First node, second node and methods performed thereby for handling a knowledge graph

Abstract

A computer-implemented method performed by a first node (111). The method is for handling a first KG. The first node (111) operates in a communications system (100). The first node (111) obtains (405) data encoding conversational speech. The first node (111) also obtains (406) the features extracted from the data, refraining from converting the data into text. The features are short-term features, The first node (111) determines (407), using an artificial neural network (121), a first KG representing words in a conversational speech and relations of the words. The determining (407) uses, as first input, the obtained features extracted from the data. The first node (111) then initiates (408) outputting an indication of the determined a first KG.

Description

[0001]FIRST NODE, SECOND NODE AND METHODS PERFORMED THEREBY FOR HANDLING A KNOWLEDGE GRAPH TECHNICAL FIELD The present disclosure relates generally to a first node and methods performed thereby for handling a first knowledge graph (KG). The present disclosure further relates generally to a second node and methods performed thereby, for handling a respective first KG. The present disclosure also relates generally to computer programs and computer-readable storage mediums, having stored thereon the computer programs to carry out these methods. BACKGROUND Computer systems in a communications network or communications system may comprise one or more nodes. A node may comprise one or more processors which, together with computer program code may perform different functions and actions, a memory, a receiving port, and a sending port. A node may be, for example, a server. Nodes may perform their functions entirely on the cloud. Computer systems may be comprised in a telecommunications network. The telecommunications network may cover a geographical area which may be divided into cell areas, each cell area being served by a type of node, a network node in the Radio Access Network (RAN), radio network node or Transmission Point (TP), for example, an access node such as a Base Station (BS), e.g., a Radio Base Station (RBS), which sometimes may be referred to as e.g., gNB, evolved Node B (“eNB”), “eNodeB”, “NodeB”, “B node”, or Base Transceiver Station (BTS), depending on the technology and terminology used. The base stations may be of different classes such as e.g., Wide Area Base Stations, Medium Range Base Stations, Local Area Base Stations and Home Base Stations, based on transmission power and thereby also cell size. A cell may be understood to be the geographical area where radio coverage may be provided by the base station at a base station site. One base station, situated on the base station site, may serve one or several cells. Further, each base station may support one or several communication technologies. The telecommunications network may also comprise network nodes which may serve receiving nodes, such as user equipments, with serving beams. The standardization organization Third Generation Partnership Project (3GPP) is currently in the process of specifying a New Radio Interface called Next Generation Radio or New Radio (NR), as well as a Fifth Generation (5G) Packet Core Network, which may be referred to as 5G Core Network (5GC). The advantages of 5G NR may include higher bandwidth, more resources, low latency and network slicing. 5G may provide services to various applications, such as enhanced Mobile Broad Band (eMBB), machine to Machine type communication (mMTC), Ultra Reliable Low Latency Communication (URLLC), etc. 5G may be understood to bring in sizeable flexibility with technological advancements along with innovations of cloud and AI. This may be understood to bring a whole new set of opportunities in the enterprise segment. For many enterprises, mobile cellular technology has already proven to bring great value to their digitalization process, which may include numerous use cases, such as autonomous robotics, enhanced video services, connected vehicles, remote operations, hazard, and maintenance sensors etc. This may be understood to not only enhance productivity in connected factories, but also make workplaces safer. In the course of operations of the telecommunications network, data may be collected via the telecommunications network, which may enable to monitor and manage different functions. The advent of for example, the Internet of Things (IoT) has exponentially increased the amount of data to be monitored. The availability of large amounts of data, such as those collected for example, from IoT devices, may be understood to enable the possibility of analysing such data to make predictions on events, with a high predictive power. To make predictions on events may be understood to refer to building mathematical models that may fit those data, which mathematical models may then be used to make predictions for such events. Within this context, machine learning models may be used to analyze the data collected, and enable an improved management of different types of operations via the telecommunications network. Machine Learning Machine learning (ML) may be understood as the study of computer algorithms that may improve automatically through experience. It is seen as a part of Artificial Intelligence (AI). ML algorithms may build a model based on sample data, known as "training data", in order to make predictions or decisions without being explicitly programmed to do so. ML algorithms may be used in a wide variety of applications, such as email filtering and computer vision, where it may be difficult or unfeasible to develop conventional algorithms to perform the needed tasks. There may be basically 3 types of ML Algorithms: Supervised Learning, Unsupervised Learning, and Reinforcement Learning (RL). Supervised Learning algorithms may comprise a target / outcome variable, or dependent variable, which may have to be predicted from a given set of predictors, that is, independent variables. Using this set of variables, a function may be generated that may map inputs to desired outputs. The training process may continue until the model may achieve a desired level of accuracy on the training data. Once an ML model may have been trained, an inference process may begin, whereby new data may be run through the ML model to calculate an output. Examples of Supervised Learning may be Regression, Decision Tree, Random Forest, KNN, Logistic Regression etc. In Unsupervised Learning algorithms, there may be no target or outcome variable to predict / estimate. It may be used for clustering a population into different groups, which may be widely used for segmenting customers in different groups for specific intervention. Examples of Unsupervised Learning may be K-means, mean-shift clustering, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Expectation–Maximization (EM) Clustering using Gaussian Mixture Models (GMM), Agglomerative Hierarchical Clustering, etc…. Cluster analysis or clustering may be understood as an ML technique which may comprise grouping a set of objects in such a way that objects in the same group, which may be called a cluster, may be understood to be more similar, in some sense, to each other than to those in other groups, that is, other clusters. It may be understood as a main task of exploratory data mining, and a common technique for statistical data analysis, used in many fields, including pattern recognition, image analysis, information retrieval, bioinformatics, data compression, computer graphics and ML. Using an RL algorithm, a machine may be trained to make specific decisions. It may be understood to work as follows: the machine may be exposed to an environment where it may train itself continually using trial and error. This machine may learn from past experience and may try to capture the best possible knowledge to make accurate decisions. An example of RL may be a Markov Decision Process (MDP). The training using RL may comprise generating an ML model. To train such an ML model, an agent, given a state of the environment, may take an action in this environment and receive a reward. The action may result in a new state of the environment. This process may be repeated in a loop. Over time, the agent may learn to take actions that may result in larger immediate and future rewards, meaning that it may be understood to be in the best interest of the agent not to take the action that may only lead to the highest reward in the next state, but the action that may cumulatively lead to the highest reward in the next state and in a future number of states. The agent may comprise a neural network which may input the state and may produce an action. There may be several ML algorithms that may be used for training the network of the agent, e.g., policy-learning based, such as actor-critic approaches, or value-based learning, such as deep-q networks. Knowledge Graphs (KGs) One of the primary modes of human communication is speech signal and the information available / transmitted in the form of speech signal is enormous. Hence, it may be understood to be important to create a KG automatically from speech signal. A KG may be understood as a digital structure that may be understood to represent knowledge as concepts and relationships between them. Such methods may be understood to have a wide range of applications in semantic communication such as data mining, speech analytics, search optimization, etc. The challenges in terms of speech signal may include diverse languages, accents, emotions, noisy environment, etc. In addition, the language adoption across the world may be understood to be varied, for example, Chinese, English and Spanish may be understood to be more widely adopted. The existing KG creation algorithms are primarily developed for written text content, which possesses a linguistically well-defined sentence structure. These algorithms may be understood to exploit such structures present in the grammar compliant text, that is, written text, to obtain the KG with reasonable accuracy. On the other hand, conversational speech may be understood to be spontaneous in nature, and it may contain disfluencies, fillers, etc. Since conversational speech may not possess any easily learnable linguistic structure, it may make the creation of KG from conversational text a challenging task. To apply the conventional text-to-graph approaches for generating KG from speech signal, speech may have to be first converted to text, see Figure 1. The errors introduced in the speech-to-text conversion may further complicate the process of KG creation from speech. There are certain languages that do not have any script, but people still use them in their daily lives, for example, Konkani and Tulu languages present in India, Yanomami, Shehri etc. and for these cases, KG creation using traditional approaches is next to impossible, since it may be necessary to create a script for these languages which is extremely costly. Figure 1 is a schematic diagram depicting a cascaded approach to obtain a KG from speech, according to existing methods. In a typical cascaded approach such as that depicted in Figure 1, to automatically obtain the KG from a speech utterance requires two major steps. A first step 1 involves an automatic speech recognition system (ASR) 2, where a speech to text module may convert the speech into the corresponding text. In the second step 3, a KG 4 may be created by a Text-to-Graph module using this text data output from the ASR. The speech-to-text processing unit may convert speech to the unpunctuated text using an ASR and a punctuation module followed by the text-to-graph, which may be understood to contain several modules in cascade such as preprocessing, chunking, resolution, extraction, and postprocessing, module generate KG from the punctuated text. Typical KG creation algorithms from text, such as that depicted in the schematic diagram of Figure 2, contain different sub-modules such as sentence extraction, co-reference resolution, entity extraction, relation extraction, etc. Figure 2 schematically represents a conventional KG creation process. As schematically depicted in the diagram of Figure 1, such KG algorithms may take text as an input at 21. From the text, a sentence extraction module 22 may extract sentences. A co-reference resolution module 23 may then find all mentions of the same entity in the text, even if it may not be explicitly mentioned. In such algorithms, a triplet extraction module 24, of the variant <h,r,t>, may then extract a set of triplets of format (h, r, t). An entity extraction module 25 in the triplet extraction module 24 may extract the two entities, h and t from the text. A relation extraction module 26 in the triplet extraction module 24 may then extract a relation r, wherein each triplet may be understood to link the two entities, h and t through the relation r. The triplets may then be input to the graph by another module 27. A triplet extraction module of the variant <h,r,t> may assume that the input text has grammatical coherence for the task of deriving the correct KG. There may be variants of such triplet representation. Typically, a KG may have verbs, predicates representative of edges of the graph and the subject, objects and other Named Entity Recognitions (NERs) being represented as nodes of the KG. The triplets may be represented in embedded space, also known as KG embeddings (KGE), and this may aid in bringing in semantical aspects of the nodes / relations and graphs from such triplets, that is, it may improve semantic representation. To obtain a KG from text data in a traditional approach, the modules in the traditional KG algorithms, such as those depicted in Figure 2, may be independently optimized for the corresponding tasks, and it is not explicitly optimizing any cost function directly related to the KG prediction error. Since there is no joint optimization present in the conventional approaches, any errors produced by the individual submodules, for conversational type of data the error produced by the individual modules are going to be high, may propagate to the downstream modules, and result in an error compounding. Hence, existing KG creation algorithms, may result in prediction error. SUMMARY As part of the development of embodiments herein, one or more problems with the existing technology will first be identified and discussed. The conventional cascaded approach has the following limitations. In real-life conversations, humans communicate through speech without following any grammatical adherence and contain disfluencies. Hence, text corresponding to a conversational speech may not possess a well-defined sentence structure, such as the one that may be present in the written text documents. Even segmenting text sequences into grammatically meaning full sentences may be a challenging task. An accurate KG generation, using these types of text data, such as in the algorithm depicted in Figure 2 is difficult since the accuracy of triplet identification, that is, of nodes and edges, may be understood to rely on well-defined word relations in sentences / text documents. Another limitation is that in a conventional cascaded approach, errors produced in one module will adversely affect the downstream processing modules and have a compounding effect on the final KG prediction error. In the conventional cascaded approach, to get text automatically from speech, an ASR may need to be employed, and the effect of errors which are obtained due to the ASR engine cannot be eliminated completely. Errors present in the ASR output further complicate graph creation, as ASR errors which propagate to the downstream module, e.g., the text-to-graph module, have a compounding effect on the overall error present in the predicted KG. For example, the ASR output typically has three types of errors such as deletion, insertion and substitution. These errors will propagate to downstream modules and hamper the KG prediction. Similarly, all the modules present in the cascaded approach will produce errors and since these modules are not jointly optimized for the KG prediction task, the errors will propagate to the downstream modules and may produce a compounding effect. Since these errors are large in the case of conversational speech data, the use of a cascaded approach may not be effective. A further limitation is that the conventional cascaded approach cannot exploit all the relevant information, for example, para-linguistic features present in the speech signal. In addition to linguistic information, speech signals may contain many useful information such as non-linguistic, para-linguistic, emotion, etc. This information may be understood to be very relevant in cross-language speech-to-KG graph creation because this information may help to get the better context of the conversation. However, learning useful information from speech and utilizing it for the prediction of KG may require an end-to-end Neural Network (NN) framework where input to the NN is speech signal, and the output is the corresponding KG. Since the input to the text-to-graph module may be understood to be text, during the graph creation, the conventional cascaded approach cannot exploit all the information present in the speech signal relevant for KG creation, such as paralinguistic features and non-linguistic features, which is not represented in the text. An additional limitation is that cascaded approaches cannot be applied to languages where there is no textual representation. Since the cascaded approach requires intermediate text representation, for languages that do not possess any script, or languages where the ASR type of data is not available / rich, using the cascaded approach for KG creation is very expensive or may not be even possible, because it may be understood to require annotation by linguistic experts, and may need to develop / define rules for creating the new script. There is a need to provide downstream tasks such as conversational AI or retrieval which may support many languages including resource poor languages. Yet another limitation is that cascaded approaches do not explicitly minimize any KG prediction error. According to the foregoing, it is an object of embodiments herein to improve the handling a KG in a computer system. According to a first aspect of embodiments herein, the object is achieved by a computer- implemented method, performed by a first node. The method is for handling a first KG. The first node operates in a computer system. The first node obtains data encoding conversational speech. The first node obtains features extracted from the data, refraining from converting the data into text. The features are short-term features. The first node determines, using an artificial neural network, a first KG. The first KG represents words in conversational speech and relations of the words. The determining uses, as first input, the obtained features extracted from the data. The first node then initiates outputting an indication of the determined first KG. According to a second aspect of embodiments herein, the object is achieved by a computer-implemented method, performed by the second node. The method is for handling a respective first KG. The second node operates in the computer system. The second node obtains a plurality of sets of data. Each set of data encodes respective conversational speech. The second node then obtains, from the plurality of sets of data, a