Monitoring system, method for operating the monitoring system and computer program

Abstract

The invention relates to a monitoring system 1 with a plurality of monitoring devices 2 for generating monitoring data, with an event database 4 for recording event data based on the monitoring data, and with an NL event interpretation module 3 for generating the event data from the monitoring data, wherein the event data is designed as natural language data, and / or with an NL query module 5 for generating queries, wherein the queries are designed as natural language queries, wherein the monitoring system 1 is designed to output monitoring events in response to the natural language queries.

Description

The invention relates to a monitoring system having the features of the preamble of claim 1. The invention also relates to a method for operating the monitoring system and to a computer programPrior ArtTraditional surveillance systems have only recorded surveillance video of surveillance areas and stored them. During recording, or quasi offline, the surveillance videos can be viewed and checked by surveillance personnel. However, such monitoring involves a high level of human complexity and the susceptibility to errors which are added by the human factor.Monitoring systems were thus developed which were able to detect events via digital image processing and report the corresponding events, so that the events could be controlled by monitoring personnel, for example. As a result, the activity of the monitoring personnel has been limited to the monitoring of the detected, highly relevant events, so that a reduction of monitoring personnel and monitoring in a constant, although automated, monitoring quality has been ensured.The document DE 10 2005 01915 3 A1 describes a method and system for processing data. The document discloses methods for processing data, comprising the steps of: recording raw data; processing, in particular digitizing, compressing and / or encoding, the raw data; analyzing, in particular performing a content analysis, the raw or the processed data; evaluating, in particular performing an event recognition, the analysis results in order to create a semantic description of the content; indexing the processed data by means of the semantic description; processing the indexed data. In the semantic description, an abstract, formal description and representation of the event that was detected with the monitoring system thus takes place.Disclosure of the InventionThe subject matter of the invention is a monitoring system having the features of claim 1, a method having the features of claim 14 and a computer program having the features of claim 15.The subject matter of the invention is thus a monitoring system which is suitable and / or designed for monitoring a monitoring region. As monitoring areas, in particular coherent and / or distributed monitoring areas in public or private environments can be considered.The monitoring system has a plurality of monitoring devices for generating monitoring data. The monitoring data can be sensor data of the monitoring devices, but it can also be provided that the monitoring data are already preprocessed by the monitoring device or an intermediate device and are designed as preprocessed monitoring data.The monitoring devices can be designed in particular as monitoring cameras, such as PTZ cameras or stationary cameras, 360° cameras. Alternatively or additionally, the monitoring devices can be designed as access control devices, intrusion sensor devices, motion sensor devices, environment detection sensor devices, etc. It is also possible for the monitoring devices to be implemented as fire, gas and / or temperature sensor devices, etc.The monitoring system has an event database for recording event data on the basis of the monitoring data. The event data are in particular designed as further-processed monitoring data. The event database may be located on a local data storage device or on a data storage device on a server, in the cloud, etc. The event database can be designed as a single device, but it is also possible for the event database to be distributed over different digital data processing devices and / or entities. An event database is to be understood in particular as a data collection independently of the architecture of the event database.Event data in the most general form is understood to mean data which can be associated with a monitoring event in the monitoring data. In particular, the event data comprise a description of the monitoring event and a connection to the associated monitoring event.In the context of the invention, it is proposed that the monitoring system has an NL event interpretation module for generating the event data from the monitoring data, wherein the event data are designed as natural language data.Alternatively or additionally, the monitoring system has an NL query module for generating queries, wherein the queries are designed as natural language queries. The monitoring system is configured to issue monitoring events in response to the natural language queries.It is thus a consideration of the invention to implement the data storage structure and / or the query structures on the basis of natural language data and / or natural language queries and thus to use natural language processing (NLP). In particular, natural language generation (NLG) data and / or natural language understanding (NLU) queries are used for natural language language data. Natural language is understood in particular to mean the natural language, so that the natural language data defines a description of the monitoring event on the basis of the natural language and / or the natural language queries on the basis of the natural language.Security systems are key components for maintaining security and monitoring in various areas. With a constantly increasing number of cameras and associated devices such as access control devices, intrusion sensors, motion sensors, intercom systems, perimeter detection systems, etc., monitoring and analyzing the vast amounts of data generated by these systems becomes a difficult task. Existing solutions typically require manual review and special knowledge to identify particular events or patterns, which is labor intensive