the invention which give first good results in our experiments, but the method according to the invention is not limited to these examples of embodiments and covers the use of many other possible parameters.

[0099] According to another embodiment of the second functional aspect of the invention, which may be implemented alternatively or cumulatively with the preceding modes, the method comprises, if the index received corresponds to a fragment of semantic data, the reconstitution of complete semantic data from the fragments of semantic data corresponding to the indexes received.

[0100] Thanks to this embodiment, the method according to the invention makes it possible to manage the fragmentation of the communicated semantic data. It is simply necessary to ensure that the complete semantic data is reconstituted from the fragments of semantic data corresponding to the different indexes received. This reconstitution makes it possible to complete the operation of the method according to the invention.

[0101] According to a first material aspect, the invention relates to a management entity capable of managing the transmission of semantic data, called data to be communicated, by a first device to a second device, characterized in that it comprises the following modules: • A module for obtaining an index corresponding to the data to be communicated, said index making it possible to find in a data structure another semantic data, called communicated data, the index being chosen according to a distance measured between the data to be communicated and the communicated data; • A module for transmitting the index obtained through a communication channel between the first device and the second device.

[0102] According to a first embodiment of this hardware aspect, the invention relates to telecommunications equipment comprising a management entity capable of managing the communication of semantic data according to the invention.

[0103] Such telecommunications equipment may be, for example, a router, or a home gateway or one in a professional environment, or a network card present in computer equipment such as a computer, or a mobile terminal, or a communicating object such as a camera or other object producing semantic data for artificial intelligence applications or not. In general, a management entity according to the invention may be embedded in any type of object or equipment producing semantic data intended for a remote receiver or having to carry out the communication of such data intended for other equipment in a computer or telecommunications network.

[0104] According to a second material aspect, the invention relates to a management entity capable of managing the reception of semantic data, characterized in that it comprises the following modules: • A module for receiving an index, called the received index; • A module for reading semantic data in a structure of data based on the received index.

[0105] According to a first embodiment of this hardware aspect, the invention relates to telecommunications equipment comprising a management entity capable of managing the reception of semantic data according to the invention.

[0106] Such telecommunications equipment may be, for example, a router, or a home gateway or one in a professional environment, or a network card present in computer equipment such as a computer, or a mobile terminal, or equipment managing the reception of data in a cloud computing system. In general, a management entity according to the invention may be embedded in any type of object or equipment receiving semantic data for backup or processing by artificial intelligence or other applications or having to carry out the reception of such data in a computer or telecommunications network.

[0107] In embodiments, the management entities managing the communication and reception of semantic data are integrated into a single device. In some of these embodiments, the devices in question may be routers performing both communication to other devices and reception from other devices of semantic data. In other embodiments, the management entities managing the communication and reception correspond to separate components of the same device. The management entity managing the communication may correspond to a component producing semantic data and the management entity managing the reception to a component for storing and reading the produced semantic data. The transmission channel between the devices in this case is a software bus which connects the two components of the same device together.

[0108] According to another material aspect, the invention relates to a computer program capable of being implemented by a management entity, the program comprising code instructions which, when executed by a processor, carries out the steps of the method for managing the communication of semantic data defined above.

[0109] According to another material aspect, the invention relates to a computer program capable of being implemented by a management entity, the program comprising code instructions which, when executed by a processor, carries out the steps of the method for managing the reception of semantic data defined above.

[0110] Finally, according to another material aspect, the invention relates to data media on which are recorded computer programs comprising sequences of instructions for implementing the methods for managing the transmission and reception of semantic data defined above.

[0111] The data carriers may be any entity or device capable of storing the programs. For example, the carriers may comprise a storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or a magnetic recording means such as a hard disk. Furthermore, the carriers may be transmissible media such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The programs according to the invention may in particular be downloaded from a network such as the Internet. Alternatively, the information carrier may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.The program according to the invention can use any type of computer technology in terms of compiled, interpreted programming languages, or a combination of both as well as in terms of operating systems. Brief description of the figures.

[0112] The invention will be better understood on reading the following description, given by way of example, and made with reference to the appended drawings in which:

[0113] [Fig. 1] represents a management entity managing the transmission and a management entity managing the reception of semantic data, included in two devices, illustrating an exemplary embodiment of the invention.

[0114] [Fig.2] illustrates an example of steps implemented within the framework of an embodiment of the invention.

[0115] [Fig.3] illustrates examples of possible distortions undergone by semantic data during its transmission and reception.

[0116] [Fig.4] illustrates experimental results obtained during the development of the methods according to the invention. Detailed description

[0117] [Fig. 1] represents a management entity 100 managing the communication of a semantic data SD from a device EQ1 to a device EQ2, through a communication channel CNL. A management entity 200 manages the reception of a semantic data SD'.

[0118] In the exemplary embodiment presented in [Fig.l], the management entity 100 comprises the following modules: • A module 101 for obtaining an index I' corresponding to the data to be communicated SD, said index I' making it possible to find in a data structure DS another semantic data SD', called communicated data, the index I' being chosen as a function of a distance D measured between the data to be communicated SD and the communicated data SD'; • A module 102 for transmitting the index I' obtained through a CNL communication channel between the first device EQ1 and the second device EQ2.

[0119] In the exemplary embodiment presented in [Fig.l], the index I' is obtained from the semantic data SD by using at one point an indexing function F.

[0120] In the exemplary embodiment presented in [Fig.l], the management entity 200 comprises the following modules: • A module 201 for receiving an index I', called the received index; • A module 202 for reading semantic data SD' in a structure of DS data, depending on the received index I'.