respective plurality of sets of features extracted from the respective set of data, refraining from converting the data into text. The second node then obtains, for each set of data of the plurality of sets of data, a respective plurality of words and a respective second KG. The respective second KG corresponds to a respective oracle KG of the respective set of data. The respective plurality of words corresponds to a respective oracle plurality of words. The second node then trains the artificial neural network to determine a respective first KG for each respective set of data using the obtained plurality of sets of data, the respective plurality of words and the respective second KG. The training is performed by minimizing a loss function representing a difference between each respective first KG and the obtained respective second KG. The training is performed iteratively until the loss function is below a threshold. The second node then initiates outputting a first indication of the trained artificial neural network. According to a third aspect of embodiments herein, the object is achieved by the first node. The first node may be understood to be for handling the first KG. The first node is configured to operate in the computer system. The first node is configured to obtain the data configured to encode conversational speech. The first node is also configured to obtain the features configured to be extracted from the data, refraining from converting the data into text. The features are configured to be short-term features. The first node is further configured to determine, using the artificial neural network, the first KG configured to represent words in the conversational speech and relations of the words. The determining is configured to use, as first input, the features configured to be extracted from the data configured to be obtained. The first node is also configured to initiate outputting the indication of the first KG configured to be determined. According to a fourth aspect of embodiments herein, the object is achieved by the second node. The second node may be understood to be for handling the respective first KG. The second node is configured to operate in the computer system. The second node is configured to obtain the plurality of sets of data. Each set of data is configured to encode the respective conversational speech. The second node is further configured to obtain, from the plurality of sets of data, the respective plurality of sets of features configured to be extracted from the respective set of data, refraining from converting the data into text. The second node is further configured to, obtain, for each set of data of the plurality of sets of data, the respective plurality of words and the respective second KG. The respective second KG is configured to correspond to the respective oracle KG of the respective set of data. The respective plurality of words is configured to correspond to the respective oracle plurality of words. The second node is further configured to train the artificial neural network to determine the respective first KG for each respective set of data using the plurality of sets of data, the respective plurality of words and the respective second KG configured to be obtained. The training is performed by minimizing the loss function configured to represent the difference between each respective first KG and the respective second KG configured to be obtained. The training is configured to be performed iteratively until the loss function is below the threshold. The second node is further configured to initiate outputting the first indication of the artificial neural network configured to be trained. According to a fifth aspect of embodiments herein, the object is achieved by a computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the first node. According to a sixth aspect of embodiments herein, the object is achieved by a computer-readable storage medium, having stored thereon the computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the first node. According to a seventh aspect of embodiments herein, the object is achieved by a computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the second node. According to an eighth aspect of embodiments herein, the object is achieved by a computer-readable storage medium, having stored thereon the computer program, comprising instructions which, when executed on at least one processing circuitry, cause the at least one processing circuitry to carry out the method performed by the second node. By obtaining the plurality of sets of data encoding the respective conversational speeches, and the respective plurality of sets of features extracted from the respective set of data, the second node may then be enabled to use the respective plurality of sets of features as input to train the artificial neural network to determine the respective first KG for each respective set of data, using as well the obtained respective plurality of words and the respective second KG as oracle KG and respective oracle plurality of words, respectively. By directly minimizing the error between the predicted graph and the oracle graph during training, the second node may be enabled to exploit the domain specific global and local linguistic structures present in the data to predict the accurate respective first KG. Furthermore, the approach may be understood to be agnostic to the domain, making it easier to use. By the second node then initiating outputting the first indication of the trained artificial neural network, the second node may enable a recipient, such as the first node, to use the trained artificial neural network to accurately predict KGs from unstructured text data, such as conversational data. By the first node obtaining the data encoding the conversational speech, and the features extracted from the data, the first node is then enabled to use the obtained features as input to the trained artificial neural network to determine the first KG of the conversational speech encoded in the data. By determining the first KG using the artificial neural network having refrained from converting the data into text, the first node may be enabled to automatically obtain, e.g., predict, an accurate KG from speech signal directly from speech, without any intermediate conversion process. The input may be understood to be the speech signal, and the output may be understood to be the respective first KG constructed from the meaning of the speech signal represented in any language, e.g., may be the same language also. Hence, the first node may enable the creation of the first KG in cross lingual speech-to-KG creation scenarios and also the case where languages may have no script available. In fact, the first node may make the KG creation accessible to resource poor languages. Moreover, by directly operating on speech signal, the first node may be enabled to exploit all the relevant information present in the speech signal to predict the first KG, in addition to the linguistic information such as the domain specific attributes, and paralinguistic features. By directly operating on speech signal, the first node may be enabled to predict the first KG, devoid of errors present in the conventional cascaded approach, such as errors in a speech-to-text module and modules present in the conventional text-to-graph creation process and its error compounding problems. The approach in embodiments herein used may be understood to be robust against grammatical mistakes and disfluencies. Furthermore, the approach may be understood to be agnostic to the grammatical mistakes of the user. The first node may be understood to automatically handle the issues with the speech on mis-spelled words, fillers and disfluencies. The approach used may be understood to be robust to lack of entities and relations in the conversational inputs modality. By the first node then initiating outputting the first indication of the determined first KG, the first node enables a receiving party to use the accurate first KG to effectively process the conversational speech in a variety of applications. BRIEF DESCRIPTION OF THE DRAWINGS Examples of embodiments herein are described in more detail with reference to the accompanying drawings, according to the following description. Figure 1 is a schematic diagram illustrating a cascaded approach to obtain a KG from speech, according to existing methods. Figure 2 is a schematic diagram illustrating a conventional KG creation process, according to existing methods. Figure 3 is a schematic diagram illustrating two non-limiting examples, in panels a) and b), of a computer system, according to embodiments herein. Figure 4 is a flowchart depicting a method in a first node, according to embodiments herein. Figure 5 is a flowchart depicting a method in a second node, according to embodiments herein. Figure 6 is a schematic diagram depicting a non-limiting example of the method performed by the first node, according to embodiments herein. Figure 7 is a schematic diagram depicting particular aspects of a non-limiting example of the method performed by any of the first node or the second node, according to embodiments herein. Figure 8 is a schematic diagram depicting particular aspects of another non-limiting example of the method performed by any of the first node and the second node, according to embodiments herein. Figure 9 is a schematic diagram depicting an English KG, generated from a French speech signal, according to embodiments herein. Figure 10 is a schematic diagram depicting another non-limiting example of the method performed by the first node, according to embodiments herein. Figure 11 is a signalling diagram depicting a further non-limiting example of the method performed by the first node, according to embodiments herein. Figure 12 is a schematic block diagram illustrating an embodiment of a first node, according to embodiments herein. Figure 13 is a schematic block diagram illustrating an embodiment of a second node, according to embodiments herein. DETAILED DESCRIPTION Certain aspects of the present disclosure and their embodiments address the challenges identified in the Background and Summary sections with the existing methods and provide solutions to the challenges discussed. Embodiments herein may be understood to relate to an end-to-end approach for KG creation from speech signals. In order to eliminate the compounding effect of error in the conventional approach, and to exploit all relevant information, including non-linguistic and para-linguistic features that may be present in the speech signal, embodiments herein may use an end-to-end NN architecture which approach may be referred to herein as a Listen, Attend, and Draw (LAD) Model for generating KG directly from speech signal, where input to the NN may be understood to be speech signal and the output may be understood to be the corresponding KG. Some of the embodiments contemplated will now be described more fully hereinafter with reference to the accompanying drawings, in which examples are shown. In this section, the embodiments herein will be illustrated in more detail by a number of exemplary embodiments. Other embodiments, however, are contained within the scope of the subject matter disclosed herein. The disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art. It should be noted that the exemplary embodiments herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. Several embodiments and examples are comprised herein. It should be noted that the embodiments and / or examples herein are not mutually exclusive. Components from one embodiment or example may be tacitly assumed to be present in another embodiment or example and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments and / or examples. Figure 3 depicts two non-limiting examples, in panels “a” and “b”, respectively, of a computer system 100, in which embodiments herein may be implemented. In some example implementations, such as that depicted in the non-limiting examples of Figure 3, the computer system 100 may be a computer network. In other example implementations, such as that depicted in panel b) of Figure 3, the computer system 100 may be implemented in a telecommunications system, sometimes also referred to as a cellular radio system, cellular network or wireless communications system. In some examples, the telecommunications system may comprise network nodes which may serve receiving nodes, such as wireless devices, with serving beams. In some examples, the telecommunications system may be, for example, a communications network, such as 5G system, or Next Gen network. The telecommunications system may also, or alternatively, support other technologies, such as LTE, e.g., LTE Frequency Division Duplex (FDD), LTE Time Division Duplex (TDD), LTE Half-Duplex Frequency Division Duplex (HD-FDD), LTE operating in an unlicensed band, Wideband Code Division Multiple Access (WCDMA), Universal Terrestrial Radio Access (UTRA) TDD, Global System for Mobile communication (GSM) / Enhanced Data Rate for GSM Evolution (EDGE) Radio Access Network (GERAN) network, Ultra-Mobile Broadband (UMB), EDGE network, network comprising of any combination of Radio Access Technologies (RATs) such as e.g., Multi-Standard Radio (MSR) base stations, multi-RAT base stations etc., any 3rd Generation Partnership Project (3GPP) cellular network, Wireless Local Area Network / s (WLAN) or WiFi network / s, Worldwide Interoperability for Microwave Access (WiMax), Internet of Things (IoT), Machine Type Communication (MTC), IEEE 802.15.4-based low-power short-range networks such as IPv6 over Low-Power Wireless Personal Area Networks (6LowPAN), Zigbee, Z-Wave, Bluetooth Low Energy (BLE), or any cellular network or system. The computer system 100 comprises nodes, whereof a first node 111 and a second node 112 are depicted in Figure 3. In some examples, such as that depicted in Figure 3 a), the first node 111 and the second node 112 may be co-located or be the same node. The computer system 100 may comprise additional nodes. As depicted in the non-limiting example of Figure 3 b), the computer system 100 may comprise a third node 113. Any of the first node 111, the second node 112 and the third node 113 may be understood, respectively, as a first computer system or server, a second computer system or server, and a third computer system or server. Any of the first node 111, the second node 112 and the third node 113 may be implemented as a standalone server in e.g., a host computer in the cloud 115, as depicted in the non-limiting example of Figure 3b) for the first node 111 and the second node 112. In other examples, any of the first node 111, the second node 112 and the third node 113 may be a distributed node or distributed server, such as a virtual node in the cloud 115, and may perform some of its respective functions locally, e.g., by a client manager, and some of its functions in the cloud 115, by e.g., a server manager. In other examples, any of the first node 111, the second node 112 and the third node 113 may perform its functions entirely on the cloud 115, or partially, in collaboration or collocated with a radio network node. Yet in other examples, any of the first node 111, the second node 112 and the third node 113 may also be implemented as processing resources in a server farm. In yet other examples, any of the first node 111, the second node 112 and the third node 113 may be comprised in a device, such as the device 140 described below, or on the edge. Any of the first node 111, the second node 112 and the third node 113 may be under the ownership or control of a service provider or may be operated by the service provider, or on behalf of the service provider. Any of the first node 111, and the second node 112 may have a capability to perform machine-implemented learning procedures, which may be also referred to as “machine learning” (ML). Any of the first node 111, and the second node 112 may have a capability to manage an artificial neural network 121. The artificial neural network 121 may be understood as a machine learning framework, which may comprise a collection of connected nodes, where in each node or perceptron, there may be an elementary decision unit. Each such node may have one or more inputs and an output. The input to a node may be from the output of another node or from a data source. Each of the nodes and connections may have certain weights or parameters associated with it. In order to solve a decision task, the weights may be learnt or optimized over a data set which may be representative of the decision task. The most commonly used node may have each input separately weighted, and the sum may be passed through a non-linear function which may be known as an activation function. The nature of the connections and the node may determine the type of the neural network, for example a feedforward network, recurrent neural network etc. That any of the first node 111 and the second node 112 may have the capability to manage the artificial neural network 121 may be understood herein as having the capability to store the training data set and the models that may result from the machine learning, to train a new model, and once the model may have been trained, to use this model for prediction. In some embodiments, such as those depicted in Figure 3 b), the system that may be used for training the model and the one used for prediction may be different. That is, the first node 111 may be different from the second node 112. In some embodiments, the first node 111 may be understood as a node having a capability to use the model, once the model may have already been trained. However, in other embodiments, the first node 111 may also have a capability to train the artificial neural network 121. Any of the second node 112, and in some embodiments, the first node 111, used for training the artificial neural network 121 may require more computational resources than the first node 111 that may use the trained model to make predictions. Therefore, any of the second node 112, and in some embodiments, the first node 111, used for training the artificial neural network 121 may, for example, support running python / Java with Tensorflow or Pytorch, Theano etc... Any of the first node 111 and the second node 112 may also have GPU capabilities. In some embodiments, any of the first node 111 and the second node 112 may be a core network node, such as, e.g., a network data analytics function (NWDAF), a Service management and orchestration (SMO) node, a positioning node, a coordinating node, a Self- Optimizing / Organizing Network (SON) node, a Minimization of Drive Test (MDT) node, etc…. In 5G, for example, any of the first node 111 and the second node 112 may be located in the Operations Support Systems (OSS). In other examples not depicted in Figure 3, any of the first node 111 and the second node 112 may be a radio network node, such as radio network node 130, depicted in Figure 3 b). The radio network node 130 may be, e.g., comprised in a Radio Access Network of the telecommunications system. That is, the radio network node 130 may be a transmission point such as a radio base station, for example a gNB, an eNB, or any other network node with similar features capable of serving a wireless device, such as a user equipment or a machine type communication device, in the computer system 100. In typical examples, the radio network node 130 may be a base station, such as a gNB or an eNB. In other