and often ineffective. Moreover, these systems do not provide an intuitive way to search for events or patterns based on natural language descriptions or configure the system with natural language commands.From 2023, large language models (LLMs), such as GPT-4 developed by OpenAl, have greatly advanced the field of natural language processing (NLP) and artificial intelligence (AI). These models are capable of generating human-like text, understanding context, responding to prompts, and even generating density, layers, and code, among other capabilities. The innovation here is to apply advanced language models, such as GPT-4, developed by OpenAl to generate natural language descriptions of events. This means that the system can articulate the events in a manner that is comprehensible to anyone. Furthermore, users can search for events with simple voice queries and configure the system even with simple commands. This increases the accessibility and efficiency of the system, which is advantageous for all users. In essence, this invention utilizes modern AI technology to make security systems more intelligent, user friendly and effective.The monitoring system thus has the advantage that the natural language data can be directly understood by a user, since they describe the natural language monitoring events (natural language). Alternatively or additionally, the monitoring system has the advantage that the queries are designed as natural language queries and are initiated in this way via a user query that corresponds to the usual speech understanding of the user.The monitoring system can be installed without the monitoring devices on a digital data processing device, one or more servers and / or in the cloud. If necessary, the monitoring system without the monitoring devices has a data-technology interface to the monitoring devices.Further possible advantages, especially depending on the design, are: Efficient event evaluation: Conventional systems often require manual interpretation of events and video material, which can be labour intensive and time consuming. The proposed monitoring system automates this process and thus saves time and resources. Enhanced accessibility: By creating natural language data as natural language event descriptions, the monitoring system makes the monitoring data more readily accessible and comprehensible to non-specialized users. This represents a considerable improvement over conventional systems, which often require special knowledge for interpreting the data. Contextual understanding: The ability of the monitoring system to include contextual information in its event descriptions can provide valuable insights that may be overseeed in conventional systems. For example, it may detect whether an event occurs at an unusual time or location, which could indicate suspect activities. Extended Search Functions: The NL Query Module of the monitoring system allows users to search for events with natural language queries. This facilitates users to search for certain events as compared to conventional systems, which often require complex query languages or manual video verification. Natural language configuration: The ability of the monitoring system to accept natural language configuration commands makes it more user-friendly and easier to manage than conventional systems. This reduces the need for special training and can save time in setting up or adapting the system. Enhanced Security: By detecting and alerting unusual events or patterns, the monitoring system can increase security and help prevent incidents before they occur. This proactive approach is a significant improvement over conventional systems, which are often reactive and rely on manual monitoring. Scalability: The automation capabilities of the monitoring system make it highly scalable. It can manage an increasing number of cameras and related apparatuses as monitoring devices without correspondingly increasing the amount of work or resources. This makes it suitable for large installations which are difficult to manage with conventional systems. Integration of AI and ML: By integrating artificial intelligence and machine learning technologies, the monitoring system can continually improve its performance over time by learning from past data to better interpret future monitoring events. This is a further advantage over conventional systems which generally do not have such learning capabilities. Overall, the proposed monitoring system significantly improves the efficiency, accessibility and performance of safety management systems as monitoring systems and is thus a valuable instrument.In a preferred embodiment of the invention, the NL event interpretation module and / or the NL query module has an NL generator submodule. For the NL event interpretation module, the NL generator submodule is designed, in particular, to generate the natural language data from detected monitoring events. Examples of such generation are described in Karpathy, A.; Fei-Fei, L. Deep visual-semantic alignments for generating image descriptions. Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, 7-12 June 2015; pp. 3128-3137.For the NL query module, the NL generator submodule is designed, in particular, to generate the Natuals Language queries from a natural voice input, optionally acoustically or in text form. Examples of such generation is ConstructGPT.In particular, the natural language queries are matched to the natural language data in such a way that they have a common natural language basis. Especially in the description of queries or monitoring events in natural language, the descriptions can deviate greatly from one another with regard to the selection of the expression, the structure of the description, etc. It is advantageous if the NL generator submodules are matched to one another in such a way that they have the same or a similar language model for the description. In particular, these can be trained with the same training data.The NL event interpretation module preferably has an event definition submodule which, on the basis of the monitoring data, defines a monitoring event as a basis for the natural language data. In particular, the monitoring event is fed to the NL generator submodule of the NL event interpretation module