[0121] Instead of directly transmitting the symbols forming the semantic data SD, the management entity 100 obtains an index F and transmits it, which greatly saves the size of data to be transmitted.

[0122] The transmission of the symbols forming the index I' is done through a CNL communication channel. Transmission errors can occur during transmission. However, in the example presented in [Fig.l], the received index I' is equal to the calculated index F. Examples in which transmission errors appear will be seen in [Fig.3]. The management entity 200 can read in a data structure DS a semantic data item SD' corresponding to the received index I'. This semantic data item SD' will then be used in subsequent artificial intelligence or other applications in place of the data item SD to be transmitted. The semantic data item SD' was therefore communicated from the equipment EQ1 to the equipment EQ2 in place of the semantic data item SD which was the semantic data item to be communicated.

[0123] In the example presented in [Fig.l], the only distortion caused by the use of the method according to the invention corresponds to the use of the calculated index F, associated with the semantic data SD', present in the data structure DS. We present in [Fig.3] all the distortions that can be introduced by the method according to the invention.

[0124] The management entity 100 included in the equipment EQ1 has the hardware architecture of a conventional computer. It includes in particular a processor, a RAM type random access memory and a read-only memory such as a Flash, ROM type memory (not shown in the figure) as well as input-output devices such as, in certain cases, keyboards and / or screens (not shown in the figure), and ports networks allowing communication with other entities and servers through a communication network such as the Internet not shown in the figure.

[0125] The management entity 200 included in the equipment EQ2 also has the hardware architecture of a conventional computer. It includes in particular a processor, a RAM type random access memory and a read-only memory such as a Flash, ROM type memory (not shown in the figure) as well as input-output devices such as, in certain cases, keyboards and / or screens (not shown in the figure), and network ports allowing communication with other entities and servers using a communication network such as the Internet, not shown in the figure.

[0126] The management entities 100 and 200 may be included, for example, in telecommunications equipment. Such equipment EQ1, EQ2 may be routers, home gateways or deployed in professional environments. The management entities 100 and 200 may be included in network cards present in computer equipment EQ1, EQ2 such as computers or equipment EQ1, EQ2 managing the reception or transmission of data in cloud computing systems. The management entities 100 and 200 may also be included in mobile terminals EQ1, EQ2 where they will be used respectively for the communication or reception of semantic data through 3G, 4G, 5G, 6G or other wireless telecommunications networks to which the mobile terminals EQ1, EQ2 are connected.In general, a management entity 100 or 200 according to the invention can be embedded in any type of object or equipment EQ1, EQ2 producing semantic data SD for saving or processing by artificial intelligence or other applications or having to carry out the reception of such data in a computer or telecommunications network.

[0127] In another technical field, the management entities 100 and 200 can be included in image and sound sensors and in all types of equipment EQ1, EQ2, for example means of transport or in robotic equipment EQ1, EQ2. Image or sound sensors can encode semantic data from the captured images or sounds. Such sensors (cameras, microphones) can include a management entity 100 for transmitting semantic data. Management entities 200 can be deployed in processing servers which will receive the transmitted semantic data and act on them. These management entities 200 can for example be found in servers embedded in means of transport where they will be used for the interpretation of data transmitted by the sensors via the management entities 100 for the transmission of semantic data.This organization of management entities 100 and 200 can be useful in . the framework of autonomous means of transport, or robots, in which sensors process captured images and sounds, produce semantic data, and this semantic data is sent for interpretation and consequent action to control servers. In this framework, the management entities 100 and 200 can be included in the same equipment.

[0128] In another domain, the management entities 100 and 200 can belong to the same equipment and manage the storage of semantic data. In this context, the management entity 100 obtains an index I' from a semantic data item SD to be stored and it is this index I' which is stored. The management entity 200 retrieves the stored index I' and reads the corresponding semantic data item SD' in the data structure DS. The semantic data item SD' is therefore ultimately the data item which has been stored in place of the semantic data item SD to be stored.

[0129] The CNL communication channel is a link between the management entities 100 and 200. The CNL channel may be a segment of a telecommunications network or, on the contrary, correspond to the chaining of several communication networks making it possible to connect the management entity 100 to the management entity 200. The CNL channel may also be a link inside a piece of equipment (autonomous means of transport, robot, computer equipment in general) between sensors (cameras, microphones) and processing servers or between a component producing semantic data and storing them and a component retrieving them. In these cases, the CNL channel may be a data bus inside a piece of computer equipment. The communications technologies deployed in the CNL channel may be wired, optical, satellite or wireless technologies or a combination of these different technologies.

[0130] The CNL channel can provide a direct link between the management entity 100 and the management entity 200. For example, the management entity 100 can be included in a network card of a personal computer; the management entity 200 can be included in a home gateway deploying a local area network with WiFi technology; and the CNL channel can be in this case the WiFi link between the computer including the management entity and the gateway including the management entity 200. Or, the management entity 100 can be included in a mobile phone using wireless telephony of 3G, 4G, 5G, 6G or other technology and the management entity can be included in a base station deploying the corresponding technology and the CNL channel will be the wireless link between the mobile phone including the management entity 100 and the base station including the management entity 200.Or the management entity 100 is associated with a camera, a microphone or a radar or lidar system and addresses semantic data to the management entity 200 included in a processing server embedded in an autonomous vehicle or a robot via an internal coaxial cable or optical fiber link, which. forms the CNL channel. Or the CNL channel is a software or data bus connecting components of computer equipment.

[0131] In other examples, the CNL channel will provide an indirect link between the management entity 100 and the management entity 200. For example, the management entities 100 and 200 may be included in routers EQ1, EQ2 present in packet-switched networks, which may be, for example, IP routers (acronym for Internet Protocol). The CNL channel may, in this example, be a path according to the IP protocol between the two routers EQ1, EQ2 which include the management entities 100 and 200. In this example embodiment, the CNL channel may therefore use several IP network segments and several IP routers and will then be an indirect link between the two management entities 100 and 200.

