{"id":3338,"date":"2023-05-24T10:38:43","date_gmt":"2023-05-24T08:38:43","guid":{"rendered":"https:\/\/www.risc-software.at\/fachbeitrag-natural-language-processing-3\/"},"modified":"2023-12-29T12:58:42","modified_gmt":"2023-12-29T11:58:42","slug":"technical-article-natural-language-processing-3","status":"publish","type":"publication","link":"https:\/\/www.risc-software.at\/en\/technicalarticles\/technical-article-natural-language-processing-3\/","title":{"rendered":"With Natural Language Understanding (NLU) from text chaos to knowledge gain"},"content":{"rendered":"\n<h2 class=\"wp-block-heading is-style-v2-telegrafico\">How Natural Language Understanding also helps your company to optimise existing processes<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">by Sandra Wartner, MSc<\/h3>\n\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<p>In many companies, there is an increasing shift towards digitalisation and automation. In the process, enormous amounts of unstructured data are continuously accumulating, the scope and complexity of which deter the stakeholders concerned from evaluating it, or the potential in the existing data is often not even recognised in the first place. Regardless of whether fault messages in production processes are to be analysed, doctor&#8217;s letters are to be filed in a structured manner or products are to be suggested automatically, Natural Language Understanding (NLU) offers a broad spectrum of industry-specific and cross-industry applications.<br><\/p>\n\n\n\n<div style=\"height:100px\" aria-hidden=\"true\" class=\"wp-block-spacer\"><\/div>\n\n\n\n<div class=\"wp-block-media-text has-media-on-the-right is-stacked-on-mobile is-vertically-aligned-center\"><div class=\"wp-block-media-text__content\">\n<p><strong>Table of contest<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>In the beginning there is the mountain of data\u2026 and now what?<\/li>\n\n\n\n<li>How do I teach the AI system what to do?<\/li>\n\n\n\n<li>Current trends and challenges<\/li>\n\n\n\n<li>Conclusion<\/li>\n\n\n\n<li>Sources<\/li>\n\n\n\n<li>Author<\/li>\n<\/ul>\n<\/div><figure class=\"wp-block-media-text__media\"><img decoding=\"async\" width=\"1024\" height=\"683\" src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/AdobeStock_518004804-1024x683.jpg\" alt=\"NLP\" class=\"wp-image-1856 size-full\" srcset=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/AdobeStock_518004804-1024x683.jpg 1024w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/AdobeStock_518004804-300x200.jpg 300w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/AdobeStock_518004804-768x512.jpg 768w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/AdobeStock_518004804-1536x1024.jpg 1536w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/AdobeStock_518004804.jpg 1920w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure><\/div>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:66.66%\">\n<p>Language is omnipresent and we encounter it in many different facets in our everyday lives as well as in our professional environment &#8211; written by humans, spoken and communicated in different languages, but also analysed, processed and synthesised by machines. With <strong>Natural Language Processing (NLP)<\/strong>, computers are able to process and generate natural language automatically and act as an interface between humans and machines (for more details on NLP see [1]). As an application area of artificial intelligence (AI), NLP is always used when monotonous processes or frequently recurring tasks in text processing are to be automated, subsequently optimised and integrated into a higher-level framework. In this way, errors can be minimised in various areas, processes can be (partially) automated and savings can be achieved (through reduced personnel costs).<\/p>\n\n\n\n<p>RISC Software GmbH supports its customers with its many years of practical experience when it comes to the development of individually tailored, AI-supported solutions, including in the area of <strong>Natural Language Understanding (NLU)<\/strong>, a sub-area of Natural Language Processing.<\/p>\n\n\n\n<p><strong>Natural Language Understanding (NLU)<\/strong> focuses on the extraction of information from written text and thus on the acquisition of text comprehension with regard to a certain aspect. Syntax (grammatical structure) and semantics (meaning of words) play an important role. Examples of this are:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Information extraction<\/strong>, e.g. the recognition of persons, places or other keywords in texts (e.g. Named Entity Recognition (NER)),\n<ul class=\"wp-block-list\">\n<li><strong>Use case &#8220;Newsadoo&#8221;<\/strong>: &#8220;Newsadoo &#8211; All the news about your interests&#8221; &#8211; allows users to access news articles from numerous sources and offers relevant news personalised according to interests. In the background, NLP is used to transform unstructured text data into structured, analysable content.<\/li>\n\n\n\n<li><strong><a href=\"https:\/\/www.risc-software.at\/referenzprojekte\/flowgoess2t\/\" target=\"_blank\" rel=\"noreferrer noopener\">Use case &#8220;FLOWgoesS2T&#8221;<\/a><\/strong>: Voice messages on current traffic events are converted into written texts, in which important information such as roads, locations, driving directions and events are then automatically recognised and stored in a structured manner using NLP. This serves to support the editors in the processing of transmitted voice messages in order to be able to quickly identify traffic-relevant events.