in order to generate the natural language data. The event definition submodule may be based on digital image processing, as is known from video surveillance, for example. Alternatively, the event definition submodule can be designed as an artificial intelligence (AI), in particular as an artificial neural network, which receives the monitoring data as input data and provides the detected monitoring event at the output. In particular, the event definition submodule is based on deep / machine learning.In a preferred development of the invention, the NL event interpretation module has a configuration submodule and a correlator submodule. The configuration submodule provides system configuration data of the monitoring system and / or event configuration data for correlated monitoring events.System configuration data is understood to mean, in particular, data about the monitoring devices; in particular, the data include the position of the monitoring devices in the monitoring area, the type of monitoring device, etc.Event configuration data for correlated monitoring events are understood in particular to mean data which define special conditions for the correlated monitoring events. These may include special time conditions, such as the occurrence of the monitoring event in a particular time window and / or correlations of monitoring devices, such as the occurrence of the monitoring event or a related monitoring event at two or more monitoring devices. For example, there is a correlation of monitoring devices if an actuation of an access control device and at the same time the occurrence of an object / person in the monitoring area in the vicinity of the access control device is detected as monitoring data.The correlator submodule is designed to define correlated monitoring events as a basis for the natural language data on the basis of the monitoring events and also, in addition, the system configuration data and / or the event configuration data of the configuration submodule. The correlator submodule is thus designed to use further information in addition to the results from the event definition submodule. In this way, monitoring events which are correlated with one another can be merged into a correlated monitoring event as a monitoring event. By being able to access system configuration data, the local proximity of monitoring devices can be used, for example, to correlate monitoring events of two or more adjacent monitoring devices with one another.It is particularly preferred that the NL query module is designed to send the natural language queries to the event database in order to obtain monitoring events, in particular correlated monitoring events from the correlator submodule or simple monitoring events from the event definition submodule as a response. This creates a database for storing monitoring events and querying monitoring events on the basis of natural language.It is particularly preferred that the NL query module is designed to generate natural language queries on the basis of natural language queries. The natural language requests are input by a user audibly or textually. In this way, a user can communicate with the monitoring system particularly easily in order to obtain monitoring results as responses.In an alternative or a development of the invention, the NL query module is designed to generate natural language queries on the basis of natural language queries and to propose them as possible natural language queries that can be selected by a user. Thus, the NL query module offers a computer-aided formulation aid for formulating the natural language queries. In this way, the user can communicate with the monitoring system particularly easily.In a further alternative or development of the invention, the NL query module has a natural language configuration instruction submodule, wherein the natural language configuration instruction submodule is designed to generate a monitoring query for a predefinable monitoring event as a natural language query. In this way, it is possible for the user or in another way to set up continuous monitoring for a specific monitoring event, comprising in particular a simple monitoring event and / or a correlated monitoring event. The natural language configuration instruction submodule configures the monitoring system such that the response to the monitoring query is output as a natural language query when / or as soon as the predefinable monitoring event occurs. For example, the natural language configuration instruction submodule communicates with the NL event interpretation module. In this way, the monitoring system may be sensitized by the user or otherwise to particular monitoring events.In this case, it is possible, on the one hand, for the natural language configuration instruction submodule to act on the configuration submodule in order to sensitize the latter and in particular to define the monitoring results or the correlated monitoring results there. Alternatively or additionally, the natural language configuration instruction submodule can act on the input of the event database, such that the response is output as soon as the monitoring event occurs. It is also possible that the natural language configuration instruction submodule queries the event database in a kind of continuous loop, so that the response is issued as soon as the monitoring event is entered in the event database.In a preferred development of the invention, the monitoring system has an output device for outputting the detected monitoring event. In particular, for example, an alarm can be initiated, a monitoring center can be activated, etc.The invention further relates to a method for operating the monitoring system as described above or according to one of the preceding claims. In the method, it is provided that the NL event interpretation module generates event data from the monitoring data, wherein the event data are designed as natural language data.Alternatively or additionally, it is provided that the NL query module generates queries, wherein the queries are designed as natural language queries, wherein the monitoring system outputs monitoring events in response to the natural language queries.A further subject matter of the invention relates to a computer program which