[0132] The CNL channel can be seen, in embodiments, as transmitting sequences of bits, that is to say sequences of binary symbols 0 or 1 between the two devices EQ1, EQ2 including the two management entities 100 and 200. But at the level of physical transmission, the CNL channel will transmit physical symbols which can take more values ​​than just two values. For example, in the field of wireless communications, the symbols transmitted by the CNL channel can be phase modulations carried by a radio wave, and these phase modulations can take up to sixteen or more distinct values. Similarly, in the field of links by optical fiber or copper support, the symbols transmitted by the CNL channel will be variations of electromagnetic waves carried by the physical supports.

[0133] The calculated index I' which is transmitted by the management entity 100 through the CNL channel will therefore be in certain embodiments a sequence of bits 0 or 1 but will be, in other embodiments, a sequence of symbols corresponding to the symbols which can be transmitted in the CNL channel.

[0134] The CNL channel introduces errors into the transmission. The received index I' is therefore, in general, different from the calculated index I' which was transmitted. In the example presented in [Fig.l], these two indexes are identical.

[0135] The management entity 200 has a data structure DS which makes it possible to match the received index I' with a semantic data SD'. This semantic data SD' is then made available by the management entity 200 and used by artificial intelligence applications or others in place of the initial semantic data SD. Instead of communicating the complete semantic data SD by transmitting all the symbols or bits constituting it, the management entity 100 only had to transmit the index I', which represents a significant saving in the size of the data transmitted.

[0136] The data structure DS can be of any type. It can be, for example, a key-value table, or a dictionary, associating indexes I, I' with respective semantic data SD, SD'. The data structure DS can also be a database where the indexes I, I' are associated with the respective semantic data SD, SD'. The data structure DS can be saved in RAM or ROM by the management entity 200 or else queried on demand upon receipt of an index F to retrieve the associated semantic data SD'.

[0137] Semantic data SD, SD' are data obtained by transforming data such as texts or images into a model. The semantic content of the textual or image data is then associated with the semantic data obtained. The fact that semantic data SD, SD' preserve the semantic content of initial data associated with them makes them particularly suitable for generative or other artificial intelligence applications. Semantic data SD, SD' generally follow a given format that will allow them to encode, integrate or represent more or less semantic information. Possible formats are listed on the MTEB site (acronym for Massive Text Embedding Benchmark) for semantic data SD, SD' that represent texts and their performance in different artificial intelligence tasks is presented.A powerful format is for example the Mistral-7B format, from the company Mistral AL. The semantic data SD, SD' in this format are vectors of 4096 reals, that is to say elements of the vector space R4096. Equivalently, we can say that the semantic data SD, SD' are points belonging to the Cartesian space of the same dimension / J4096. Other formats can be used, for example of dimension 1024. .

[0138] The size of an SD semantic data item according to the Mistral-7B format is therefore 4096 times the 32 bits used to encode a real number in the float32 format. If the communication of an SD semantic data item is replaced by the transmission of an index of length 32 bits, the gain in size of transmitted data is then a factor of 4096. Other smaller or larger gains can be obtained depending on the indexing choices made during the implementation of a possible embodiment of the invention from a whole set of possible embodiments.

[0139] The transmission of an index I' instead of the transmission of the symbols or bits forming the semantic data SD, then the transmission errors caused by the CNL channel, and finally the reading of a data SD' corresponding to the received index I' involve distortions with respect to the semantic data to be communicated SD. These distortions are presented in more detail in [Fig.3]. In the example of [Fig.l], an indexing function F is chosen so as to minimize these distortions according to a distance D measured between semantic data. In [Fig.l], this is represented by a minimum distance D between the semantic data SD and the semantic data SD'.

[0140] The distance D is measured by a function measuring the proximity between two semantic data SD and SD'. This function can be a distance, in the mathematical sense of the term, in the vector space to which the semantic data SD, SD' belong. In embodiments, the distance D can be for example the Euclidean distance between points of the Cartesian space associated with the vector space of the semantic data SD, SD'. In other embodiments, the distance D can be obtained by calculating the cosine similarity between vectors of the vector space of the semantic data SD, SD'. Other distances can be defined which will be specific to the subsequent tasks which will use the semantic data SD, SD' so that the proximity between semantic data SD, SD' measured by the distance D is in direct relation with the proximity of the semantic contents associated with the data themselves in the subsequent task which uses them.

[0141] In embodiments, the distance D can be learned from a training set representative of the semantic data SD, SD' which will have to be transmitted. Indeed, if the semantic data SD to be transmitted are rather concentrated in a precise zone of the vector space to which they belong, it may be relevant for the function measuring the distance D to be more discriminating in this zone, to take into account differences between data concentrated in this zone, and less discriminating outside this zone since it is expected to find less semantic data SD there.

[0142] We recall that the index I' is chosen as a function of a distance D measured between the semantic data to be communicated SD and that communicated SD' and that, in embodiments, this measured distance D is calculated by a function preserving the semantic proximity between the semantic contents associated with the two semantic data SD, SD'.

[0143] In the example shown in [Fig.l], the indexing function F is chosen so as to ensure minimal distortion between the semantic data to be transmitted SD and the transmitted data SD'. Furthermore, in embodiments, the indexing function F is chosen in such a way that for two distinct semantic data SD, SD', the closer these semantic data SD, SD' are to each other according to the distance D measured between them, the closer the calculated indices I, I' for these two semantic data SD, SD' are to each other. The proximity between calculated indices can be determined by the Hamming distance or by another appropriate calculation method. An advantage of this choice is to ensure a certain robustness of the method in the presence of transmission errors in the CNL channel which affect the transmitted index.

