Documentation system with AI-supported and rule-based structuring logic

Abstract

Offline-capable documentation system for the social pedagogical and social work sector, comprehensive a) a speech input unit that can be executed locally on a mobile or stationary device to convert spoken language into text without an Internet connection, b) a locally embedded AI module for automatic linguistic smoothing, neutralization and pseudonymization of the transcribed text, c) a rule-based assignment logic that automatically recognizes and assigns areas of life, measures, methods and objectives from the prepared text using predefined technical rules and keyword lists, d) a local storage unit for the encrypted storage of the processed and categorised documentation entries, (e) a user interface for entering the speech input, checking the processing results and outputting the final documentation.

Description

Technical area:

[0001] The invention relates to a documentation system in the field of information technology, in particular a system for digital documentation for the social education and social work sectors. The system enables the capture and processing of documentation entries using voice input, combining speech recognition, artificial intelligence, and rule-based components. It automatically assigns the captured information to subject categories and is offline-capable, allowing data processing without a constant internet connection. State of the art:

[0002] In the social pedagogical and social work environments, documentation is often still created manually in written form or using simple electronic text systems. Although voice assistance systems and speech-to-text technologies that enable text dictation are already known from other fields, these often rely on a stable online connection to convert speech into text. Specific solutions for social work that use voice input are still in development or at the pilot stage. For example, the first voice-controlled documentation approaches exist in youth welfare, but these typically do not provide for comprehensive offline processing. Furthermore, the existing systems lack functions for automatic linguistic smoothing of input or for the automated pseudonymization of personal data. The categorization orAssigning content to subject areas is usually done manually by the specialist or subsequently through separate evaluation steps. Overall, the current state of the art is characterized by the lack of an integrated solution that combines all relevant functions—offline speech recognition, AI-supported speech processing, and rule-based subject-specific assignment—in a single system. Object of the invention:

[0003] The invention is based on the object of significantly facilitating and improving the documentation process in social work and social pedagogy through technical means. In particular, the aim is to create a way to record documentation directly via voice input, without being dependent on an internet connection, and to simultaneously optimize the quality of the text recorded in this way, prepare it in compliance with data protection regulations, and classify it in a structured manner. The aim is to reduce the time required for documentation, automatically correct input errors or irregularities in expression, and automatically pseudonymize sensitive personal data during recording. Furthermore, the invention is to ensure that the content of the documentation is immediately assigned to the correct subject categories or form sections in order to facilitate later use of the data (e.g., for reports or statistics). Technical solution:

[0004] This problem is solved by an offline-capable documentation system that has the features mentioned in the protection claim. The system according to the invention comprises, in particular, a local speech input unit for converting spoken language into text, an AI-supported speech processing module for automatically improving and pseudonymizing the text, and a rule-based classification logic for categorizing the content. The combination of these components enables incoming spoken documentation to be seamlessly converted into cleaned, data protection-compliant, and categorized text documents. Furthermore, by eliminating the need for a constant internet connection, high data security and availability are ensured in every situation. Innovative components:

[0005] What is particularly innovative about the invention is the combination of the following components in a single system: • Offline Speech Capture: An offline-capable speech input component with direct on-device speech recognition that converts spoken speech into text without an internet connection. • Mini-AI for text processing: A compact, locally executed AI module (“Mini-AI”) that immediately smooths the captured text inputs linguistically (i.e., standardizes grammatically and stylistically), checks the content (e.g., for key terms or completeness) and automatically pseudonymizes personal information. • Rule-based subject classification logic: A rule-based subject classification logic based on industry-specific rules and criteria of social work. This module uses defined key terms or language patterns to identify the subject area or section of a given documentation structure to which the content belongs and classifies it accordingly.