examples, the radio network node 130 may be a distributed node, such as a virtual node in the cloud 115, and may perform its functions entirely on the cloud 115, or partially, in collaboration with a radio network node. The telecommunications system may cover a geographical area, which in some embodiments may be divided into cell areas, wherein each cell area may be served by a radio network node 130, although, one radio network node 130 may serve one or several cells. In the example of Figure 3, the cells are not depicted to simplify the figure. The radio network node 130 may be of different classes, such as, e.g., macro eNodeB, home eNodeB or pico base station, based on transmission power and thereby also cell size. In some examples, the radio network node 130 may serve receiving nodes with serving beams. The radio network node 130 may be directly connected to one or more core networks. Any of the first node 111, the second node 112 and the third node 113, and / or any of the nodes comprised in the computer system 100 may support one or several communication technologies, and its name may depend on the technology and terminology used. A device 140 may be comprised in the telecommunication network. The device 140 comprised in the computer system 100 may be a wireless communication device such as a 5G UE, or a UE, which may also be known as e.g., mobile terminal, wireless terminal and / or mobile station, a Customer Premises Equipment (CPE) a mobile telephone, cellular telephone, or laptop with wireless capability, just to mention some further examples. The device 140 comprised in the telecommunications system may be, for example, portable, pocket-storable, hand-held, computer-comprised, or a vehicle-mounted mobile device, enabled to communicate voice and / or data, via the RAN, with another entity, such as a server, a laptop, a Personal Digital Assistant (PDA), or a tablet, Machine-to-Machine (M2M) device, device equipped with a wireless interface, such as a printer or a file storage device, modem, sensor, IoT device, or any other radio network unit capable of communicating over a radio link in a communications system. In typical examples, the device 140 may be, or comprise, a microphone. The device 140 comprised in the telecommunications system may be enabled to communicate wirelessly in the telecommunications system. The communication may be performed e.g., via a RAN, and possibly the one or more core networks, which may be comprised within the telecommunications system. It may be understood that the telecommunications network may comprise additional radio network nodes 130 and / or additional devices 140. The first node 111 may be configured to communicate within the computer system 100 with the second node 112 over a first link 141, e.g., a radio link, or a wired link. The first node 111 may be configured to communicate within the computer system 100 with the radio network node 130 over a second link 142, e.g., a radio link, or a wired link. The second node 112 may be configured to communicate within the computer system 100 with the radio network node 130 over a third link 143, e.g., a radio link, or a wired link. The radio network node 130 may be configured to communicate within the computer system 100 with the device 140 over a fourth link 144, e.g., a radio link. The second node 112 may be configured to communicate within the computer system 100 with the third node 113 over a fifth link 145, e.g., a radio link, or a wired link. The radio network node 130 may be configured to communicate within the computer system 100 with the third node 113 over a sixth link 146, e.g., a radio link, or a wired link. The device 140 may be configured to communicate within the computer system 100 with the third node 113 over a seventh link 147, e.g., a radio link, or a wired link. The first node 111 may be configured to communicate within the computer system 100 with the third node 113 over an eighth link 148, e.g., a radio link, or a wired link. Any of the first link 141, the second link 142, the third link 143, the fourth link 144, the fifth link 145, the sixth link 146, the seventh link 147 and the eighth link 148 may be a direct link or may be comprised of a plurality of individual links, wherein it may go via one or more computer systems or one or more core networks in the computer system 100, which are not depicted in Figure 3, or it may go via an optional intermediate network. The intermediate network may be one of, or a combination of more than one of, a public, private or hosted network; the intermediate network, if any, may be a backbone network or the Internet; in particular, the intermediate network may comprise two or more sub-networks, which is not shown in Figure 3. In general, the usage of “first”, “second”, “third”, “fourth”, “fifth”, “sixth”, “seventh” and / or “eighth”, herein may be understood to be an arbitrary way to denote different elements or entities, and may be understood to not confer a cumulative or chronological character to the nouns they modify. Some of the embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. Other embodiments, however, are contained within the scope of the subject matter disclosed herein, the disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art. Embodiments of a computer-implemented method, performed by the first node 111, will now be described with reference to the flowchart depicted in Figure 4. The method is for handling a first KG. The first node 111 operates in the computer system 100. Several embodiments are comprised herein. In some embodiments all the actions may be performed. In some embodiments, some actions may be optional. In Figure 4, optional actions are indicated with dashed lines. It should be noted that the examples herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Action 401 Embodiments herein may be understood to provide an approach for KG creation from conversational speech data by using the first artificial neural network 121. In this Action 401, the first node 111 may obtain a plurality of sets of data. Each set of data may encode a respective conversational speech. The data may encode recordings of conversations. A set of data may correspond to the same topic, or the same speaker, or a certain time period, or may have a trait in common, so that the data within each set may be analyzed together in order to create a respective KG for that set of data. Obtaining in this Action 401 may comprise receiving from another node in the computer system 100, for example, from the device 140, via the radio network node 130, or fetching or retrieving from a memory. Action 402 In this Action 402, the first node 111 may obtain, from the plurality of sets of data, a plurality of respective sets of features extracted from the respective sets of data. That is, for each set of data obtained, a respective, or corresponding, set of features, may be obtained, e.g., extracted. Therefore, a plurality of respective sets of features may be obtained for the plurality of sets of data. Features may be understood as data, e.g., variables or attributes, extracted from, in embodiments herein signal corresponding to conversational speech. In some examples, the features may be short term features. Since, conversational speech may be understood to be a continuous signal, which may be continuously sampled, e.g., sampled at regular intervals, it may contain lot of information. Hence, the data may be windowed, that is, the data may be divided into windows of equal length, and a frequency transformation may be obtained to extract frequency level information. This may be performed by transformation techniques. A windowing transform such as Mel Frequency Cepstral Coefficients (MFCC), windowed Fourier transform, etc. may be used to extract features per window. These features may be referred to as short-time features, which may be understood be windowed features. Obtaining in this Action 401 may comprise receiving from another node in the computer system 100, for example, from the device 140, via the radio network node 130, fetching or retrieving from a memory, or the first node 111 extracting itself the plurality of respective sets of features from the respective sets of data. The first node 111 obtains the plurality of respective sets of features refraining from converting the data into text, e.g., refraining from using an ASR. This may therefore enable provide a framework to automatically generate KG from speech signal directly. The plurality of the respective sets of features may be extracted from the respective sets of data, that is, speech data, using, for example, MFCC, wave-2-vec [2], etc. The features may be short-time features. Accordingly, in some examples of embodiments herein, the first node 111, in this Action 402, or another node in the computer system 100 in a similar action, may convert speech to short-time features. By, in this Action 402, obtaining the plurality of respective sets of features extracted from the respective set of data, refraining from converting the data into text, the first node 111 may enable to eliminate the requirement of the intermediate text representation in the cascaded approach of existing methods, as described in Background section. Since the obtained plurality of the respective sets of features may then be used to generate respective KG for the respective set of data, the first node 111 may thereby ultimately enable KG based semantic communication in a wide variety of problems where human speech may be the primary source of information sharing. Hence, the first node 111 may enable that all the relevant information present in the speech signal, including non-linguistic and paralinguistic information, may be exploited to obtain the accurate KG. Action 403 The conversational speech signal comprised in the data in the plurality of sets of data obtained in Action 401 may comprise noisy words, grammatical errors, disfluencies, etc. Hence, such data may be understood to be a difficult type of data for the artificial neural network 121 to use in order to learn the correct prediction of a KG. In order to enable that the artificial neural network 121 may be able to learn all the relevant information present in the speech data for predicting the correct graph from it, a human annotator may listen to the speech signal and create the corresponding oracle KG (^^), referred to herein as a “second” KG, and words for the training of the artificial neural network 121. The second KG may be optimum in the sense that the second KG may be understood to need to closely match the content in the speech signal while using only the words present in the speech data. For this purpose, while creating the second KG, the annotator may need to remove errors or fillers comprised in the speech signal. The human annotator may then, in some examples, provide the second KG to the third node 113. Similarly, the human annotator may identify the words in each of the obtained respective sets of data. The second KG as well as the identified words in each of the respective sets of data, which may be understood to be in digital form, may be provided to the third node 113, or stored, in digital form, in a computer memory as a respective plurality of words and a respective second KG for each set of data of the plurality of sets of data. In this Action 403, the first node 111 may obtain, for each set of data of the plurality of sets of data, the respective plurality of words and the respective second KG. The respective second KG may correspond to a respective oracle KG of the respective set of data, and the respective plurality of words may correspond to a respective oracle plurality of words. The respective oracle KG may therefore be understood to be the true KG, which may be understood as the true output label. The oracle KG may be understood as the KG corresponding to the meaning of the input speech signal, and it may be in a different language too. ^^ and ^^1, ^^2, ... , ^^^^ may denotes the oracle graph, that is the respective second KG,and words, that is, the respective plurality of words, corresponding to the input speech signal, respectively. In some embodiments, the respective second KG may be received from the third node 113 operating in the computer system 100. By, in this Action 403, obtaining the respective plurality of words and the respective second KG for each set of data of the plurality of sets of data, the first node 111 may be enabled to use the obtained the respective plurality of words and the respective second KG for training of the artificial neural network 121, as will be described in the next Action 404, thereby enabling to provide the respective second KG as a KG that may be devoid of irrelevant words and disfluencies present in the corresponding speech as the ground truth KG. This may enable the artificial neural network 121 to learn to eliminate the effects of these noises present in the speech signal during KG creation for a respective set of data. Action 404 In this Action 404, the first node 111 may train the artificial neural network 121 to determine a respective first KG for each respective set of data using the obtained plurality of sets of data, the respective plurality of words and the respective second KG. The training inthis Action 404 uses, as input, the respective set of features, e.g.,  ^^2 ,  .  .  . ,  ^^^^, extractedfrom the respective set of data. The artificial neural network 121 may have an end-to-end neural network (NN) architecture. An end-to-end architecture may be understood as a single artificial neural network which may directly predict a KG from input data, and it may be trained to explicitly minimize the KG prediction error. A non-limiting example of the end-to-end neural network architecture for KG prediction according to embodiments herein is depicted in Figure 8. Embodiments herein may be understood to use a new end-to-end NN framework which may be referred to herein as a Listen, Attend, and Draw (LAD) model, which may automatically create KG, in this Action 404, the respective first KG, from speech signals, in this Action 404, the respective set of data, where an NN module, e.g., with Long Short-Term Memory Networks (LSTM) / Recurrent NN (RNN) stacks and attention mechanism, may be directly predicting KG, in this Action 404, the respective first KG, from the speech signal, in this Action 404, the respective set of data. In the LAD module, an encoder-decoder architecture may be employed to obtain, in this Action 404, the respective first KG from the speech signal, that is, the respective set of data. Particularly, the artificial neural network 121 may transform the obtained features extracted from the data in a set, which may be referred to as the speech signal features, into a latent representation by using an encoder. A latent representation may be understood as a representation of data in a transformed domain. For example, the latent representation may be a lower-dimensional representation or another type of transformation. In embodiments herein, the encoder may generate the appropriate latent representation. The encoder may be generated by one of: a bidirectional Gated Recurrent Unit (bi- GRU), bidirectional Long Short-Term Memory Networks (bi-LSTM) and encoder blocks available in Transformer architecture. The artificial neural network 121 may comprise a decoder. The decoder may take as second input the latent representation of the data, map the latent representation to words, e.g.,  ^̂^2,  .  .  .  ,  ^̂^^^  , and output the first KG by using a neural network having a capability tomodel sequences of data. The decoder may use the latent representation through an attention mechanism to generate the words corresponding to the meaning of the speech signal, e.g.,  ^̂^2,  .  .  .  ,  ^̂^^^, and may thereby be enabled to draw, in this Action 404, the final respectivefirst KG. For each set of data, the encoder module may consider the obtained features extractedfrom the data, that is, the input short-time speech features, e.g., ^^1  ,  ^^2 ,  .  .  . ,  ^^^^, and maygenerate embedding ^^ = [ℎ1   ℎ2 ,  .  .  . ,  ℎ^^′]. Short-time speech features may be any standardspeech features, e.g., Wave2vec [2], may be used as features. The embedding may be performed using one of the bi-GRU, the bi-LSTM and the encoder blocks available in Transformer architecture, with subsampling. The decoder may then attend the embedding ^^ and generate the necessary informationfor word prediction and graph creation.  ^̂^2,  .  .  .  ,  ^̂^^^   may be understood to represent theestimates of words in the respective first KG, the output ^̂^^^,^^may be understood to refer to a two-dimensional vector, that may be understood to represent the relation of word ^^ with word ^^ . That is, [1, 1] may be understood to represent the word ^^ is connected to word ^^ and the word ^^ is a node, [1,0] may be understood to represent word ^^ is an edge and it is connected to the word ^^ , [0,0] and [0,1] may be understood to represent ^^ should not be considered for the graph creation. The functional relationship between the variables in the decoder may be provided in a non-limiting example of embodiments herein, by the following equations:^^^^  = ^^^^^^^^^^^^^^^^^^(^^^^ ,  ^^)wherein: ^^^^may be understood to be an NN hidden layer representation; ^^ may be understood to be the embeddings obtained by the encoder, encoder generated latent representations; ^^^^  may be understood as a representation obtained by a linear combination of embeddings ^^, based on attention weights calculated based on ^^^^and ^^; ^^^^^^^^^^^^^^^^^^ may be understood to compute the linear combination of embeddings ^^ , based on attention weights calculated based on ^^^^and ^^;^^^^ = ^^^^^^^^ − ^^^^^^^^^^1(^^^^−1,  ^̂^^^−1,  ^^^^−1),wherein: ^^^^^^^^ may be understood to refer to Long short-term memory network;^^^^−1 may be understood to be the ^^ − 1 th state of the LSTMs in LSTM-Stack1;^̂^^^−1 may be understood to be the ^^ − 1 th word predicted;^^^^−1 may be understood to be the ^^ − 1 th state of the LSTMs in LSTM-Stack1;^^^^^^^^^^1 may be understood to refer to an LSTM or more than one LSTM connected in cascade.^^(^̂^^^ ;  ^̂^<^^,  ^^) = ^^^^^^ − ^^^^^^^^^^^^^^(^^^^,  ^^^^),A Feed forward Network followed by a Softmax, may provide the probability of word connections. wherein: ^̂^^^may be understood to be the ^^ th word predicted; ^̂^<^^may be understood to be all predicted words before the ^^ th word; ^^ may be understood to be input to the NN architecture, that is, the short-time speech or short- time speech features; ^^^^^^ may be understood to refer to Multi-layer perceptron, or a cascade of dense NN layers with non-linear activations; ^^^^^^^^^^^^^^ may be understood to refer to an activation function that may scale numbers / logits to probabilities. wherein: ^^^^may be understood to be ^^ th state of the LSTMs in LSTM-Stack2; ^^^^+1may be understood to be input to the LSTM stack 2; ^^^^+1may be understood to be input to the LSTM stack 2; ^^^^−1 may be understood to refer to the ^^ − 1 th state of the LSTMs in LSTM-Stack2;^^^^^^^^^^2 may be understood to refer to a second stack of LSTMs where the stack may be formed by one LSTM or connecting more than one LSTMs in cascade. ^^^^=1,^^ = ^^^^^^^^ − ^^^^^^^^^^3( 1), set initial state as and input 1= [1,1]wherein: ^^^^=1,^^may be understood to be the first output of the LSTM stack 3 for the corresponding predicted word ^^; ^^^^−1may be understood to be the initial state of the LSTM stack 3 corresponding to the predicted word ^^; ^^^^^^^^^^3 may be understood to refer to the third stack of LSTMs. wherein: ^̂^^^−1,^^may be understood to be refer to a two-dimensional vector, that may be understood torepresent the relation of word ^^ with word ^^ − 1 . That is, [1, 1] may be understood to representthe word ^^ is connected to word ^^ − 1 and the word ^^ is a node, [1,0] may be understood torepresent word ^^ is an edge and it is connected to the word ^^ − 1 , [0,0] and [0,1] may beunderstood to represent ^^ should not be considered for the graph creation; ^^^^,^^may be understood to be LSTMs output after taking the ^̂^^^−1,^^as input; ^^ may be understood to be a word index; ^^^^^^^^^^3 may be understood to refer to a third stack of LSTMs where the stack may be formed by one LSTM or connecting more than one LSTMs in cascade.