[0144] The function F calculates an index of size limited to a maximum of a few bits (or symbols in the case where the CNL channel transmits different symbols on its physical layer) from a semantic data item formed of several thousand real numbers. For example, we will use indexes of length 10 bits, which gives 1024 possible index values, or 16 bits, which gives 65,536 possible index values. These indexes are associated in a DS data structure with semantic data SD, SD'. Given the size in number of bits of the semantic data (16 kilobytes for a SD semantic data item of size 4096 real numbers encoded cn floal 32), if the DS data structure includes 65,536 data items associated with all the possible values ​​taken by an index of length 16 bits, it will have a size of 1 gigabyte, which could be seen as a maximum size.

[0145] It is therefore clear that the number of possible values ​​for the indexes is very small compared to the number of possible values ​​for the semantic data.

[0146] Furthermore, it is not mandatory that there exists in the data structure DS a semantic data associated with all the possible index values. The calculated index I' from the semantic data SD can therefore be obtained in two ways: • Or a first index I is initially calculated from the semantic data SD, but this index is not present in the data structure DS. Two cases can then arise depending on the implementation modes: • The method can search in the data structure DS for an index I' close to the initially calculated index I and it is this index I' which will be the index obtained by the module 101 and finally transmitted by the module 102 in the CNL channel. The proximity between indexes could for example be defined by the Hamming distance and must in any case ensure the proximity between the data to be transmitted SD and the transmitted data SD' associated with the calculated index I' in the data structure DS. • Or the index I obtained from the semantic data SD is not present in the data structure DS but is nevertheless transmitted by the module 102 in the CNL channel. It is the management entity 200 of the reception which will have to search from the index received (I, in the case where the CNL channel does not introduce transmission errors) a close index present in the data structure DS, which will make it possible to read in the data structure DS a semantic data item associated with this close index. • Or the index obtained from the semantic data SD is present in the data structure DS, for example because the latter covers the set of all values ​​that an index of length can take data. In this case, the method transmits the obtained index I' but this corresponds, in all probability, in the data structure DS, to a semantic data SD' different from SD. The indexing function F or another method of obtaining the index I' must then ensure the proximity between the semantic data SD to be communicated and the semantic data SD' actually communicated via the obtained index I'.

[0147] In embodiments, the indexing function F is chosen such that, for two distinct semantic data SD, SD', if the given distance D between these two semantic data SD, SD' is less than a given threshold, then the probability that the calculated indexes I, F by the indexing function F for these two semantic data SD, SD' are the same is greater than a given probability. The choice of an indexing function F having this property makes it possible to increase the chances that two semantic data close according to the distance D measured between them will have the same calculated index.

[0148] An indexing function F having these properties is given by the family of locality-sensitive hashing functions (LSH functions). These functions produce indexes that are close to each other, or even equal, for starting elements that are close to each other for a distance D. Other examples of functions having this property can be used in different embodiments.

[0149] The indexing function F chosen may be an LSH function for cosine similarity. In other embodiments, it may be an LSH function for Euclidean distance.

[0150] In embodiments, the calculation of the index F is carried out by searching in the data structure DS for the semantic data SD' closest to the data to be transmitted SD according to the measured distance D. In this way, it is ensured that the data transmitted SD' via the calculated index F is indeed the semantic data SD' closest among the possible data present in the data structure DS.

[0151] Here, the indexes present in the data structure DS have been obtained beforehand for a set of semantic data present in the data structure DS which achieve an appropriate coverage of the values ​​which can be taken by the semantic data SD to be communicated in the vector space to which they belong. In this way, it is not necessary to calculate for each semantic data SD to be transmitted an index F but simply to find in the data structure DS the semantic data SD' closest to the data to be communicated SD according to the distance D, then to transmit this index I' found in the data structure DS.

[0152] The DS data structure therefore performs a quantification of the vector space in which the semantic data SD, SD' to be communicated can take values. This quantification makes it possible to transmit, instead of the bits or symbols constituting a complete SD semantic data, an index calculated I' by the function F, an index which is associated in the DS data structure with a SD semantic data close to the SD semantic data according to the distance D.

[0153] In general, the methods according to the invention operate when the data structure DS is known both by the communication management entity 100 and the reception entity 200. This is not an obligation: the communication management entity 100 can simply calculate an index I for a semantic data item SD and transmit it, without consideration of the presence of the index I in a data structure DS, and it is upon reception that the management entity 200 will make the connection between the received index (I, if the CNL channel does not introduce transmission errors, or I' if there have been transmission errors) and those present in the data structure DS.

[0154] In exemplary embodiments, the method for managing the communication of semantic data SD by the management entity 100 comprises a fragmentation step.

[0155] We have seen that the DS data structure performs a quantification of the vector space in which the semantic data SD, SD' take values. But the size of the DS data structure is limited. We have seen for example that a DS data structure associating 65,536 semantic data SD, SD' with a size of 4096 reals encoded on 32 bits would have an overall size close to a gigabyte. Such a size would probably be maximum. It may therefore be interesting instead to fragment the semantic data SD to be communicated, for example into fragments of dimensions 4, 16, 64 etc. The DS data structure then used would associate fragments of semantic data of the given dimension with indexes. The same principles regarding the indexing calculation relative to the given distance D then apply but in the vector space of the fragments, of reduced dimension compared to the complete vector space of the semantic data.The index I' obtained is then an index making it possible to find in the data structure DS a fragment of given dimension. In associated exemplary embodiments, the method for receiving semantic data by the management entity 200 then comprises a defragmentation step. The management entity 200 receives in these embodiments several indexes corresponding to the different fragments, indexes transmitted by the management entity 100. After having found in the data structure DS the fragments corresponding to the received indexes, the management entity 200 reconstructs a complete semantic data by combining the different fragments found.