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>Classification of text into predefined categories<\/strong>\n<ul class=\"wp-block-list\">\n<li><strong><a href=\"https:\/\/www.risc-software.at\/ergebnis-projektabschluss-act4whistleblowing\/\" target=\"_blank\" rel=\"noreferrer noopener\">Use Case &#8220;ACT4&#8221;<\/a><\/strong>: In an expansion stage of the existing platform solution ACT4 of Compliance 2b GmbH, RISC Software GmbH is developing a trustworthy AI component together with the company, which on the one hand supports whistleblowers in submitting the report and on the other hand should enable the responsible officers to process the reports more efficiently and less error-prone. The system will automatically derive information (e.g. category or roles of the persons involved) from the textual information and compare it with already structurally recorded data in the form of a plausibility check.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>Sentiment and opinion analysis (sentiment analysis)<\/strong>\n<ul class=\"wp-block-list\">\n<li><strong>Use case &#8220;Intelligent Twitter Analysis&#8221;<\/strong>: Are positive emotions in tweets about listed companies related to their share price development? Sentiment analysis can be used to analyse a text in terms of sentiment (positive, negative, etc.) and evaluate how much information is actually contained between the lines.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:33.33%\">\n<figure class=\"wp-block-image size-full is-style-rounded\"><img decoding=\"async\" width=\"720\" height=\"405\" sizes=\"(max-width: 720px) 100vw, 720px\" src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2021_08_26_risc_animatedBanner_AnnaLyze_NLPoverview_v1.gif\" alt=\"NLP Explained\" class=\"wp-image-1868\"\/><\/figure>\n\n\n\n<figure class=\"wp-block-image size-large is-style-rounded\"><img decoding=\"async\" width=\"1024\" height=\"683\" sizes=\"(max-width: 1024px) 100vw, 1024px\" src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock-602578049-1024x683.jpg\" alt=\"Research\" class=\"wp-image-3021\" srcset=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock-602578049-1024x683.jpg 1024w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock-602578049-300x200.jpg 300w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock-602578049-768x512.jpg 768w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock-602578049-1536x1024.jpg 1536w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock-602578049.jpg 1920w\" \/><\/figure>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<h3 class=\"wp-block-heading\">In the beginning there is the mountain of data\u2026 and now what?<\/h3>\n\n\n\n<p>The first steps are almost always the hardest. The following (certainly not exhaustive) checklist provides an overview of the most relevant questions that every project team should clarify before the concrete planning or implementation of NLU or AI systems in general.<\/p>\n\n\n\n<div style=\"height:100px\" aria-hidden=\"true\" class=\"wp-block-spacer\"><\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/circle-question-regular.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">Is the problem formulated sufficiently well?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>What requirements must the AI system meet in order to be used beneficially in operations?<\/li>\n\n\n\n<li>Are the expected results clearly defined?<\/li>\n\n\n\n<li>Do all stakeholders have the same expectations?<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/magnifying-glass-solid.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">Is the type of problem to be solved known or clearly delimited (e.g. classification of words or documents, sentiment analysis)?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>If not, can the problem be solved in several sub-problems that can be clearly delineated?<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/database-solid.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">Can I solve the problem using the existing database?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>If not, are there ways to get this data, e.g. by using data from other\/public sources, or by collecting your own data?<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/check-double-solid.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">Is the data quality sufficiently &#8220;good&#8221;?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Data quality results from the interaction of different criteria that depend on the use case (see [2]).<\/li>\n\n\n\n<li>If the data quality is not sufficient &#8211; what measures can be taken to improve it? Is it possible to establish a robust data strategy in the company in the long term?<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/laptop-code-solid.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">Is a ground truth (correctly annotated examples) available?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>If not, can this be created? Are resources available or is technical\/domain-specific know-how available to annotate it?<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/ruler-solid.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">How do I assess whether a solution works &#8220;well enough&#8221;? How can I &#8220;measure&#8221; errors?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>On the one hand, metrics are needed for the accuracy of the models themselves, and on the other hand, evaluation strategies are needed to determine whether and what added value is generated by the use of the solution, e.g. a certain percentage increase in one or more of the company&#8217;s KPIs.