^̂^^^,^^ = ^^^^^^^^^^^^^^(^^^^,^^)wherein: ^̂^^^,^^may be understood to be refer to a two-dimensional vector, that may be understood to represent the relation of word ^^ with word ^^ . That is, [1, 1] may be understood to represent the word ^^ is connected to word ^^ and the word ^^ is a node, [1,0] may be understood to represent word ^^ is an edge and it is connected to the word ^^ , [0,0] and [0,1] may be understood to represent ^^ should not be considered for the graph creation. are dense layers, and ^^^^,^^ = wherein: ^^ may be understood to be an index; ^^^^,^^may be understood to be an attention value, e.g., a value between 0 and 1, which may be considered as a probability value; ℎ^^may be understood to be a vector which may be the ^^ th column of the encoder output ^^; ^^(^^^^) may be understood to be a function applied on ^^^^; ^^(ℎ^^) may be understood to be a function applied on ℎ^^; ^^^^,^^may be understood to refer to an inner product between ^^(^^^^) and ^^(ℎ^^). During the supervised training stage of the artificial neural network 121, the input may be understood to be the speech signal, that is, a set of data of the sets of data, and the respectiveoutput label ^^^^,^^ , where ^^ = 2,3, ... , ^^ , ^^ = 1, 2, ... , ^^ − 1 may be understood to represent theoracle, true, KG, that is, the respective second KG obtained in Action 403 corresponding to that set of data. This may be understood to be repeated, iteratively, for the whole plurality of sets of data being input to the artificial neural network 121 to train it in this Action 404. ^^ may be understood to denote the number of words used to represent the meaning of the speech signal / number of words in the respective second KG. The training in this Action 404 may be performed by minimizing a loss function representing a difference between each respective first KG and the obtained respective second KG. The first node 111 may use a binary cross entropy as loss for the KG prediction part, and, for the word prediction part, the first node 111 may use categorical cross entropy loss. The combined loss for a training sample may become: log ^̂^^^,^^ − (1 − ^^^^,^^) log(1 − ^̂^^^,^^)) wherein: ^^ may be understood to denote the total number of words in a dictionary available to the first node 111, the dictionary comprising all the words in the language of the respective first KG, ^^ may be understood to be an index over the dictionary elements and ^^ may be understood to be an index over the number of words in the given speech; , and^^ > 0.The total loss, that is, the average of ^^ over all the training samples, may be minimized to obtain the optimum weights for the artificial neural network 121. During the training in this Action 404, the first node 111 may give input to the LSTM-STACK-1 and LSTM-STACK-3, for^^ > 1, through a Teacher Forcing [3] method. That is, the model output obtained at a previoustime step may be provided as input at the next time step. The training in Action 404 may be performed iteratively until the loss function may be below a threshold. That is, the first node 111 may iterate the same procedures for each set of data of the obtained plurality of sets of data. During training the parameters of the artificial neural network 121 may be optimized to minimize the error between the predicted respective first KG and the actual KG, that is, the second KG. Any standard optimization algorithm such as Adam may be used to obtain the parameters of the artificial neural network 121. The objective of the artificial neural network 121 may be understood to be to construct a KG from conversation speech data, which may contain many noisy words, grammatical issues, disfluencies, etc. For this objective, a direct KG construction method such as that described for embodiments herein, may be understood to be beneficial since it may be understood to directly minimize the error between the second KG, that is, the oracle graph ^^^^,^^, and the respective first KG, that is, the predicted graph ^̂^^^,^^, through a supervised learning strategy. This training strategy may be understood to enable the end-to-end artificial neural network 121 to learn the patterns present in the speech and reduce the effect of grammatical mistakes and disfluencies while predicting the respective first KG, from speech data. Since the artificial neural network 121 may be understood to learn the relation between the respective set of data and the corresponding respective first KG through a supervised training, by minimizing the error between the predicted respective first KG and the respective second KG obtained in Action 403 for the same respective set of data, it may be possible to use the artificial neural network 121 to better exploit all the relevant information present in the speech signal, including non-linguistic and paralinguistic, to obtain the accurate respective first KG. Hence, the respective first graph generated using the proposed formulation may be understood to be accurate even, in the case of conversational speech. Hence, the first node 111 may be enabled to construct a respective first KG from unstructured, conversation speech data, even if it may contain many noisy words, grammatical issues, etc. By, in this Action 404, training the end-to-end artificial neural network 121 to determine the respective first KG for each respective set of data by minimizing the loss function representing the difference between each respective first KG and the obtained respective second KG, the first node 111 may be enabled to directly explicitly minimize the error between the obtained respective second KG, that is, the oracle, true, KG, and the predicted KG through a supervised learning strategy. Since the first node 111 may be understood to perform the direct optimization of the error, this may be understood to automatically enable the artificial neural network 121 to learn all the relevant information present in the speech for predicting the correct KG, that is, the respective first KG. Since, as stated earlier, the conversational speech may be understood to be difficult data, because of grammatical errors, presence of disfluencies, etc., the artificial neural network 121 may be automatically enabled to learn all the relevant information present in the speech signal for the correct prediction of the graph. That is, this training strategy may enable the artificial neural network 121 to learn the pattern of the grammatical errors, lack of structure, and disfluencies present in the conversational speech data and help to reduce its effect during the inferencing stage. The respective first KG and the respective plurality of words may be understood to represent the meaning of the speech signal, which may be in a different language. In other words, the respective plurality of words and the respective first KG may be from a different language than the input speech signal. In some embodiments, the training in Action 404 of the artificial neural network 121 may comprise using as input the plurality of sets of data, comprising sets of data in conversational speech of different respective first languages, and using for each set of data of the plurality of sets of data, a respective second KG in a different second language than the respective first language. Since the input to the artificial neural network 121 may be understood to be a speech signal and the output used for supervised training may be understood to be the oracle graph corresponding to the meaning of the speech signal, embodiments herein may be used for cross-language speech-to-KG creation and also in the scenario where no script may be available for the speech signal. For the cross lingual scenario, the language of the respective first KG and the respective plurality of words may be different from the input speech signal language. For example, the case may be considered where the first node 111 may need to predict a respective first KG in English where the input speech signal may be in French. To train the artificial neural network 121 for cross-lingual KG prediction, the first node 111 may need to create data samples corresponding to these training scenarios. For the above discussed scenario, the input may be a speech sample in French and the output may be the respective first KG in English, and the respective plurality of English words, which may capture the meaning of the speech signal uttered in French. A non-limiting example of an input speech signal sample in French may be: "Eirene a un bateau, ouais, ouais, elle va pour pêcher, elle attraper du thon, mm mm, elle vivre en Israël où le thon est plus beaucoup cher". The corresponding English translation provided here for the benefit of the reader to understand the context may be understood to be “eirene has a boat yeah yeah she goes for fishing she catch tuna fish mm mm she live in Israel where tuna is costlier". It may be understood that the first node 111 may not be required to perform any translation according to embodiments herein. In multi-lingual KG creation, the goal may be understood to be to find a first KG in the given target language where input to the KG creation module, that is, to the artificial neural network 121, may be speech signals from different languages. For example, in a non-limiting example, the input speech signals in the sets of data may be from the different respective first languages French, English, Hindi, and Swedish, and the output respective first KG created from these speech signal may need to be in English. Embodiments herein may be understood to ultimately enable an automatic multi-lingual creation of a corresponding KG. In such a multilingual scenario the first node 111 may need to perform the training in this Action 404 of the artificial neural network 121 using the data where input may be in a speech signal from different languages, and the output label may be understood to be the corresponding respective first KG in the given target language. Consider the example multi-lingual scenario, where the input speech signal may be from the first languages French, English, Hindi, and Swedish, and the goal may be to obtain the corresponding first KG in English. Then to achieve this goal, the artificial neural network 121 may need to be trained with the plurality of sets of data where the training samples may need to have the form, {Input: French Speech Signal , Output: English KG}, {Input: English Speech Signal, Output: English KG}, {Input: Hindi Speech Signal , Output: English KG}, and {Input: Swedish Speech Signal , Output: English KG}. During inferencing, the first node 111 may obtain in Action 405 speech signal from any of these four languages as an input to the artificial neural network 121, the artificial neural network 121 may then produce the corresponding KG in English as output in Action 407. In some embodiments, the first node 111 may refrain from training the artificial neural network 121 in Action 404. The trained artificial neural network 121 may be received in a first indication from the second node 112 operating in the computer system 100. By the first node 111 training the artificial neural network 121 to determine the respective first KG, wherein the artificial neural network 121 may for each respective set of data using the obtained plurality of sets of data and the respective plurality of words, wherein the artificial neural network 121 may use the obtained features extracted from the data refraining from converting the data into text, the first node 111 may be able to generate KG from speech signal directly where the input to the artificial neural network 121 may be speech signal and output may be a KG corresponding to the meaning of speech signal. By the first node 111 training the artificial neural network 121 according to embodiments herein, the first node 11 may enable the cross-lingual and multi-lingual KG creations and may help the semantic communication accessible to resource-poor languages, wherein an intermediate step conversion of the speech to text may not even be possible or very costly. In addition, the respective first KGs, after training, may be robust to the language errors in the speech as the firs node 111 may use speech embeddings and the second KG, that is, the correct oracle graph, for the supervised learning. Since there is no cascading of different modules, by the first node 111 training the artificial neural network 121 according to embodiments herein, refraining from converting the data into text, the first node 111 may enable to avoid error compounding present in the traditional approaches and the direct minimization of KG prediction error may enable the artificial neural network 121 to learn all relevant features present in the speech signal, which may be understood to lead to a more accurate KG. Action 405 Once the first artificial network 121 may have been trained, the first node 111 may use it to make predictions on newly obtained data, that is, the first node 111 may start an inferencing phase once the training phase may have concluded once an accuracy of the artificial neural network 121 may have reached a desired level. Accordingly, in this Action 405, the first node 111 obtains data encoding conversational speech. That is, the first node 111 may obtain new conversational, unstructured data, for which the first node 111 may want to obtain a KG. The data obtained in this Action 405 may be understood to be equivalent to any set of data obtained in Action 401, but this time it may be understood to be new data. The data may be obtained, e.g., from the device 140, in a similar way as to how it was obtained in Action 401. Action 406 In this Action 406, the first node 111 obtains features extracted from the data. The features may be short-term features, refraining from converting the data into text. The features obtained in this Action 406 may be understood to be equivalent to the any of the respective sets of features obtained in Action 402, but this time it may be understood to be for the new data. That is, the extraction of features in this Action 406 may be performed in a similar manner than that described for Action 402. Hence, obtaining in this Action 406 may comprise receiving from another node in the computer system 100, for example, from the device 140, via the radio network node 130, fetching or retrieving from a memory, or the first node 111 extracting itself the features from the data obtained in Action 405. As stated earlier, the artificial neural network 121 may be understood to have an end-to- end neural network architecture, wherein the artificial neural network 121 may transform the obtained features extracted from the data into the latent representation by using the encoder. The artificial neural network 121 may comprise the decoder, wherein the decoder may take as second input the latent representation of the data, map the latent representation to words, and output the first KG artificial neural network 121 by using the neural network having the capability to model sequences of data. Action 407 In this Action 407, the first node 111 determines, using the artificial neural network 121, a first KG. The first KG represents words in the conversational speech and relations of the words. The determining in this Action 407 uses, as first input, the obtained features extracted from the data. The first KG determined in this Action 407 may be understood to be equivalent to a respective first KG, as described before, but for the new data obtained in Action 405. Moreover, the first KG may have been determined by the first node 111 using the artificial neural network 121, once the artificial neural network 121 may have been trained. The artificial neural network 121, once trained, may, in this Action 407 take the speech features as input and the output the corresponding text and KG. Determining may be understood as calculating, estimating, deriving, or similar, or obtaining or receiving from another node. The determining in Action 407 of the first KG may be performed using the trained artificial neural network 121 from Action 404. The conversational speech may be in a first language and the first KG may be in a second language. As explained earlier, in regard to Action 404, embodiments herein may allow to generate the first KG in a different language than the one in present in the speech signal. That is, the input speech signal and corresponding predicted KG may be in two different languages. For example, input may be provided to the first node 111 as speech signal of Hindi, a language present in India, and generate the corresponding first KG in English, where the first KG in English may be understood to correspond to the meaning of the input speech signal. This cross-language speech-to-KG may be understood to have two important applications in the context of semantic communication. First, the first node 111 may be able to generate the first KG for a language where there may be no script available for the speech. For example input may be provided to the first node 111 as Konkini speech signal and the output first KG may be in English. A second advantage may be understood to be that the first node 111 may be able to generate the first KG in a language which may be popular, for example English, for the speech signal from a non-popular language. This may be understood to be advantageous since a KG generated in a common language such as English may be be a good source to represent and implement for any tasks, new tasks also. That the conversational speech may be in a first language and the first KG may be in a second language may be understood to be additionally advantageous as it may enable to avoid using a translation framework applied to a cascaded approach of existing methods, since the amount of effort needed for obtaining data by annotation would be very large in such cases as it may be understood to require expert annotation in both, speech-to-text, e.g., for the ASR type data, and text-to-text, for the Machine translation type data creation. Furthermore, such an approach may be understood to not be able to eliminate the error compounding effect described in the Summary Section. In this Action 407, the first node 111 may, in some embodiments, use the artificial neural network 121 it may have trained itself as described in the previous actions. Alternatively, the first node 111 may used the artificial neural network 121, as trained by the second node 112. In such embodiments, the first node 111 may receive a second indication from the second node 112. The second indication may indicate the trained artificial neural network 121. That is, Actions 401-404 may be understood to be performed optionally by the first