[0156] In embodiments, the data structure DS is defined beforehand according to the format of the semantic data SD to be transmitted.

[0157] For a format frequently used in artificial intelligence applications, it may be interesting to define in advance one or more DS data structures that perform a suitable quantification of the vector space corresponding to the semantic data to be transmitted. For a given format, several different DS data structures may exist that will be more or less suitable for the applications using the communicated semantic data. Semantic data corresponding to visual content may, for example, be more concentrated in a specific area of ​​the vector space where they take values ​​that are different from the areas where semantic data corresponding to textual content are found, and this for the same semantic data format. The quantifications of interest, and the DS data structures corresponding to these two tasks will therefore be distinct.

[0158] For a given format of semantic data to be communicated, the DS data structure can be defined using a machine learning algorithm. Indeed, the best way to concentrate the quantification performed by the DS data structure at the places where it is useful is to perform machine learning from a corpus of semantic data. This corpus can be representative of a specific task using the semantic data to be transmitted, or be generic for all types of semantic data using a given format.

[0159] In the case of automatic learning of the DS data structure, a suitable learning algorithm is the k-means method where k is the number of semantic data present in the data structure. Here, the automatic learning will perform a clustering of a set of data SD, SD' communicated or to be communicated, which form a training set, around a reduced number of representative semantic data which will be recorded in the DS data structure. Such a clustering can be carried out using the k-means method, where k is the size of the data structure.Thanks to this embodiment, the grouping is uniform, in the sense that the data belonging to the training set are well distributed in an egalitarian manner into groups whose center is a representative semantic data which will be found in the DS data structure to represent all the semantic data of the same grouping.

[0160] Associatedly, an indexing function can be learned that will index each semantic data present in the data structure DS so that the data close according to the given distance D present in the data structure DS will be associated with indexes close to each other, for example according to the Hamming distance between indexes.

[0161] In embodiments, the indexing function F is defined in advance for a given format of semantic data or learned according to a training corpus. The training corpus can be generic, for all tasks expected from the semantic data in a given format, or specific to the semantic data to be communicated for a given task.

[0162] Some functions are particularly suited to machine learning. This is the case of locality-sensitive hash functions (LSH functions). These functions are defined according to hyperplanes that divide the vector space in which the semantic data take values ​​into two parts. One part, located on one side of the hyperplane, will correspond to the value 0 for a bit of given rank in the calculated index while the other part, located on the other side, will correspond to the value 1 for the bit of the same rank. For a given index length L, an equal number of hyperplanes is fixed that will divide the vector space into 2 to the power of L, if we consider that the index is a sequence of bits. Similar calculations can be done when the index is a sequence of other symbols, by reducing these other symbols to bits.

[0163] The placement of these hyperplanes, which define an LSH indexing function, can be obtained by learning according to the density of the data to be communicated in the vector space, in order to produce the most suitable cuts.

[0164] To improve the operation of the invention in the embodiment shown in [Fig.l], it is preferable that the indexing function F and the data structure DS are known both by the communication management entity 100 and by the reception management entity 200. One way of proceeding is for example, for the management entity 100, to transmit, in a preamble to the transmission which may be found in a packet header or in a datagram intended for this role, references to an indexing function F and / or a data structure DS, references which are known by the management entity 200.

[0165] In other embodiments, the management entity 100 may transmit the DS data structure itself in a prior exchange. In this case, the size of the DS data structure (and therefore the number of semantic data present in the DS data structure, performing the quantification of the vector space of the semantic data to be communicated) must be taken into account in evaluating the performance of the methods according to the invention. A prior transmission of a DS data structure adapted to the semantic data to be communicated subsequently must be made profitable by numerous communications of semantic data.

[0166] Embodiments of the invention include the negotiation of parameter values ​​useful for the transmission of semantic data for the invention. Thanks to these embodiments, more advanced possibilities of communication between the management entity 100 and the receiving management entity 200 are implemented. We have seen previously that values ​​useful for the execution of the method could be transmitted, for example in a packet header, and make it possible to find elements already defined, such as the data structure DS used in reception to find the semantic data SD' close to the data to be communicated SD. But this previous mode is not sufficient when, for example, the data structure DS and / or the indexing function F are learned by automatic learning. The receiver cannot in this case know a priori the elements useful for the operation of the method.

[0167] In this embodiment, a negotiation phase prior to the transmission of the index(es) makes it possible to agree on all the parameters useful for the communication of semantic data according to the method. This mode can be combined with the previous one: for example, a first transmission can give the references to a DS data structure, then a negotiation allows the transmitter and the receiver to agree on another parameter such as the size of the packets exchanged.

[0168] [Fig.2], for its part, illustrates an example of steps implemented within the framework of an embodiment of the invention.

[0169] In this example, the management entities 100 and 200 carry out a negotiation phase making it possible to set up the transfer of semantic data from the management entity 100 to the management entity 200 through a CNL channel not shown in [Fig.2]. For this, the management entities 100 and 200 exchange several messages transmitted through the same CNL channel or by another means.