executes the method as soon as it is formed on the monitoring system and / or on a digital data processing device. The computer program can be installed on a digital data processing device, a server and / or on the cloud, in particular in an executable manner. An optional subject matter of the invention relates to a digital storage medium for storing the computer program, in particular a machine-readable storage medium, preferably a non-transitory machine-readable storage medium, on which the described computer program is stored.This invention thus aims to revolutionize the manner in which security systems are managed by introducing an integrated system that uses natural language processing (NLP) and machine learning (ML) to generate descriptions of the natural language events recorded by the system. This system allows users to search for events with natural language queries and configure the system with natural language commands, which significantly improves accessibility and efficiency of the monitoring system.Further features, advantages and effect of the invention will become apparent from the following description of a preferred exemplary embodiment and from the attached figure. This shows: FIG. 1 is a schematic block diagram of a monitoring system as an embodiment of the invention.FIG. 1 shows a schematic block diagram of a monitoring system 1 as an exemplary embodiment of the invention. The monitoring system has the function of monitoring a monitoring region with respect to a monitoring event. The monitoring event may be any, such as entering a safe area, a content event, in particular a human cluster, an unsupervised object, such as a luggage case in a station, a fire, a fire, a gas concentration, etc.The monitoring system 1 has a plurality of monitoring devices 2 which generate monitoring data during operation. The monitoring devices 2 can be selected as desired from a group of: monitoring cameras, in particular stationary or mobile monitoring cameras, PTZ cameras, 360° cameras, fire alarm devices, temperature sensor devices, access control devices, etc. The monitoring data are designed as raw monitoring data or as preprocessed monitoring data.The monitoring system 1 has an NL event interpretation module 3, wherein the NL event interpretation module 3 is designed to generate event data of monitoring events on the basis of the monitoring data, wherein the event data are designed as natural language data. The natural language data describes the natural language monitoring events. For example, the natural language data may describe the monitoring events as follows:an object enters the surveillance scene;a fire occurs in the surveillance scene;an access control device was activated;a fire alarm device has registered a fire.In addition, for example, a description of the location and / or the time of the monitoring event, in particular likewise as natural language data, can be added.The monitoring system 1 has an event database 4 which stores the event data which are generated on the basis of the monitoring data. The event database 4 can be present centrally or decentrally in any desired architecture. In particular, monitoring data from the monitoring devices 2 are assigned to the event data, so that the event data has the natural language data and a connection, such as a pointer or a reference, to the generated monitoring data.The monitoring system 1 has an NL query module which can generate queries, wherein the queries are designed as natural language queries. Thus, the queries are made in natural language. The monitoring system 1 is designed to output suitable monitoring events in response to the natural language queries.The monitoring system 1 thus allows the monitoring events to be stored as event data in the form of natural language data in the event database 4 and asked in natural language via the NL interrogation module 5. The advantage of the monitoring system 1 is above all that monitoring events are no longer stored and queried in a formal, semantic computer language, but that the event data and the query are implemented on a natural language basis. As a result, the monitoring system 1 allows access and control of the event data in a user-friendly and intuitively understood natural language sphere or coding.The NL event interpretation module 3 can be designed as an AI, which evaluates the monitoring data on the basis of deep and / or machine learning and generates the event data as natural language data.The NL event interpretation module 3 is in particular the first start point for raw data which originate from different safety devices as monitoring devices 2 in the monitoring system 1 as safety system. It is used to interpret and categorize various monitoring events that occur in the safety system. If, for example, an access system device is activated as monitoring device 2 at the main entrance, this module interprets the event as "activation of the access system at the main entrance". This module analyzes the events in real time and, in particular, correlates finally related events in correlator submodule 8 in order to obtain a more comprehensive overview of the situation. For example, if a motion detection event occurs followed by activation of the access system, the system could correlate these events and interpret them as a single event: "Motion detected near the main entrance and simultaneously activate the access system.".By way of example, FIG. 1 shows an optional substructure in the NL event interpretation module 3:The NL event interpretation module 3 comprises an event definition submodule 6. The event definition submodule is designed to define, in particular to recognize and describe in any desired manner, the monitoring events on the basis of the monitoring data. The definition of monitoring events may be implemented via known digital image processing algorithms. Alternatively, any artificial neural network, in particular based on deep / machine learning, can be implemented for the detection, description and thus definition of the monitoring events.Optionally, the NL event interpretation module has a storage device 7, wherein the storage device 7 is designed as a FIFO memory, which represents a buffer in the information and / or data flow. The input of the storage device 7 