node 111. By determining the first KG using the artificial neural network 121 in this Action 407, the first node 111 may be enabled to automatically obtain, e.g., predict, an accurate KG from speech signal directly from speech, without any intermediate conversion process. This may be achieved by the end-to-end NN-based speech-to-graph creation module of the first node 111, which may be able to, once trained, understand the meaning of the speech signal and generate the first KG corresponding to the meaning of the speech signal. This may be achieved through the supervised training of the artificial neural network 121 in Action 404, where the input may be understood to be the speech signal, and the output may be understood to be the respective first KG constructed from the meaning of the speech signal represented in any language, e.g., may be the same language also. Hence, the first node 111 may enable the creation of the first KG in cross lingual speech-to-KG creation scenarios and also the case where languages may have no script available. In fact, the first node 111 may make the KG creation accessible to resource poor languages. Moreover, by directly operating on speech signal, the first node 111 may be enabled to exploit all the relevant information present in the speech signal to predict the first KG, in addition to the linguistic information such as the domain specific attributes, and paralinguistic features. By directly operating on speech signal, the first node 111 may be enabled to predict the first KG, devoid of errors present in conventional cascaded approach, such as errors in a speech-to-text module and modules present in the conventional text-to-graph creation process and its error compounding problems. The approach used may be understood to be robust against grammatical mistakes and disfluencies. Furthermore, the approach may be understood to be agnostic to the grammatical mistakes of the user. The first node 111 may be understood to automatically handle the issues with the speech on mis-spelled words, fillers and disfluencies. The approach used may be understood to be robust to lack of entities and relations in the conversational inputs modality. By directly minimizing the error between the predicted graph and the oracle graph during training, the first node 111 may be enabled to exploit the domain specific global and local linguistic structures present in the data to predict the accurate first KG. Furthermore, the approach may be understood to be agnostic to the domain, making it easier to use. Action 408 In this Action 408, the first node 111 initiates outputting an indication of the determined first KG. The indication output in this Action 208 may be understood to be a second indication. Initiating may be understood as triggering, enabling, beginning, or similar. Outputting in this Action 408 may be understood as displaying, e.g., on an interface or sending, for example, via the a link to another node. The second indication may be for example, as depicted in the non-limiting example of Figure 9. In another example, the second indication may be an interface through which the determined first KG may be searchable. For the search, standard searching techniques such as elastic search, etc. may be used. By initiating outputting, the (second) indication in this Action 408, the first node 111 may be enabled to share the outcome of the determination of Action 407 efficiently with another node or among users, enabling to achieve the advantages already explained in relation to Action 407. This will now be illustrated with a few use cases. Embodiments herein may have wide variety of use cases, and accordingly, the second indication may be outputted in one way or another, based on the use case. The end-to-end approach described in embodiments herein may be understood to have a wide range of applications in semantic communication and natural language processing tasks. Next, different non-limiting examples of use cases are described to illustrate the usage of embodiments herein. The (second) indication may be outputted in one of the following. Example Use case 1: Connected Healthcare: Intelligent way to process conversations According to a first option, the (second) indication of the determined first KG may be outputted in a healthcare application. In such embodiments, the conversational speech may describe a condition of a patient. Most hospitals have a four-step process to understand the condition of a patient and dispatch an ambulance. As the request may be typically received via a telephone call to the hospital, it may usually be a caller agent that may take the details of the request, such as location, seriousness of the situation, e.g., accident, and patient details, subject to the medical condition of the patient, or in case an attendee to the patient may be able to provide. Based on the location, which may be a Global Positioning System (GPS) enabled system, the nearest ambulance may be dispatched to the site. In these steps, one of the most important aspects may be understood to be to understand the condition of the patient from the patient, or from the attendee, who may be in a state of distress or panic. The initial conversation may be understood to be the most important in order to understand the condition of the patient. However, it has been found that in many cases it is very difficult to understand the symptoms explained by the patient or the attendee due to multiple reasons, such as network quality, the ability to explain the real problem, the state the caller may be in, e.g., in a panic, and may not be clear with speech. Many times lingual factors may also play a role in understanding the real picture. Additionally, and with such impediments, the caller may have to repeat the same set of questions / answers, first to the caller agent, and then to a doctor who may then identify which type of ambulance may need to be deployed in this scenario. The approach described embodiments herein may now be used to generate KG from the speech, and thereby enable to understand the problem described better. A KG may be generated based on the discussion between patient or patient attender with different persons in the hospital, e.g., front desk employee, supervisor, doctor, etc. The information shared earlier with the emergency responder may be shared with the doctor, instead of the patient explaining, in form of the graph. A doctor or nurse may now be able to access the second indication output according to embodiments herein, and go through this KG to understand the problem affecting the patient better, which may help to provide a better diagnosis. Optionally, to make the communication in an intelligent way, conversation analytics may be performed in sematic-aware networking and at the time of response, any biases may be avoided, and the exact knowledge may be advantageously enabled to be shared with the requested parties. Finally, the first node 111 may, according to Action 408, share the generated first KG among different parties, which may be understood to facilitate information sharing and enable to provide better treatment. For example, sharing the second indication may enable to setup the equipment in a laboratory in advance, before the arrival of the patient to the hospital. Since we are a globalized society callers may have different mother tongue, hence there are scenarios where the different callers may be talking in different languages. A cross- lingual / multi-lingual KG creation algorithm helps to create KG in a language which is comfortable for service providers making the semantic communication an easier task. Example Use case 2: Tele-operated Driving: 5G Connected and Automated Driving According to a second option, the (second) indication of the determined first KG may be outputted in a tele-operated driving application. In such embodiments, the conversational speech may describe instructions to a vehicle by a remote driver. Tele-operated Driving (ToD) may be understood to be an application that may operate with different configurations and Quality of Service (QoS) levels. ToD may be understood to enable a remote driver to control a vehicle through proper conversations. The environmental conditions as observed by sensors in the vehicle may be transferred to the remote driver as perception data. In such a setup, it may be possible to understand and develop the conversation through sensing through different sensors in a vehicle. The remote driver may use the sensors details and operate the driving either via text inputs or speech input. To realize ToD, data may be exchanged through a cellular network. The appropriate configuration of the perception may be understood to be needed for enabling the remote control of the vehicle at a command center (CC). Different sensor outputs in a vehicle, such as speech texts, may be used to generate a KG regarding the specific issue the vehicle may have encountered, or for the solution of the issue. Embodiments herein may therefore enable to avoid the occurrence of accidents through error-free graph generation process. Example Use case 3: Vehicular-to- Everything Communication According to a third option, the (second) indication of the determined first KG may be outputted in a vehicle-to-everything application. In such embodiments, the conversational speech may describe vehicle-related problems between stakeholders. The anticipated development of autonomous driving and connected vehicles will bring numerous vehicle-to-everything (V2X) applications, including both safety services and entertainment-related services which may be provided through conversation. Communication in vehicles may be understood to be used in many forms. This may comprise the traditional cellular mobile connectivity utilizing 3GPP technologies 2G, 3G, and 4G LTE. For the past decade, vehicles have been including direct connectivity, known as vehicle-to-everything (V2X). V2X may be understood to enable vehicles to communicate with other road participants, e.g., other vehicles, pedestrians, etc. In such a context, the problem may be understood to be how to establish the conversation intelligently to solve the problems, and also to avoid any biases in the procedure. The first node 111 of embodiments herein may be introduced as an end-to-end module to attend any vehicle related problems on the route through robust semantic communication established between different stakeholders, such as repair shops, payments, required parts, etc. Through search in the KG, an FSO may then get the details that may be necessary to sort out the issue. Example Use case 4: FSOs Solving Network Problems in Customer Site According to a third option, the (second) indication of the determined first KG may be outputted in a field service operation application. In such embodiments, the conversational speech may describe a problem to be solved at a customer site. In the present servicing industry, a solution document may be usually provided to Field Service Operations (FSOs) to apply the knowledge, that is, the explanation and relatedness of the problem, with their experiences to shorten the life cycle of addressing the issues. These FSOs may not be proficient in English and may have different accents different. Also, they may be conversing in a language different from English. A search may be based on the KG which may be generated from existing documents and different use cases. To give the better search results to the FSOs, it may be beneficial to construct a KG from the speech. Since grammatical errors may be made, the KG creation may be understood to need to be immune to the grammatical errors present in the text. Hence, embodiments herein may be used to arrive at the KG, which may be further used to generate search results. Also, some problems may be handled with different FSOs together. In such scenarios, information sharing among FSOs may be understood to be important to solve the problem in finite time. In such a scenario, embodiments herein may also help in solving the problem in earlier time. Whenever a first FSO, or whoever may solve the problem earlier, may input speech signal as input, the speech signal may be translated to KG. This KG may then be stored in a database. A second FSO may use the KG and understand the problem. The end- to-end system of embodiments herein may be completely utilized to enhance the performance of FSOs during field visits. Embodiments herein may enable to solve the problem faster and effectively. Also, the information storage in a KG format may be a vital data to solve future problems through easy search and retrieval methodology. Even in situations where the FSO’s may be conversing in different languages than the target language for KG creation, for example, English, the approach of embodiments herein may enable the KG creation in target language. The sequence diagram of a non-limiting example of the use case 4 is shown later, in Figure 8. Example Use case 5: Solving problem at Telecommunication provider sites In another use case according to the third option, the customer site may be a telecommunication provider site. Typically, the customer may inform the problem to the telecommunication provider engineer, and depending on the type of problem, the engineer may raise a ticket to solve it. During the ticket creation, the telecommunication provider engineer may give a description of the problem and he / she may have more details on the problem. In such a scenario, the ticket may be assigned to another engineer, and he / she may read only the description of the problem to understand the issue. The solution may take time, as the description of the problem may not have all the details on what the first engineer attending to the issue may have had. According to embodiments herein, details may be shared between engineers, so that the problem may be solved faster and in an efficient way. In this process, multiple sub-graphs may be connected to arrive at a suitable solution to close the ticket asap. In some examples, the KG may not be visible to a mechanic. The mechanic may only search through the KG to understand the problem and arrive at a solution. Embodiments of a computer-implemented method, performed by the second node 112, will now be described with reference to the flowchart depicted in Figure 5. The method is for handling the respective first KG. The second node 112 operates in the computer system 100. Several embodiments are comprised herein. In some embodiments all the actions may be performed. In some embodiments, some embodiments of the actions may be optional. It should be noted that the examples herein are not mutually exclusive. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the second node 112, and will thus not be repeated here. For example, the data may encode recordings of conversations. Action 501, 502, 503 and 504 may be performed by the second node 112 as they have been described for the first node 111, respectively in Action 401, 402, 403, and 404. Action 501 As was mentioned earlier in relation to the description of Figure 1, the second node 112 may be understood to be a node having a capability to train the artificial neural network 121. In this Action 501, the second node 112 obtains the plurality of sets of data. Each set of data encodes the respective conversational speech. Action 502 In this Action 502, the second node 112 obtains, from the plurality of sets of data, the respective plurality of sets of features extracted from the respective set of data, refraining from converting the data into text. Action 503 In this Action 503, the second node 112 obtains, for each set of data of the plurality of sets of data, the respective plurality of words and the respective second KG. The respective second KG corresponds to the respective oracle KG of the respective set of data, and the respective plurality of words corresponds to the respective oracle plurality of words. Action 504 In in this Action 504, the second node 112 trains the artificial neural network 121 to determine the respective first KG for each respective set of data, using the obtained plurality of sets of data, the respective plurality of words and the respective second KG. The training in this Action 504 is performed by minimizing the loss function representing the difference between each respective first KG and the obtained respective second KG. The training in this Action 504 is performed iteratively until the loss function is below the threshold. In some embodiments, the artificial neural network 121 may have the end-to-end neural network architecture. The artificial neural network 121 may transform the obtained respective plurality of sets of features extracted from the respective set of data into the respective latent representation by using the encoder. In some embodiments, the encoder may be generated by one of: the bidirectional Gated Recurrent Unit, the bidirectional Long Short-Term Memory Networks and the encoder blocks available in Transformer architecture. In some embodiments, the artificial neural network 121 may comprise the decoder and the decoder may take as second input the respective latent representations of the data, map the respective latent representations to respective words, and output the respective first KGs by using the neural network having the capability to model the sequences of data. In some embodiments, the training in this Action 504 of the artificial neural network 121 may comprise using as input the plurality of sets of data, comprising the sets of data in conversational speech of different respective first languages, and using for each set of data of the plurality of sets of data, the respective second KG in the different second language than the respective first language. Action 505 In this Action 505, the second node 112 initiates outputting the first indication of the trained artificial neural network 121. In some embodiments, at least one of the following may apply. According to a first option, the respective second KG may be received from the third node 113 operating in the computer system 100. According to a second option, the initiating in this Action 505 of the outputting of the first indication may comprise sending the first indication to the first node 111 operating in the computer system 100. The first indication of the trained artificial neural network 121 may be outputted in one of: i) the healthcare application, and the conversational speech may describe the condition of the patient, ii) the tele-operated driving application, and the conversational speech may describe instructions to the vehicle by the remote driver, iii) the vehicle-to-everything application, and the conversational speech may describe vehicle-related problems between stakeholders, and iv) the field service operation application, and the conversational speech may describe the problem to be solved at the customer site. Figure 6 is a schematic diagram depicting a non-limiting example of embodiments herein, wherein a direct approach may be used for generating the first KG from speech. As indicated in Figure 6, a speech signal obtained according to Action 405 may be input the artificial neural network 121 for the determination performed in accordance with Action 407, and the output in accordance with Action 408 may be the corresponding KG, referred to herein as the first KG. Figure 7 is a schematic diagram depicting a non-limiting example of the direct approach of embodiments herein, to obtain the KG from the speech signal, that may be followed according to embodiments herein, by any of the first node 111 and the second node 112 during the training phase. As depicted in Figure 7, the framework of embodiments herein may comprise to convert speech to short-time features, the respective sets of features, such as MFCC, wave-2-vec [2], etc. in accordance with Action 502 and Action 402. The end-to-end artificial neural network 121 may then, take the speech feature as input and in accordance with Action 504, Action 505, Action 404 and Action 405 output the respective first KG, ^̂^, and the corresponding text, that is,the estimates of words in the respective first KG,  ^̂^2,  .  .  .  ,  ^̂^^^  . To obtain the respective firstKG from speech, the first node 111, or the second node 112, may employ the artificial neuralnetwork 121, which may predict the first KG from the speech signal, where ^^ and ^^1, ^^2, ... , ^^^^denote the oracle graph, that is, the respective second KG, and the respective plurality of words corresponding to the input speech signal, respectively. As explained earlier, embodiments herein provide a NN based approach to directly predict the KG from speech signal, where input to the NN may be a speech signal and the output may be a KG corresponding to the meaning of the speech signal. During training in accordance with Action 504 and Action 404, NN parameters may be optimized to explicitly minimize the error between the predicted KG, ^̂^, and the oracle KG, ^^, that is, the true KG, obtained according to Action 503 and Action 403, which may be understood as the true output label. The oracle KG may be understood as the KG corresponding to the meaning of the input speech signal, and it may be in a different language too. Figure 8 is a schematic diagram depicting a non-limiting example of the architecture of the end-to-end artificial neural network 121 of embodiments herein, where an attention mechanism and hierarchical RNNs may be employed to predict the first KG from the speech signal, according to Listen, attend and draw (LAD) model. The dotted boxes contain the LSTM / GRU / RNNs. The encoder 801 may generate the appropriate latent representation and the decoder 802 may use this latent representation through an attention mechanism to generate the words corresponding to the meaning of the speech signal and draw the graph. The words and first KG may be from a different language than the input speech signal. In Figure 8,^^1  ,  ^^2 ,  .  .  . ,  ^^^^ represents input speech features 803. Any standard speech features, forexample use Wave2vec [2] as features may be used.  ^̂^2,  .  .  .  ,  ^̂^^^  represent the estimatesof words 804 in the first KG. The output ^̂^^^,^^, a two-dimensional vector, represents the relation 805 of word ^^ with word ^^ . That is, [1, 1] represents the word ^^ is connected to word ^^ and the word ^^ is a node, [1,0] represents word ^^ is an edge and it is connected to the word ^^ , [0,0] and [0,1] represent ^^ should not be considered for the graph creation. The dotted boxes indicate the unidirectional RNN / LSTMs stack. The encoder module 802 may consider the input short- time speech features 803 and may generate the embedding ^^ 806. The decoder 802 may attend the embedding ^^ 806, and generate the necessary information for word prediction and graph creation according to the equations providing the functional relationship between the variables in the decoder 802, already provided in the description of Action 404:^^^^  = ^^^^^^^^^^^^^^^^^^(^^^^ ,  ^^)^^^^ = ^^^^^^^^ − ^^^^^^^^^^1(^^^^−1,  ^̂^^^−1,  ^^^^−1),^^(^̂^^^ ;  ^̂^<^^,  ^^) = ^^^^^^ − ^^^^^^^^^^^^^^(^^^^,  ^^^^),(Feed forward Network followed by a Softmax, gives the probability of words)^^^^ = ^^^^^^^^ − ^^^^^^^^^^2(^^^^+1,  ^^^^+1,  ^^^^−1)^^^^=1,^^ = ^^^^^^^^ − ^^^^^^^^^^3( 1), set initial state as and input 1= [1,1] During the supervised training stage of the artificial neural network 121, the input may beunderstood to be the speech signal, and the output label ^^^^,^^ , where ^^ = 2,3, ... , ^^ , ^^ = 1, 2,... , ^^ − 1, may be understood to represent the oracle (true) KG. ^^ may be understood to denotethe number of words used to represent the meaning of the speech signal / number of words in the KG. The artificial neural network 121 may use the binary cross entropy as loss for KG prediction part and for the word prediction part one can use categorical cross entropy loss. The loss for a training sample may become: log(1 − ^̂^^^,^^)) where ^^ denotes the total number of words in the dictionary and ^^ > 0. The total loss, theaverage of ^^ over all the training samples, may be minimized to obtain the optimum weights of the artificial neural network 121. During the training, the first node 111 may needs to give inputto the LSTM-STACK-1 and LSTM-STACK-3 ( for ^^ > 1) through Teacher Forcing [3] method.In Figure 8, for sequence modeling done using RNN / LSTM, the artificial neural network 121 may use alternatively Transformers encoder units [1] also. Also, another alternative may be, in the proposed NN architecture, the artificial neural network 121 may use Transformer architecture [1] also in the place of the ENCODER and LSTM-STACK-1. SOS may be understood to indicate start of the sequence. EOS may be understood to indicate end of the sequence. Figure 9 is a schematic diagram depicting a non-limiting example of an oracle KG in English generated from the French speech signal "Eirene a un bateau, ouais, ouais, elle va pour pêcher, elle attraper du thon, mm mm, elle vivre en Israël où le thon est plus beaucoup cher", where the errors or fillers are removed. In this example, the words for the English KG are ^^1= “Eirene” , ^^2= =”has” , ^^3= ”a”, ^^4=”boat”, ^^5=”She” ^^6= =”goes”, ^^7= ”for”, ^^8=”fishing”, ^^9=”and”, ^^10=”catches”, ^^11=”tuna”, ^^12=”fish”, ^^13=”she” ^^14=”lives”, ^^15=”in”, ^^16=”Israel”, ^^17=”where”, ^^18=”tuna”, ^^19=”is”, and ^^20=”costlier”. The corresponding adjacency matrices ^^ and ^^ may be understood to have a size 20x20, as the number of words is ^^ =20. The nodes ‘eirene’ and ‘boat’ are related by the word ‘has’, hence the matrix elements ^^^^^^^^^^^^^^,ℎ^^^^, ^^ℎ^^^^,^^^^^^^^^^^^, ^^ℎ^^^^,^^^^^^^^and ^^^^^^^^^^,ℎ^^^^are zero, that is, in ^^, may be understood to not be encoding the relation between ‘eirene’ and ‘boat’, that is ‘has’. On theother hand, the matrix ^^, which may be understood to correspond to the adjacency matrix ofthe directed graph, may be used to encode the relation between nodes, where the value one for the elements ^^^^^^^^^^^^^^,ℎ^^^^and ^^ℎ^^^^,^^^^^^^^may be understood to encode the relation {eirene, has, boat}. From the nonzero elements ^^^^^^^^^^^^^^,^^^^^^^^, ^^^^^^^^^^,^^^^^^^^^^^^, ^^^^^^^^^^^^^^,ℎ^^^^, and ^^ℎ^^^^,^^^^^^^^the KG branch {eirene, has, boat} may be inferred. To represent the complete graph, in ^^ , theelements ^^^^^^^^^^^^^^, ℎ^^^^ , ^^ℎ^^^^, ^^^^^^^^ , ^^^^^^^^^^^^^^, ^^^^^^^^^^ , ^^^^^^^^^^^^, ^^^^^^^^^^^^ ^^^^^^^^^^^^^^, ^^^^^^^^ ,^^^^^^^^^^, ^^^^^^ℎ^^^^^^ , ^^^^^^^^^^^^^^, ^^^^^^^^ℎ^^^^ , ^^^^^^^^^^ℎ^^^^, ^^^^^^^^ , ^^^^^^^^^^, cos ^^ ^^^^^^^^ , and^^ cos ^^ ^^^^^^^^, ^^^^^^^^^^^^ may be understood to be one and all other elements may be understood tobe zero. In ^^, values of the elements ^^^^^^^^^^^^^^, ^^^^^^^^ , ^^^^^^^^^^, ^^^^^^^^^^^^ , ^^^^^^^^ℎ^^^^^^, ^^^^^^^^^^^^ ,^^^^^^^^^^^^^^^^,^^^^^^ℎ^^^^^^,^^^^^^^^^^^^, ^^^^^^^^^^^^ , ^^^^^^^^^^^^^^, ^^^^^^^^^^^^ , ^^^^^^^^^^^^^^, ^^^^^^^^ , ^^^^^^^^^^, ^^^^^^^^^^^^ ,^^^^^^^^^^, ^^^^^^^^^^^^ , and ^^^^^^^^^^^^^^, ^^^^^^^^ may be understood to all be one and all other elementsmay be understood to be zero. The corresponding KG depicted in Figure 9 may be able to be constructed uniquely from ^^ and ^^. Figure 10 is a schematic diagram depicting a non-limiting example of the execution process of the method, that may be followed according to embodiments herein, by the first node 111 during the inference phase. The direct approach to automatically obtain the first KG from a speech utterance may involve that, first, speech obtained according to Action 405 may be transformed to short term features, according to Action 406. Then, according to Action 407, using the proposed LAD model used by the artificial neural network 121, the first KG may be determined and output according to Action 408. Figure 11 is a signalling diagram depicting a non-limiting example of the method performed by the first node 111 during the inference phase, according to embodiments herein. A first FSO, FSO 1101, may record a speech and send it to the first node 111, which may receive it in accordance with Action 405. In accordance with Action 406 and 407, the artificial neural network 121, depicted as a “KG model”, comprised in the first node 111, may convert the speech to the first KG. The first node 111 may then, in accordance with Action 408, send the constructed first KG to a database 1102. The database 1102 may be comprised in the first node 111 or may be a separate node. Saving the constructed first KG in the database 1102 may enable that at 1103, a second FSO 1104, FSO 2, may send his or her suggestion or request as input in form of speech to the database 1102. The database 1102 may then search on the KG and at 1105 send the results to the FSO2. The same process may be repeated at 1106 and 1107 for a third FSO 1108, FSO 3. As a summarized overview of the foregoing, embodiments herein may be understood to provide a new framework wherein a neural-network module may be used to, automatically, directly obtain the KG from speech data. Certain embodiments herein may provide one or more of the following technical advantage(s). Embodiments herein may be understood to solve the problems present in the conventional approaches particularly provide the following advantages. As a first advantage, since a single NN may directly predict the KG from speech, embodiments herein may be understood to enable to completely avoid the error propagation and its compounding effects present in the cascaded approach of existing methods. For example, in the conventional approach, an error occurred during the ASR is propagated to the downstream processing module. However, in embodiments herein there may be understood to be no such issue. A further advantage may be understood to be that unlike the conventional speech-to- graph creation process, that is, cascaded approach, the approach of embodiments herein may be understood to be completely devoid of the effect of errors happening in the speech-to-text conversion process. Yet another advantage may be understood to be that the end-to-end NN of embodiments herein, may have direct access to speech signal and may be enabled to exploit all the relevant information present in the speech signal, including the para-linguistic and non-linguistic information. A further advantage of embodiments herein may be understood to be that the end-to- end NN may enable creation of a KG in any language and it may handle cross-language KG creation, where the input speech and the corresponding KG may be in different languages. In the telecommunications domain, a group of Field service Professionals (FSO’s) from different origins may interact on their own languages to solve a problem and the LAD according to embodiments herein may provide a tool which may capture the conversations and understand the core approach to solve the problem and prepare a consolidated report in a universally understandable format for multinational companies. For example, NN architecture of embodiments herein may also support a multilingual scenario where input speech signals may be from different languages and the predicted KG may need to be in a specific language. Furthermore, another advantage of embodiments herein may be to enable the KG creation accessible to resource poor languages. Embodiments herein may even enable the KG creation for languages that do not have any script. Embodiments herein may directly minimize the error between the predicted graph and the oracle graph during training and may enable to exploit the domain specific global and local linguistic structures present in the data to predict the accurate KG. During the supervised training phase, the NN parameters may be explicitly optimized to minimize the error between the predicted KG and the true KG. The direct optimization may help to learn the local, and global contexts and all relevant information present in the speech signal, which may be helpful in scenarios where speech may be more unstructured and noisier, e.g., conversational speech. Another advantage may be understood to be that embodiments herein may be robust to lack of entities and relations that may be extracted from conversational inputs modality, e.g., text / speech. A further advantage may be understood to be that embodiments herein may be agnostic to the grammatical mistakes of the user. An additional advantage may be understood to be that embodiments herein may be agnostic to the domain, making it easier to use, as the artificial neural network 121 may be trained to obtain a supervised graph. Yet another advantage may be understood to be that embodiments herein may automatically handle the issues with the speech on mis-spelled words, fillers, and disfluencies, whereas the cascaded approach performance is severely affected by these factors.. A further advantage may be understood to be that embodiments herein may enable to be extended to use in different domains that may relate to the establishment of explainable and robust semantic communication. Figure 12 depicts an example of the arrangement that the first node 111 may comprise to perform the method described in Figure 4 and / or Figures 6-11. The first node 111 may be understood to be for handling the first KG. The first node 111 is configured to operate in the computer system 100. Several embodiments are comprised herein. It should be noted that the examples herein are not mutually exclusive. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the first node 111, and will thus not be repeated here. For example, the data may be configured to encode recordings of conversations. In some embodiments, the first node 111 is configured to obtain the data configured to encode the conversational speech. In some embodiments, the first node 111 is configured to obtain the features configured to be extracted from the data, wherein the features are configured to be short-term features, refraining from converting the data into text. The first node 111 is configured to determine, using the artificial neural network 121, the first KG configured to represent words in the conversational speech and the relations of the words. The determining is configured to use, as first input, the features configured to be extracted from the data configured to be obtained. The first node 111 is also configured to initiate outputting the (second) indication of the first KG configured to be determined. In some embodiments, the artificial neural network 121 may be configured to have the end-to-end neural network architecture. The artificial neural network 121 may be configured to transform the features, configured to be obtained, extracted from the data, into the latent representation by using the encoder. In some embodiments, the encoder is configured to be generated by one of: the bidirectional Gated Recurrent Unit, the bidirectional Long Short-Term Memory Networks and the encoder blocks available in Transformer architecture. In some embodiments, the artificial neural network 121 is configured to comprise the decoder, and the decoder is configured to take as second input the latent representation of the data, map the latent representation to words, and output the first KG artificial neural network 121 by using the neural network configured to have the capability to model sequences of data. following. In some embodiments, the first node 111 may be further configured with at least one of the following four configurations. In some embodiments, the first node 111 may be further configured to obtain the plurality of sets of data. Each set of data may be configured to encode the respective conversational speech. In some embodiments, the first node 111 may be further configured to obtain, from the plurality of sets of data, the respective plurality of sets of features configured to be extracted from the respective set of data, refraining from converting the data into text. In some embodiments, the first node 111 may be obtain, for each set of data of the plurality of sets of data, the respective plurality of words and the respective second KG. The respective second KG may be configured to correspond to the respective oracle KG of the respective set of data, and the respective plurality of words may be configured to correspond to the respective oracle plurality of words. In some embodiments, the first node 111 may be further configured to train the artificial neural network 121 to determine the respective first KG for each respective set of data using the plurality of sets of data, the respective plurality of words and the respective second KG configured to be obtained. The training may be configured to be performed by minimizing the loss function configured to represent the difference between each respective first KG and the respective second KG configured to be obtained. The training may be configured to be performed iteratively until the loss function may be below the threshold. The determining of the first KG may be configured to be performed using the artificial neural network 121 configured to trained. In some embodiments, at least one of the following may apply: a) the respective second KG may be configured to be received from the third node 113 configured to operate in the computer system 100, and b) the first node 111 may be configured to refrain from training the artificial neural network 121, and the trained artificial neural network 121 may be configured to be received in the first indication from the second node 121 configured to operate in the computer system 100. In some embodiments, the conversational speech may be configured to be in the first language and the first KG may be configured to be in the second language. In some embodiments, the training of the artificial neural network 121 may be configured to comprise using as input the plurality of sets of data, comprising the sets of data in conversational speech of different respective first languages, and using for each set of data of the plurality of sets of data, the respective second KG in the different second language than the respective first language. In some embodiments, the (second) indication of the determined first KG may be configured to be outputted in one of: i) the healthcare application, and the conversational speech may be configured to describe the condition of the patient, ii) the tele-operated driving application, and the conversational speech may be configured to describe instructions to the vehicle by the remote driver, iii) the vehicle-to-everything application, and the conversational speech may be configured to describe vehicle-related problems between stakeholders, and iv) the field service operation application, and the conversational speech may be configured to describe the problem to be solved at the customer site. The embodiments herein in the first node 111 may be implemented through one or more processors, such as a processing circuitry 1201 in the first node 111 depicted in Figure 12, together with computer program code for performing the functions and actions of the embodiments herein. A