[0170] The first REQ message is sent by the management entity 100 to the management entity 200 to indicate its intention to communicate semantic data and to propose parameters for performing its data communications as best as possible. Such parameters may indicate the format of the semantic data to be communicated, which may correspond to a known model name. Another parameter may indicate the intention to fragment the data as well as the size of the fragments obtained after fragmentation. Another parameter may be a reference to the data structure that will be used to quantify the fragments, i.e. the dictionary that groups values ​​associated with indexes. This dictionary may be known to both the management entity 100 and the management entity 200 and already present in memory. It may, for example, be known as performing good quantification for fragments of a given size for semantic data of a given format.The management entity. 100 may also indicate its desire to transmit a dictionary not known by the management entity 200 but particularly suited to the future communication. The REQ message may also include references to the indexing function F which will be used in embodiments which, again, may be a function already known or a function learned for the purposes of communicating semantic data. Another useful parameter to transmit is the length of the indexes which will be transmitted in place of the fragment values ​​which will be known elsewhere thanks to the data structure. Any other type of parameter useful for the communication may be proposed in the REQ message sent by the management entity 100 to the management entity 200.

[0171] The management entity 200 may respond with an ANS message that indicates the parameter values ​​it is willing to use for communication. For example, if the management entity proposes a data structure suitable for a certain fragment size, but the management entity 200 does not have this dictionary in memory, it may propose another fragmentation corresponding to a dictionary it has. Several exchanges of REQ-ANS messages may take place between the management entities 100 and 200, forming a negotiation of the parameters useful for the communication of semantic data between them.

[0172] The management entity 100 closes the actual negotiation phase by sending a SET message which transmits the actual values ​​on which the management entities 100 and 200 have just agreed. For example, the contents of the dictionary may be transmitted here whereas previously the management entities 100 and 200 negotiated the size of the dictionary that they were going to use.

[0173] An ACK acknowledgment message can be sent by the management entity 200 to the management entity 100 to indicate that the management entity 200 has received all of the values ​​useful for the communication of semantic data and is ready to receive indexes transmitted by the management entity 100. This exchange of messages corresponds to a possible embodiment of a phase of negotiation of the parameter values ​​useful for the communication of semantic data according to the invention, but other embodiments are possible.

[0174] In other embodiments, the parameter values ​​useful for the communication are not negotiated between the management entities 100 and 200 but are included in a datagram forming part of the index transmission. This mode can be used for communications of semantic data expressed in standard formats for which data structures carrying out appropriate quantifications are known and maintained in memory by all the management entities 100 and 200 carrying out the methods according to the invention.

[0175] The management entity 100 begins the communication by performing a FRAG fragmentation of the semantic data to be communicated. This operation consists of cutting the semantic data to be communicated in several fragments according to a given size that could have been negotiated or that corresponds to a known communication mode for a given semantic data format. For example, semantic data expressed as a 4096-dimensional vector can be divided into 512 fragments of dimension 8. Subsequent operations will apply to each of these fragments individually. The interest of such a division is, for example, to reduce the size of the data structure to be transmitted or saved in memory, since the elements present in this data structure will no longer be semantic data of dimension 4096 but fragments of dimension 8.

[0176] The management entity 100 continues, in this exemplary embodiment, with a quantification operation QNT. This operation may consist of searching in the data structure for the closest fragment according to a given distance from a fragment of the data to be communicated. The indexing operation IND then consists of finding in the data structure the index corresponding to this closest fragment. In other embodiments, an index is calculated directly for the fragment of the data to be communicated and the quantification operation QNT corresponds to the construction of the data structure used by the management entities 100 and 200.

[0177] The management entity 100 continues with a transmission operation TRS which comprises sending a message comprising the index I found in the data structure which has been negotiated, index I corresponding to a fragment present in the data structure close, according to the given distance, to the fragment of the data to be communicated. These quantification operations QNT, indexing IND and transmission TRS are repeated for all the fragments corresponding to the semantic data to be communicated which were obtained by the fragmentation operation FRG.

[0178] The management entity 200 will then carry out a succession of RCP reception operations of transmitted index I and RDG reading in the data structure of fragments corresponding to the received indexes I. A DFRG defragmentation operation then makes it possible to reconstruct the semantic data to be communicated from the read fragments which correspond to the received indexes. This reconstructed semantic data will be that which will be used in artificial intelligence applications in place of the semantic data to be communicated. The QNT quantification and / or IND indexing operations ensure that the indexes obtained are chosen according to a distance measured between the data to be communicated and that reconstructed from the fragments corresponding to the received indexes.

[0179] [Fig.3], for its part, illustrates examples of possible distortions undergone by semantic data during its communication and reception.

[0180] In the example of [Fig.3], to better visualize the distortions introduced by the communication and reception management methods according to the invention, it is assumed that the semantic data considered are vectors in a 3-dimensional space represented by the X, Y, Z reference frame located at the origin of these vectors. This example is not realistic because the semantic data or fragments that will be communicated will often be of a larger dimension, and which is a power of two, but the principles remain the same in the vector space concerned.

[0181] The objective is to communicate the semantic data SD.

[0182] It is assumed that, in this example, the method uses an indexing function F and that the indexing function F directly calculates an index I from SD, but that this index I is not present in the data structure DS. In certain embodiments, the method will not directly transmit the index I but will look for a close index F, which is indeed associated with a semantic data item SD' in the data structure DS. In other embodiments, the index I calculated by the function F will be considered as the index obtained and directly transmitted. The index F is transmitted in the CNL channel but this introduces transmission errors. The index received is then the index I”, which is assumed here not to be present in the data structure DS. The method according to the invention will then look for an index I'” close to the index I” such that I'” is present in the data structure DS.The semantic data SD'” associated with the index I'” in the data structure DS will then be used in various applications instead of the initial data SD. The accumulation of distortions means that the data used SD”' is not the initial data SD, but these distortions are reduced to a minimum by suitable choices of indexes based on a distance D measured between semantic data, the indexing function F, the function calculating the distance D measured between semantic data, and the quantification by the data structure DS of the vector space XYZ in which the semantic data are located.