is connected in terms of data technology to the event definition submodule 6.The in NL event interpretation module 3 comprises a correlator submodule 8, wherein the correlator submodule 8 is connected to the output of the storage device 7, so that the monitoring events from the event definition submodule 6 are provided to the correlator submodule 8.Optionally, the NL event interpretation module 3 can have a configuration submodule 9, wherein the configuration submodule 9 holds system configuration data of the monitoring system 1, in particular of the monitoring devices 2. In particular, the system configuration data can comprise a type of the respective monitoring device 2 (monitoring camera, sensor device, etc.). Alternatively or additionally, the system configuration data comprise a spatial arrangement and / or a model of the monitoring area with the monitoring devices 2, so that the local, relative relationship between the monitoring devices 2 and the monitoring area is known.It can also be provided that the configuration submodule 9 has event configuration data for correlated monitoring events. The event configuration data can comprise special conditions for evaluating and / or evaluating and / or linking the monitoring events.The special conditions can define time ranges for monitoring events, for example: a relevant monitoring event is present, for example, if the monitoring event falls within a time range in which, for example, no persons should be present in the monitoring area.The event configuration data can also relate to special conditions with respect to the simultaneous or related occurrence of monitoring events: for example, a special condition can relate to the activation of an access control device as monitoring device 2 as a monitoring event and the detection of an object in a monitoring region of a monitoring camera specified thereto as a further monitoring device 2 as a monitoring event.The correlator submodule 8 is designed to detect and / or define correlated monitoring events on the basis of the monitoring events and the system configuration data and / or event configuration data. A correlated monitoring event thus comprises in particular individual monitoring events from different monitoring devices 2, the occurrence of a monitoring event in a specific time range or other more complex monitoring events in which the special conditions are fulfilled.The correlator submodule 8 can operate according to a user-assisted and / or manually set rule-box, wherein the input data, comprising individual monitoring events, system configuration data and / or event configuration data, are evaluated according to the rule-box. Alternatively, the correlator submodule 8 can be designed as an artificial neural network, in particular on the basis of deep / machine learning, which merges the input data and defines and outputs correlated monitoring events.The monitoring events, in particular the individual monitoring events and / or the correlated monitoring events, are transferred to an NL generator submodule 10.The NL generator submodule 10 translates the monitoring events into natural language data, that is to say a description of the respective monitoring event in natural language. The NL generator submodule 10 can be designed in particular as an artificial neural network.The NL generator submodule 10 is a natural language generator: this component converts the interpreted events into detailed, natural language descriptions. It utilizes advanced natural language generation (NLG) techniques to generate human-like, easily comprehensible descriptions of the events. For example, it could describe a series of events as follows: "The main entry access system was activated at 10:15 o'clock. Simultaneously, movement has been detected near the main entrance, possibly indicating that someone has entered the building. This component would add a relevant context to the generated descriptions, making them more informative and useful. For example, if the access system was activated at an unusual time, the natural language generator could insert this context into the description, e.g., "The main entry access system was activated 3:00 a.m., an unusual time for this location.". The events are written in the event database 4 as a natural language database. The natural language generator is adaptable to the needs of the particular system installation. This component would adapt the generated descriptions to the preferences of the system configurator. For example, some customers prefer sparse descriptions, while others prefer more detailed enumerations. A personalization component could learn these preferences over time and adjust the descriptions accordingly. In addition, the system provides a feedback mechanism for the module. This mechanism would allow system configurators to provide feedback to the generated descriptions which natural language generator 10 could use to improve its performance over time.Optionally, the NL event interpretation module 3 has an NL video generator submodule 11, which is designed to output monitoring scenes in the monitoring data of the monitoring cameras as monitoring devices 2 as natural language data, that is to say to describe them in natural language. The NL video generator submodule 11 can be designed in particular as an artificial neural network. The natural language data is held in the event database 4.The NL video generator submodule 11 is a video description generator: this module is responsible for analyzing the recorded video material and creating natural language descriptions of scenes. It uses available computer vision techniques to identify objects, people, and activities in video material. For example, a description could be made such as "A big person in a blue jacket was at 10:15 am at the main entrance and carried a big box.". The events are stored in the event database 4 as a natural language database.The monitoring system 1 has an NL query module 5, an NL query module for generating queries, wherein the queries are designed as natural language queries.The NL query module 5 is a search module: this component of the system is supported by a highly-modal language model available on site or in the cloud. It processes complex requests in natural language and offers users