processor, as used herein, may be understood to be a hardware component. The program code mentioned above may also be provided as a computer program product, for instance in the form of a data carrier carrying computer program code for performing the embodiments herein when being loaded into the first node 111. One such carrier may be in the form of a CD ROM disc. It is however feasible with other data carriers such as a memory stick. The computer program code may furthermore be provided as pure program code on a server and downloaded to the first node 111. The first node 111 may further comprise a memory 1202 comprising one or more memory units. The memory 1202 is arranged to be used to store obtained information, store data, configurations, schedulings, and applications etc. to perform the methods herein when being executed in the first node 111. In some embodiments, the first node 111 may receive information from, e.g., the second node 112, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100, through a receiving port 1203. In some embodiments, the receiving port 1203 may be, for example, connected to one or more antennas in first node 111. In other embodiments, the first node 111 may receive information from another structure in the computer system 100 through the receiving port 1203. Since the receiving port 1203 may be in communication with the processing circuitry 1201, the receiving port 1203 may then send the received information to the processing circuitry 1201. The receiving port 1203 may also be configured to receive other information. The processing circuitry 1201 in the first node 111 may be further configured to transmit or send information to e.g., the second node 112, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100, through a sending port 1204, which may be in communication with the processing circuitry 1201, and the memory 1202. Those skilled in the art will also appreciate that the units comprised within the first node 111 described above as being configured to perform different actions, may refer to a combination of analog and digital circuits, and / or one or more processors configured with software and / or firmware, e.g., stored in memory, that, when executed by the one or more processors such as the processing circuitry 1201, perform as described above. One or more of these processors, as well as the other digital hardware, may be included in a single Application-Specific Integrated Circuit (ASIC), or several processors and various digital hardware may be distributed among several separate components, whether individually packaged or assembled into a System-on-a-Chip (SoC). The first node 111 may be configured to perform any of the Actions described in relation to Figure 4 and / or Figures 6-11, e.g., by means of the processing circuitry 1201 within the first node 111, configured to perform any of such actions. Also, in some embodiments, different units comprised within the first node 111 may be configured to perform the different actions described above, implemented as one or more applications running on one or more processors such as the processing circuitry 1201. Thus, the methods according to the embodiments described herein for the first node 111 may be respectively implemented by means of a computer program 1205 product, comprising instructions, i.e., software code portions, which, when executed on at least one processing circuitry 1201, cause the at least one processing circuitry 1201 to carry out the actions described herein, as performed by the first node 111. The computer program 1205 product may be stored on a computer-readable storage medium 1206. The computer- readable storage medium 1206, having stored thereon the computer program 1205, may comprise instructions which, when executed on at least one processing circuitry 1201, cause the at least one processing circuitry 1201 to carry out the actions described herein, as performed by the first node 111. In some embodiments, the computer-readable storage medium 1206 may be a non-transitory computer-readable storage medium, such as a CD ROM disc, or a memory stick. In other embodiments, the computer program 1205 product may be stored on a carrier containing the computer program 1205 just described, wherein the carrier is one of an electronic signal, optical signal, radio signal, or the computer-readable storage medium 1206, as described above. The first node 111 may comprise a communication interface configured to facilitate, or an interface unit to facilitate, communications between the first node 111 and other nodes or devices, e.g. the third node 113, The interface may, for example, include a transceiver configured to transmit and receive radio signals over an air interface in accordance with a suitable standard. In other embodiments, the first node 111 may comprise a radio circuitry 1207, which may comprise e.g., the receiving port 1203 and the sending port 1204. The radio circuitry 1207 may be configured to set up and maintain at least a wireless connection with the second node 112, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100. Circuitry may be understood herein as a hardware component. Hence, embodiments herein also relate to the first node 111 operative to operate in the computer system 100. The first node 111 may comprise the processing circuitry 1201 and the memory 1202, said memory 1202 containing instructions executable by said processing circuitry 1201, whereby the first node 111 is further operative to perform the actions described herein in relation to the first node 111, e.g., in Figure 4 and / or Figures 6-11. Figure 13 depicts an example of the arrangement that the second node 112 may comprise to perform the method described in Figure 5 and / or Figures 7-10. The second node 112 may be understood to be for handling the respective first KG. The second node 112 is configured to operate in the computer system 100. Several embodiments are comprised herein. It should be noted that the examples herein are not mutually exclusive. One or more embodiments may be combined, where applicable. All possible combinations are not described to simplify the description. Components from one embodiment may be tacitly assumed to be present in another embodiment and it will be obvious to a person skilled in the art how those components may be used in the other exemplary embodiments. The detailed description of some of the following corresponds to the same references provided above, in relation to the actions described for the second node 112, and will thus not be repeated here. For example, the data may be configured to encode recordings of conversations. The second node 112 is configured to obtain the plurality of sets of data. Each set of data is configured to encode the respective conversational speech. The second node 112 is further configured to obtain, from the plurality of sets of data, the respective plurality of sets of features configured to be extracted from the respective set of data, refraining from converting the data into text. The second node 112 is further configured to, obtain, for each set of data of the plurality of sets of data, the respective plurality of words and the respective second KG. The respective second KG is configured to correspond to the respective oracle KG of the respective set of data, and the respective plurality of words is configured to correspond to the respective oracle plurality of words. The second node 112 is further configured to train the artificial neural network 121 to determine the respective first KG for each respective set of data using the plurality of sets of data, the respective plurality of words and the respective second KG configured to be obtained. The training is configured to be performed by minimizing the loss function configured to represent the difference between each respective first KG and the respective second KG configured to be obtained. The training is configured to be performed iteratively until the loss function is below the threshold. The second node 112 is further configured to initiate outputting the first indication of the artificial neural network 121 configured to be trained. In some embodiments, the artificial neural network 121 may be configured to have the end-to-end neural network architecture. The artificial neural network 121 may be configured to transform the respective plurality of sets of features, configured to be obtained, extracted from the respective set of data into the respective latent representation by using an encoder. In some embodiments, the encoder may be configured to be generated by one of: the bidirectional Gated Recurrent Unit, the bidirectional Long Short-Term Memory Networks and the encoder blocks available in Transformer architecture. In some embodiments, the artificial neural network 121 may be configured to comprise the decoder, and the decoder may be configured to take as second input the respective latent representations of the data, map the respective latent representations to respective words, and output the respective first KGs by using the neural network having the capability to model sequences of data In some embodiments, at least one of the following may apply: a) the respective second KG may be configured to be received from the third node 113 configured to operate in the computer system 100, and b) the initiating of the outputting of the first indication may be configured to comprise sending the first indication to the first node 111 configured to operate in the computer system 100. In some embodiments, the training of the artificial neural network 121 may be configured to comprise using as input the plurality of sets of data, configured to comprise the sets of data in conversational speech of different respective first languages, and using for each set of data of the plurality of sets of data, the respective second KG in the different second language than the respective first language. In some embodiments, the first indication may be configured to be outputted in one of: i) the healthcare application, and the conversational speech may be configured to describe the condition of the patient, ii) the tele-operated driving application, and the conversational speech may be configured to describe instructions to the vehicle by the remote driver, iii) the vehicle- to-everything application, and the conversational speech may be configured to describe vehicle-related problems between stakeholders, and iv) the field service operation application, and the conversational speech may be configured to describe the problem to be solved at the customer site. The embodiments herein in the second node 112 may be implemented through one or more processors, such as a processing circuitry 1301 in the second node 112 depicted in Figure 13, together with computer program code for performing the functions and actions of the embodiments herein. A processor, as used herein, may be understood to be a hardware component. The program code mentioned above may also be provided as a computer program product, for instance in the form of a data carrier carrying computer program code for performing the embodiments herein when being loaded into the second node 112. One such carrier may be in the form of a CD ROM disc. It is however feasible with other data carriers such as a memory stick. The computer program code may furthermore be provided as pure program code on a server and downloaded to the second node 112. The second node 112 may further comprise a memory 1302 comprising one or more memory units. The memory 1302 is arranged to be used to store obtained information, store data, configurations, schedulings, and applications etc. to perform the methods herein when being executed in the second node 112. In some embodiments, the second node 112 may receive information from, e.g., the first node 111, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100, through a receiving port 1303. In some embodiments, the receiving port 1303 may be, for example, connected to one or more antennas in second node 112. In other embodiments, the second node 112 may receive information from another structure in the wireless communications network 130 through the receiving port 1303. Since the receiving port 1303 may be in communication with the processing circuitry 1301, the receiving port 1303 may then send the received information to the processing circuitry 1301. The receiving port 1303 may also be configured to receive other information. The processing circuitry 1301 in the second node 112 may be further configured to transmit or send information to e.g., the first node 111, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100, through a sending port 1304, which may be in communication with the processing circuitry 1301, and the memory 1302. Those skilled in the art will also appreciate that the units comprised within the second node 112 described above as being configured to perform different actions, may refer to a combination of analog and digital circuits, and / or one or more processors configured with software and / or firmware, e.g., stored in memory, that, when executed by the one or more processors such as the processing circuitry 1301, perform as described above. One or more of these processors, as well as the other digital hardware, may be included in a single Application-Specific Integrated Circuit (ASIC), or several processors and various digital hardware may be distributed among several separate components, whether individually packaged or assembled into a System-on-a-Chip (SoC). The second node 112 may be configured to perform any of the Actions described in relation to Figure 5 and / or Figures 7-10, e.g., by means of the processing circuitry 1301 within the second node 112, configured to perform any of such actions. Also, in some embodiments, different units comprised within the second node 112 may be configured to perform the different actions described above, implemented as one or more applications running on one or more processors such as the processing circuitry 1301. Thus, the methods according to the embodiments described herein for the second node 112 may be respectively implemented by means of a computer program 1305 product, comprising instructions, i.e., software code portions, which, when executed on at least one processing circuitry 1301, cause the at least one processing circuitry 1301 to carry out the actions described herein, as performed by the second node 112. The computer program 1305 product may be stored on a computer-readable storage medium 1306. The computer- readable storage medium 1306, having stored thereon the computer program 1305, may comprise instructions which, when executed on at least one processing circuitry 1301, cause the at least one processing circuitry 1301 to carry out the actions described herein, as performed by the second node 112. In some embodiments, the computer-readable storage medium 1306 may be a non-transitory computer-readable storage medium, such as a CD ROM disc, or a memory stick. In other embodiments, the computer program 1305 product may be stored on a carrier containing the computer program 1305 just described, wherein the carrier is one of an electronic signal, optical signal, radio signal, or the computer-readable storage medium 1306, as described above. The second node 112 may comprise a communication interface configured to facilitate, or an interface unit to facilitate, communications between the second node 112 and other nodes or devices, e.g., the first node 111, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100. The interface may, for example, include a transceiver configured to transmit and receive radio signals over an air interface in accordance with a suitable standard. In other embodiments, the second node 112 may comprise a radio circuitry 1307, which may comprise e.g., the receiving port 1303 and the sending port 1304. The radio circuitry 1307 may be configured to set up and maintain at least a wireless connection with the first node 111, the third node 113, the radio network node 130, the device 140, and / or another structure in the computer system 100. Circuitry may be understood herein as a hardware component. Hence, embodiments herein also relate to the second node 112 operative to operate in the wireless communications network 130. The second node 112 may comprise the processing circuitry 1301 and the memory 1302, said memory 1302 containing instructions executable by said processing circuitry 1301, whereby the second node 112 is further operative to perform the actions described herein in relation to the second node 112, e.g., in Figure 5 and / or Figures 7-10. When using the word "comprise" or “comprising”, it shall be interpreted as non- limiting, i.e., meaning "consist at least of". The embodiments herein are not limited to the above-described preferred embodiments. Various alternatives, modifications and equivalents may be used. Therefore, the above embodiments should not be taken as limiting the scope of the invention. Generally, all terms used herein are to be interpreted according to their ordinary meaning in the relevant technical field, unless a different meaning is clearly given and / or is implied from the context in which it is used. All references to a / an / the element, apparatus, component, means, step, etc. are to be interpreted openly as referring to at least one instance of the element, apparatus, component, means, step, etc., unless explicitly stated otherwise. The steps of any methods disclosed herein do not have to be performed in the exact order disclosed, unless a step is explicitly described as following or preceding another step and / or where it is implicit that a step must follow or precede another step. Any feature of any of the embodiments disclosed herein may be applied to any other embodiment, wherever appropriate. Likewise, any advantage of any of the embodiments may apply to any other embodiments, and vice versa. Other objectives, features and advantages of the enclosed embodiments will be apparent from the following description. As used herein, the expression “at least one of:” followed by a list of alternatives separated by commas, and wherein the last alternative is preceded by the “and” term, may be understood to mean that only one of the list of alternatives may apply, more than one of the list of alternatives may apply or all of the list of alternatives may apply. This expression may be understood to be equivalent to the expression “at least one of:” followed by a list of alternatives separated by commas, and wherein the last alternative is preceded by the “or” term. Any of the terms processor and circuitry may be understood herein as a hardware component. As used herein, the expression “in some embodiments” has been used to indicate that the features of the embodiment described may be combined with any other embodiment or example disclosed herein. As used herein, the expression “in some examples” has been used to indicate that the features of the example described may be combined with any other embodiment or example disclosed herein. REFERENCES 1. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin, “Attention is all you needed” , arXiv:1706.03762. 2. Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli, ‘’ wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations”, arXiv:2006.11477. 3. Ronald J Williams and David Zipser.1989. A learning algorithm for continually running fully recurrent neural networks. Neural computation, 1(2):270– 280.