[0183] In the previous example, the indexes I and I” are not present in the DS data structure. In some embodiments, the DS data structure includes an index-semantic data association for all possible values ​​of the indexes. In this case, the index I directly calculated from the SD data will be transmitted directly by the CNL channel. And the semantic data that will be used finally will be directly that associated with the received index I” after the transmission errors of the CNL channel have transformed the index I into I”.

[0184] Furthermore, it is important to understand that, even if all possible index values ​​are present in the DS data structure, the semantic data associated with an index calculated in the DS data structure will not generally be the semantic data for which the index was calculated. Indeed, there is a finite number of possible index values ​​while the semantic data, which are vectors in a vector space of real numbers, can take an infinite number of values. A calculated index I will therefore be the same for a whole set of semantic data while a single semantic data will be associated with this index in the data structure DS. To improve the efficiency of the method according to the invention, the semantic data which are sent to the same index by the indexing function F must be close to each other according to the distance D measured between them. These data close to each other then form a distinct zone in the vector space. The semantic data which will be chosen as being associated with the index calculated for the semantic data of this zone can be for example the data located at the barycenter of this zone to improve the representativeness of the semantic data which have the same calculated index associated with a single semantic data.

[0185] [Fig.4], for its part, illustrates experimental results obtained during the development of the methods according to the invention.

[0186] In this experiment, we wanted to show to what extent the distortions introduced by the methods for managing communication and reception of semantic data according to the invention impacted the artificial intelligence tasks in which the communicated semantic data are used.

[0187] The Hugging Face research community (accessible via the website http: / / huggingface.co) offers examples of tasks, models, and data sets for comparing algorithms in the field of artificial intelligence. For our experiments, we considered the banking77 data set (accessible at https: / / huggingface.co / datasets / banking77). This data set corresponds to several tens of thousands of sentences (in English) corresponding to 77 possible requests in the banking field, such as requesting an opposition in the event of theft of a bank card, or requesting the balance of an account, or other examples.

[0188] We converted this set of textual data into semantic data according to the Mistral-7B model with a size of 4096 real encoded cn floal32. The semantic data therefore represents a semantic content which is that of the initial texts. These semantic contents associated with the semantic data must be preserved as much as possible despite the distortions introduced by the methods according to the invention allowing the communication of semantic data.

[0189] We took a portion of this converted semantic data (and the known classification of the semantic contents associated with it) to train a neural network to classify a semantic data item into one of the 77 categories of requests in the banking domain. This is therefore a task of classic classification of application of artificial intelligence, and which uses semantic data for its manipulations.

[0190] The performances of the trained neural network, measured on semantic data other than those used for training the neural network, are presented in [Fig.3] by the dashed line, labeled nominal model, close to the top. They are greater than 90% correct classification. In contrast, random performances correspond to a success of 1 / 77. They are represented by the dotted line, labeled random model, close to zero (indeed, 1 / 77 = 1.3%).

[0191] Between these two extremes, we find three curves which represent the results of classification tasks which are obtained after using three distinct ways of communicating semantic data with the methods according to the invention in the presence of a CNL channel introducing more and more disturbances.

[0192] To obtain the curves in question, we simulated the communication of semantic data between two devices EQ1, EQ2 in a CNL channel and submitted the communicated data to the neural network that we trained. The difference with the nominal model line therefore shows the impact of the communication of semantic data according to the method on the classification task carried out by the neural network. The CNL channel is simulated as having a variable error rate in the transmission of bits, a rate which is called BER (acronym for bit error ratio). The horizontal scale is logarithmic and shows simulations for a CNL channel having an error rate which varies from 1Q'3, i.e. one chance in 1000 of a bit transmission error up to more than 10'1, i.e. more than one chance in ten of a transmission error, passing through 10'2 (one chance in 100 of a transmission error).

[0193] In the three simulated communication examples, we chose to fragment the semantic data into fragments of dimension 8. A semantic data which initially comprises 4096 reals will therefore be divided into 512 fragments which comprise 8 reals encoded in float32, i.e. 512 fragments of 8 * 32 = 256 bits.

[0194] We then performed quantifications adapted to the 8-dimensional semantic data fragments. For this, we took another corpus proposed by the Hugging Face research community, namely the Locutusque UltraTextbooks 2.0 corpus, accessible at the link http: / / huggingface.co / datasets / Locutusque / UltraTextbooks-2.0, and performed training using a clustering algorithm. It should be noted that this corpus of textual data has nothing to do with the banking domain. The quantifications learned here could therefore be used to transfer semantic data representing texts outside the banking domain in a probably equally efficient way.

[0195] We carried out three quantifications from this corpus to learn 2, then 16, then 1024 semantic data which are used to quantify fragments of dimension 8. The 2 fragments of quantified semantic data can then be represented by an index of length 1 bit; the 16 fragments of semantic data can be represented by an index of length 4 bits; and the 1024 fragments of semantic data by an index of length 10 bits.

[0196] An indexing function is then used to calculate an index of length 1, 4 or 10 bits. The calculated indexes are transferred and subjected to simulated variable error rates. The fragments corresponding to the received indexes, having undergone the transmission errors, are recovered, then a complete semantic data is constructed from the fragments and is submitted to the trained neural network, and the classification performance is measured. The classification performances are of course between the nominal model line (classification without transmission of semantic data) and the random model line (random classification).