an intuitive and powerful tool for investigating the event data. It may process queries such as "Show me events in which the principal entrance was entered at unusual times" or "Find instances of movements detected near the principal entrance while the access system was activated", for example. Existing language models are capable of producing accurate and contextual results when provided with comprehensive historical data in the form of natural language descriptions of events. To improve the functionality of this module, a predictive search function is incorporated. This function proposes search requests during user input based on the most frequent or recent search requests, thereby improving user friendliness and accelerating the search process. This component understands the context of the user's search query and uses it to provide more relevant search results. For example, if the user searches for "unusual access events", the search component understands that the user is likely to be interested in access events that occurred at unusual times or locations.The NL query module 5 can thus act on the monitoring system 1, in particular on the NL event interpretation module 3, in different ways:Optionally, the NL query module 5 has a further NL generator submodule 12 which is designed to translate natural language queries as natural language queries from a user 13 into queries in the form of natural language queries. This allows the natural language user 3 to make a natural language request which is converted by the NL generator submodule 12 into a natural language request which is subsequently made to the monitoring system 1, in particular to the event database 4. The advantage of the further NL generator submodule 12 is that the natural language queries are unified on a natural language language basis.In response, the NL query module 5 receives monitoring events, in particular comprising simple / single and / or correlated monitoring events. In particular, the responses are output as natural language responses and provided to an output device 14. The output device 14 can be designed as a monitoring center and provide the natural language responses to the user 13.In one possible variation, it can be provided that the user 13 makes a natural language request and the NL query module 5, in particular the NL generator submodule 12, makes available a plurality of natural language queries as selection options, which are subsequently selected by the user 13, for example via a human-machine interface. Subsequently, the selected natural language query is transferred to the monitoring system 1, as was described above.Optionally, the NL query module 5 has a natural language configuration instruction submodule 15, which is designed to generate a monitoring request, in particular a natural language request, from the user 13 of a predefinable monitoring event as a monitoring natural language query. The natural language configuration instruction submodule 15 is designed to configure the monitoring system 1 such that the response to the monitoring query is output as a natural language response when or as soon as the predefinable monitoring event occurs. With the natural language configuration instruction submodule 15, in particular, permanent queries or specific queries can be generated which are answered as soon as the contents of the query enter.The monitoring natural language query can be sent to the event database 4 so that the monitoring event occurs in response to the monitoring natural language query as soon as the requested monitoring event is entered in the event database 4.Alternatively, the monitoring natural language query may be sent to the NL event interpretation module 3, wherein a monitoring result is obtained in response to the monitoring natural language query as soon as the NL event interpretation module 3 detects the queried monitoring event.It is also possible for the configuration submodule 9 to be designed to define a predefinable monitoring event as a correlated monitoring event on the basis of the monitoring natural language query. In this case, the correlated monitoring event is detected by the correlator submodule 8 and can be output via the NL generator submodule 10 in response to the monitoring natural language query as soon as it occurs.From the data point of view, the response can be fed back directly from the NL event interpretation module 3, in particular the correlator submodule 8 and / or the NL generator submodule 10 or the event database 4, to the NL query module 3 and passed to the output device 14. It is possible that the response is provided as a natural language response at the output device 14 and transmitted to the user 13.Alternatively or additionally, the natural language configuration instruction submodule 15 can be designed to configure the monitoring system 1 with natural language commands, i.e. with commands in natural language, from the user 13 in order to increase operability. The commands can relate to the integration of new monitoring devices 2 or their parameters. In particular, the configuration takes place in the configuration submodule 9.The natural language configuration instruction submodule 15 is in particular a natural language configuration module: this component has been developed to provide a user-friendly interface for the configuration of the system. It allows users to use natural language commands, making the system more accessible and intuitive. For example, the users may give commands such as "generate an alarm at night when a person attempts to enter a certain area" or "release an alarm when someone shows his identification after the regular working times on an access system". A learning mechanism may be incorporated in this module. This mechanism would learn from the user's configuration commands and improve the ability of the system to understand and execute these commands over time. For example, if the user frequently configures certain types of alerts, the natural language configuration module could learn to suggest these alerts when the user begins configuring a new alert. A feedback mechanism is integrated into the module. This mechanism would allow users to provide feedback on the interpretation of their commands by the system, which the natural language configuration instruction submodule 15 could use to improve their performance over time.More specifically, a user command is input to the natural language configuration instruction submodule 15 for natural language, which then interprets and translates the command into a system configuration. The module also includes a contextual understanding component that provides contextual configurations based on the user command, as well as a learning mechanism that improves the module's ability to interpret commands over time.References included in the specificationThis list of documents cited by the applicant has been produced in an automated manner and is only included for the better information of the reader. The list is not part of the German patent application or utility model application. The DPMA does not take any adhesion for any faults or omissions.Patent Literature citedDE 10 2005 01915 3 A1