Claims

CLAIMS:

1. A computer-implemented method performed by a first node (111), the method being for handling a first KG, the first node (111) operating in a computer system (100), the method comprising: − obtaining (405) data encoding conversational speech, and − obtaining (406) features extracted from the data, wherein the features are short- term features, refraining from converting the data into text, − determining (407), using an artificial neural network (121), a first KG representing words in the conversational speech and relations of the words, wherein the determining (407) uses, as first input, the obtained features extracted from the data, and − initiating (408) outputting an indication of the determined first KG.

2. The method according to claim 1, wherein the artificial neural network (121) has an end-to-end neural network architecture, wherein the artificial neural network (121) transforms the obtained features extracted from the data into a latent representation by using an encoder.

3. The method according to claim 2, wherein the encoder is generated by one of: a bidirectional Gated Recurrent Unit, bidirectional Long Short-Term Memory Networks and encoder blocks available in Transformer architecture.

4. The method according to any of claims 2-3, wherein the artificial neural network (121) comprises a decoder, wherein the decoder takes as second input the latent representation of the data, maps the latent representation to words, and outputs the first KG artificial neural network (121) by using a neural network having a capability to model sequences of data.

5. The method according to claim any of claims 1-4, further comprising at least one of: − obtaining (401) a plurality of sets of data, wherein each set of data encodes a respective conversational speech, − obtaining (402), from the plurality of sets of data, a respective plurality of sets of features extracted from the respective set of data, refraining from converting the data into text, − obtaining (403), for each set of data of the plurality of sets of data, a respective plurality of words and a respective second KG, wherein the respective secondKG corresponds to a respective oracle KG of the respective set of data, and the respective plurality of words corresponds to a respective oracle plurality of words, − training (404) the artificial neural network (121) to determine a respective first KG for each respective set of data using the obtained plurality of sets of data, the respective plurality of words and the respective second KG, wherein the training (404) is performed by minimizing a loss function representing a difference between each respective first KG and the obtained respective second KG, wherein the training (404) is performed iteratively until the loss function is below a threshold, and wherein the determining (407) of the first KG is performed using the trained artificial neural network (121).

6. The method according to claim 5, wherein at least one of: − the respective second KG is received from a third node (113) operating in the computer system (100), and − the first node (111) refrains from training (404) the artificial neural network (121), and wherein the trained artificial neural network (121) is received in a first indication from a second node (112) operating in the computer system (100).

7. The method according to any of claims 1-6, wherein the conversational speech is in a first language and the first KG is in a second language.

8. The method according to claim 7 and any of claims 5-6, wherein the training (404) of the artificial neural network (121) comprises using as input the plurality of sets of data, comprising sets of data in conversational speech of different respective first languages, and using for each set of data of the plurality of sets of data, a respective second KG in a different second language than the respective first language.

9. The method according to any of claims 1-8, wherein the indication of the determined first KG is outputted in one of: i. a healthcare application, and wherein the conversational speech describes a condition of a patient, ii. a tele-operated driving application, and wherein the conversational speech describes instructions to a vehicle by a remote driver, iii. a vehicle-to-everything application, and wherein the conversational speech describes vehicle-related problems between stakeholders, andiv. a field service operation application, and wherein the conversational speech describes a problem to be solved at a customer site.

10. A computer-implemented method performed by a second node (112), the method being for handling a first KG, the second node (112) operating in a computer system (100), the method comprising: − obtaining (501) a plurality of sets of data, wherein each set of data encodes a respective conversational speech, − obtaining (502), from the plurality of sets of data, a respective plurality of sets of features extracted from the respective set of data, refraining from converting the data into text, − obtaining (503), for each set of data of the plurality of sets of data, a respective plurality of words and a respective second KG, wherein the respective second KG corresponds to a respective oracle KG of the respective set of data, and the respective plurality of words corresponds to a respective oracle plurality of words, − training (504) the artificial neural network (121) to determine a respective first KG for each respective set of data using the obtained plurality of sets of data, the respective plurality of words and the respective second KG, wherein the training (504) is performed by minimizing a loss function representing a difference between each respective first KG and the obtained respective second KG, wherein the training (504) is performed iteratively until the loss function is below a threshold, and − initiating (505) outputting a first indication of the trained artificial neural network (121).

11. The method according to claim 10, wherein the artificial neural network (121) has an end-to-end neural network architecture, wherein the artificial neural network (121) transforms the obtained respective plurality of sets of features extracted from the respective set of data into a respective latent representation by using an encoder.

12. The method according to claim 11, wherein the encoder is generated by one of: a bidirectional Gated Recurrent Unit, bidirectional Long Short-Term Memory Networks and encoder blocks available in Transformer architecture.

13. The method according to any of claims 11-12, wherein the artificial neural network (121) comprises a decoder, wherein the decoder takes as second input the respectivelatent representations of the data, maps the respective latent representations to respective words, and outputs the respective first KGs by using a neural network having a capability to model sequences of data.

14. The method according to any of claims 10-13, wherein the training (504) of the artificial neural network (121) comprises using as input the plurality of sets of data, comprising sets of data in conversational speech of different respective first languages, and using for each set of data of the plurality of sets of data, a respective second KG in a different second language than the respective first language.

15. A first node (111), for handling a first KG, the first node (111) being configured to operate in a computer system (100), the first node (111) being further configured to: − obtain data configured to encode conversational speech, and − obtain features configured to be extracted from the data, wherein the features are configured to be short-term features, refraining from converting the data into text, − determine, using an artificial neural network (121), a first KG configured to represent words in the conversational speech and relations of the words, wherein the determining is configured to use, as first input, the features configured to be extracted from the data configured to be obtained, and − initiate outputting an indication of the first KG configured to be determined.

16. The first node (111) according to claim 15, being further configured to at least one of: − obtain a plurality of sets of data, wherein each set of data is configured to encode a respective conversational speech, − obtain, from the plurality of sets of data, a respective plurality of sets of features configured to be extracted from the respective set of data, refraining from converting the data into text, − obtain, for each set of data of the plurality of sets of data, a respective plurality of words and a respective second KG, wherein the respective second KG is configured to correspond to a respective oracle KG of the respective set of data, and the respective plurality of words is configured to correspond to a respective oracle plurality of words, − train the artificial neural network (121) to determine a respective first KG for each respective set of data using the plurality of sets of data, the respective plurality of words and the respective second KG configured to be obtained, wherein the training is configured to be performed by minimizing a loss functionconfigured to represent a difference between each respective first KG and the respective second KG configured to be obtained, wherein the training is configured to be performed iteratively until the loss function is below a threshold, and wherein the determining of the first KG is configured to be performed using the artificial neural network (121) configured to be trained.

17. A second node (112), for handling a first KG, the second node (112) being configured to operate in a computer system (100), the second node (112) being further configured to: − obtain a plurality of sets of data, wherein each set of data is configured to encode a respective conversational speech, − obtain, from the plurality of sets of data, a respective plurality of sets of features configured to be extracted from the respective set of data, refraining from converting the data into text, − obtain, for each set of data of the plurality of sets of data, a respective plurality of words and a respective second KG, wherein the respective second KG is configured to correspond to a respective oracle KG of the respective set of data, and the respective plurality of words is configured to correspond to a respective oracle plurality of words, − train the artificial neural network (121) to determine a respective first KG for each respective set of data using the plurality of sets of data, the respective plurality of words and the respective second KG configured to be obtained, wherein the training is configured to be performed by minimizing a loss function configured to represent a difference between each respective first KG and the respective second KG configured to be obtained, wherein the training is configured to be performed iteratively until the loss function is below a threshold, and − initiate outputting a first indication of the artificial neural network (121) configured to be trained.

Images