[0197] We see that, for indexes of length 4 or 10 bits, the classification performances are very close to the nominal performance up to an error rate of 102. However, the use of these indexes corresponds to a reduction in the size of the transmitted data of 8 times 32 = 216 bits to 4 or 10 bits. We therefore have here a reduction factor of at least 20 and up to 54 for the index of length 4 bits if we had chosen to communicate the semantic data by directly transmitting the bits which represent the 512 fragments of 8 reals encoded on 32 bits. This very strong reduction is achieved with a minimal loss at the level of the classification task. The performances remain above 80% up to a fairly high error rate of 101.

[0198] 1-bit quantization has maximum performance a little below 80%. These results are therefore not very good, but not negligible either given a reduction factor of 216 in the size of the transmitted data.

[0199] Finally, when the BER increases to approach a rate of one error in 2, the classification performance drops and reaches the performance of a random classification.

[0200] These experiments therefore clearly show the interest of the invention which offers a reduction in the number of bits transmitted compared to a direct transmission of the bits forming the communicated semantic data by a very appreciable factor and a slight drop in classification performance, even in the presence of transmission errors at a significant rate.

[0201] Finally, let us point out here that, in the present text, the term "module" can correspond to a software component as well as to a hardware component or a set of hardware and software components, a software component itself corresponding to one or more computer programs or sub-programs or in a more general to any element of a program capable of implementing a function or a set of functions as described for the modules concerned. In the same way, a hardware component corresponds to any element of a hardware assembly capable of implementing a function or a set of functions for the module concerned (integrated circuit, smart card, memory card, etc.).

Claims

Claims

1. Method for managing the communication of semantic data (SD), called data to be communicated, by a first device (EQ1) to a second device (EQ2), characterized in that it comprises obtaining an index (F) corresponding to the data to be communicated (SD), said index (!') making it possible to find in a data structure (DS) another semantic data (SD'), called communicated data, the index (F) being chosen as a function of a distance (D) measured between the data to be communicated (SD) and the communicated data (SD'), followed by the transmission of the index obtained (!') through a communication channel (CNL) between the first device (EQ1) and the second device (EQ2).

2. Management method according to claim 1 characterized in that the distance (D) measured between the data to be communicated (SD) and the communicated data (SD') is a function of the proximity between the semantic contents associated with the two semantic data (SD, SD').

3. Management method according to one of claims 1 or 2 characterized in that the index (F) corresponding to the data to be communicated (SD) is obtained by searching in the data structure (DS) for the semantic data (SD') closest to the data to be transmitted (SD) according to the measured distance (D), and the transmitted index is the index (!') of said closest semantic data (SD').

4. Management method according to one of claims 1 to 3, characterized in that the index (!') is obtained by using an indexing function (F) chosen in such a way that, for two distinct semantic data, the closer these semantic data are according to the distance (D) measured between them, the closer the indexes calculated for these two semantic data by the indexing function (F) are to each other.

5. Management method according to one of claims 1 to 4 characterized in that the distance (D) between semantic data is measured using cosine similarity.

6. Management method according to one of claims 4 or 5 characterized in that the indexing function (F) is a locality-sensitive hash function.

7. Management method according to one of claims 1 to 6 characterized in that the method comprises a prior step of fragmenting the semantic data to be communicated (SD) into several fragments and the subsequent steps of the method apply to said several fragments.

8. Management method according to one of claims 1 to 7 characterized in that the data structure (DS) is defined beforehand according to the format of the semantic data (SD) to be communicated.

9. Management method according to one of claims 1 to 8 characterized in that the data structure (DS) is obtained by automatic learning based on the semantic data already communicated and / or intended to be communicated.

10. Management method according to claim 9 characterized in that the automatic learning of the data structure (DS) uses the k-means method where k is the number of semantic data present in the data structure (DS).

11. Method for managing the reception of semantic data (SD') characterized in that it comprises the reception of an index (!'), called received index, followed by the reading of semantic data (SD') in a data structure (DS) as a function of the received index (!').

12. Management method according to claim 11 characterized in that the method comprises, if the received index (!') does not correspond to any index present in the data structure (DS), a search in the data structure (DS) for the index closest to the received index (!') and in that the semantic data (SD') read in the data structure (DS) is that which corresponds to said closest index.

13. Management method according to one of claims 11 or 12 characterized in that, the method comprises, if the received index (!') corresponds to a fragment of semantic data, the reconstitution of complete semantic data from the fragments of semantic data corresponding to the received indexes.

14. Management entity (100) capable of managing the communication of semantic data (SD), called data to be communicated, by a first device to a second device, characterized in that it comprises the following modules: A module (101) for obtaining an index (!') corresponding to the data to be communicated (SD), said index (!') making it possible to find in a data structure (DS) another semantic data (SD'), called communicated data, the index (!') being chosen as a function of a distance (D) measured between the data to be communicated (SD) and the communicated data (SD'); A module (102) for transmitting the index obtained (!') through a communication channel (CNL) between the first equipment (EQ1) and the second equipment (EQ2).

15. Telecommunications equipment (EQ1) comprising a management entity (100) capable of managing the communication of semantic data (SD) according to claim 14.

16. Management entity (200) capable of managing the reception of semantic data (SD') characterized in that it comprises the following modules: A module (201) for receiving an index (!'), called received index; A module (202) for reading semantic data (SD') in a data structure (DS) as a function of the received index (!').

17. Telecommunications equipment (EQ2) comprising a management entity (200) capable of managing the reception of semantic data (SD') according to claim 16.

18. A computer program capable of being implemented by a management entity (100), the program comprising code instructions which, when executed by a processor, performs the steps of the management method defined in claim 1.

19. A computer program capable of being implemented by a management entity (200), the program comprising code instructions which, when executed by a processor, performs the steps of the management method defined in claim 11.

Images