[0006] The linking of these elements in a consistent workflow is previously unknown and represents the core of the invention. In particular, the integration of pseudonymization during text creation and the use of an adaptable rule engine for subject assignment are new and offer significant added value compared to the current state of the art. Application function in the social-educational and social work documentation process:

[0007] In practical use, the invention means that social work professionals, for example, can immediately dictate their report after a conversation with a client. The system records the spoken documentation via a microphone and converts it into text on-site. This text is automatically processed by the AI ​​module: linguistic inaccuracies or colloquial language are converted into more professional language, key points of content are highlighted or supplemented if necessary, and all personal names or identifying characteristics are replaced with neutral terms (pseudonymization). Immediately afterwards, the classification logic classifies the text and assigns the information to the appropriate categories of the professional system—such as "follow-up process," "measure planning," "observation," or similar, depending on the predefined schema of the documenting organization.The specialist thus receives a fully prepared and sorted report in real time, which they can immediately review, correct as needed, and transfer to the digital file. This significantly accelerates the documentation process and improves its quality, while consistently adhering to data protection requirements. Areas of application:

[0008] The system according to the invention can be used in various areas of social pedagogy and social work. These include, in particular, child and youth welfare (e.g., in child-rearing support, residential care, or juvenile court services), family support, school social work, counseling centers, addiction support, and other areas of social services. The system can be used wherever professionals regularly need to document facts—for example, in progress reports, case documentation, assistance plans, or report books. Related fields, such as clinical social work or psychosocial counseling, can also benefit from such technology. In principle, the system is suitable for all areas of application where mobile, offline-capable, and data protection-compliant documentation via voice input is useful. Technical features:

[0009] The invention is characterized by a number of technical features which, in combination, result in the described functionality: • The system comprises hardware in the form of a mobile or stationary electronic device with a microphone unit and a computing unit on which the software components are executed locally. • A software component for automatic speech recognition converts spoken language into text without the need for a connection to cloud services (offline speech recognition). • An AI-supported word processing system analyses the transcribed text in real time and improves the linguistic quality (grammar, syntax, technical terminology), marks or checks relevant content aspects and replaces identified personal data with pseudonyms. • A rule-based module with stored technical rules and keyword lists automatically categorizes the cleaned text and assigns it to the corresponding fields or categories of the intended documentation schema. • All processed data is stored in a local database or storage structure, whereby pseudonymization ensures that no real names of sensitive persons are stored. • A user interface allows the user to start recording, view and edit the generated text, and transfer or export the final documentation to other systems.

[0010] These technical features interact in the inventive combination to achieve the stated purpose. Depending on the embodiment, the described modules can be implemented as standalone software units or as an integrated software package. It is crucial that all processing steps take place locally and in a coherent sequence, thus providing a robust, data protection-compliant, and efficient documentation tool for practice.

Claims

[1] Offline-capable documentation system for the social pedagogical and social work sector, comprehensive a) a speech input unit that can be executed locally on a mobile or stationary device to convert spoken language into text without an Internet connection, b) a locally embedded AI module for automatic linguistic smoothing, neutralization and pseudonymization of the transcribed text, c) a rule-based assignment logic that automatically recognizes and assigns areas of life, measures, methods and objectives from the prepared text using predefined technical rules and keyword lists, d) a local storage unit for the encrypted storage of the processed and categorised documentation entries, (e) a user interface for entering the speech input, checking the processing results and outputting the final documentation. [2] System according to claim 1,characterized by that the AI ​​module is based on a mini-AI model (e.g. Phi-2, TinyLLaMA) and checks the inputs in real time for ethical and data protection criteria (e.g. exclusion of personal data, non-discriminatory language). [3] System according to claim 1 or 2, characterized by that the rule-based assignment logic uses positive and negative lists as well as context triggers to recognize semantic patterns and synonyms and to assign the entries to a technical documentation schema (e.g. LVR guidelines, BEI_NRW, ICF). [4] System according to one of claims 1-3, characterized by that it provides a formatting export of the generated documentation in standardized JSON or PDF structure. [5] System according to any one of claims 1-4, characterized by that it includes optional interfaces for calendar synchronization, Nextcloud or exports to specialist software (e.g. BeWoPlaner, Vivendi).