[0004] Cited Non-Patent LiteratureKarpathy, A.; Fei-Fei, L. Deep visual-semantic alignments for generating image descriptors. Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, 7-12 June 2015; pp. 3128-3137

[0019]

Claims

Monitoring system (1) having at least one, preferably a plurality of monitoring devices (2) for generating monitoring data, having an event database (4) for recording event data on the basis of the monitoring data, characterized an NL event interpretation module (3) for generating the event data from the monitoring data, wherein the event data are designed as natural language data, and / or an NL query module (5) for generating queries, wherein the queries are designed as natural language queries, wherein the monitoring system (1) is designed to output monitoring events in response to the natural language queries.Monitoring system (1) according to Claim 1, characterized in that the NL event interpretation module (3) and / or the NL interrogation module (5) has an NL generator submodule (10, 12).Monitoring system (1) according to Claim 1 or 2, characterized in that the monitoring system (1) is designed to output the monitoring results as natural language responses.Monitoring system (1) according to one of the preceding claims, characterized in that the NL event interpretation module (3) has an event definition submodule (6), wherein the event definition submodule (6) is designed to define a monitoring event as a basis for the natural language data on the basis of the monitoring data.Monitoring system (1) according to Claim 4, characterized in that the NL event interpretation module (3) has a configuration submodule (9) and a correlator submodule (8), wherein the configuration submodule (9) has system configuration data of the monitoring system (1) and / or event configuration data for correlated monitoring events, wherein the correlator submodule (8) is designed to define correlated monitoring events on the basis of the monitoring events and of the system configuration data and / or of the event configuration data of the configuration submodule (8) as a basis for the natural language dataMonitoring system (1) according to one of the preceding claims, characterized in that the NL query module (5) is designed to send the natural language queries to the event database (4) in order to obtain monitoring events in response to the natural language queries.Monitoring system (1) according to one of the preceding claims, characterized in that the NL query module (5) is designed to generate natural language queries on the basis of natural language queries from a user (13).Monitoring system (1) according to one of the preceding claims, characterized in that the NL query module (5) is designed to generate natural language queries on the basis of natural language queries and to propose them as possible natural language queries which can be selected by one or the user (13).Monitoring system (1) according to one of the preceding claims, characterized in that the NL query module (5) has a natural language configuration instruction submodule (15), wherein the natural language configuration instruction submodule (15) is designed to generate a monitoring query for a predefinable monitoring event as a monitoring natural language query, wherein the natural language configuration instruction submodule (15) is designed to configure the monitoring system (1) such that the response to the monitoring query is output as a natural language response if / as soon as the predefinable monitoring event occurs.Monitoring system (1) according to claim 9, characterized in that the natural language configuration instruction submodule (15) is configured to send the monitoring natural language query to the event database (4) in order to obtain monitoring events in response to the monitoring natural language query.Monitoring system (1) according to claim 9 or 10, characterized in that the natural language configuration instruction submodule (15) is configured to send the monitoring natural language query to the NL event interpretation module (3) in order to obtain monitoring events in response to the monitoring natural language query.Monitoring system (1) according to Claim 10 or 11, characterized in that the configuration submodule (9) is designed to define a predefinable monitoring event as a correlated monitoring event on the basis of the monitoring natural language query.Monitoring system (1) according to one of the preceding claims, characterized byan output device (14) for outputting the monitoring events.Method for operating the monitoring system (1), in particular a monitoring system (1) according to one of the preceding claims, wherein at least one monitoring device (2) generates monitoring data, wherein an event database (4) records event data on the basis of the monitoring data, characterized in that an NL event interpretation module (3) generates event data from the monitoring data, wherein the event data are designed as natural language data, and / or an NL query module (5) generates queries, wherein the queries are designed as natural language queries, wherein monitoring events are output in response to the natural language queries.Computer program for carrying out the method according to claim 14, when the program is executed on the monitoring system (1) or on a digital data processing device.

Images