<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/diagram-project-solid.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\"><br>Are there already solutions to similar problems or does the project have a high degree of innovation? How risk-tolerant is my organisation?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>If the degree of innovation is high and there are many risk factors, funding opportunities can also be used to be able to implement the project nevertheless, but with less risk (see [3]).<\/li>\n\n\n\n<li>If the risk factors are (still) unknown or unclear, a feasibility study can help to assess them (see [4]).<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div data-aos=\"fade-up\"  data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\" class=\"icon-box is-style-bg-blue\">\n  <div class=\"icon-overlay\">\n          <picture>\n        \n        \n        \n        \n        <img decoding=\"async\"  class=\"\" width=\"44\" height=\"44\"\n             src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/robot-solid-1.png\"\n             alt=\"\">\n      <\/picture>\n      <\/div>\n  \n\n<h5 class=\"has-text-align-center wp-block-heading\">How can I create a trustworthy AI system?<\/h5>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Which areas are relevant for my use case, e.g. comprehensibility, fairness, technical robustness (see [5])?<\/li>\n\n\n\n<li>Can I use methods from the field of Explainable AI to check my black box (see [6])?<\/li>\n<\/ul>\n\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<h3 class=\"wp-block-heading\">How do I teach the AI system what to do?<\/h3>\n\n\n\n<p>To move from raw data to a successfully implemented NLU component, a number of steps are necessary. The concrete measures vary from one project to the next, but the basic procedure follows the scheme shown in Figure 1.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img decoding=\"async\" width=\"923\" height=\"63\" sizes=\"(max-width: 923px) 100vw, 923px\" src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-Fachbeitrag_RISC_Software.png\" alt=\"NLU\" class=\"wp-image-3312\" srcset=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-Fachbeitrag_RISC_Software.png 923w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-Fachbeitrag_RISC_Software-300x20.png 300w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-Fachbeitrag_RISC_Software-768x52.png 768w\" \/><\/figure>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div class=\"accordion\">\n  <div class=\"accordion-header p-1.5 md:px-3 md:py-2 flex items-center justify-between \" id=\"headingblock_940990e2171cead34c553b735cbfe0a7\">\n    <h3 class=\" inline-block \">\n        Data basis    <\/h3>\n    <span class=\"accordion-icon-toggle inline-block\"><\/span>\n  <\/div>\n  <div id=\"collapseblock_940990e2171cead34c553b735cbfe0a7\" class=\"collapse\" aria-labelledby=\"headingblock_940990e2171cead34c553b735cbfe0a7\">\n    <div class=\"accordion-body p-1.5 md:p-3 \">\n      <p>The existing raw data can be in many different formats, e.g. as text fields in databases, contents of web pages, text files or text in images or scans. If texts are contained in (complex-)structured PDFs or web pages, relevant content can be extracted with some effort. For scans of documents, the Optical Character Recognition (OCR) method is used, which recognises texts in a two-dimensional image and stores them with their position for further processing. OCR systems already achieve very good results for images with structured, typewritten texts (e.g. scans or photos of analogue documents), but this step is often a challenge for photos (e.g. of street signs) or handwritten texts. Audio files can also be transcribed into written text using Speech-To-Text technologies. Depending on the quality of the recording, language and dialect, this can also involve considerable effort until the texts are available in sufficiently good quality for further processing.<\/p>\n    <\/div>\n  <\/div>\n<\/div>\n\n\n\n<div class=\"accordion\">\n  <div class=\"accordion-header p-1.5 md:px-3 md:py-2 flex items-center justify-between \" id=\"headingblock_78e428d499dfcd5949c31b0b869005c2\">\n    <h3 class=\" inline-block \">\n        Data preparation    <\/h3>\n    <span class=\"accordion-icon-toggle inline-block\"><\/span>\n  <\/div>\n  <div id=\"collapseblock_78e428d499dfcd5949c31b0b869005c2\" class=\"collapse\" aria-labelledby=\"headingblock_78e428d499dfcd5949c31b0b869005c2\">\n    <div class=\"accordion-body p-1.5 md:p-3 \">\n      <p>Next, the texts must be prepared for further processing. Depending on the application, this step requires, for example, removing certain punctuation marks and\/or excess spaces or converting texts to lower case. Although some information is lost as a result, this makes both manual and machine processing of the texts by AI models much easier. Another essential step is to tokenise the texts. Since computers cannot \u201ccalculate\u201d with words, a unique number is assigned to each word and all texts are converted into this uniform number scheme.<\/p>\n    <\/div>\n  <\/div>\n<\/div>\n\n\n\n<div class=\"accordion\">\n  <div class=\"accordion-header p-1.5 md:px-3 md:py-2 flex items-center justify-between \" id=\"headingblock_3feaf4ee98c8c94e836964cb72dafb74\">\n    <h3 class=\" inline-block \">\n        Language models    <\/h3>\n    <span class=\"accordion-icon-toggle inline-block\"><\/span>\n  <\/div>\n  <div id=\"collapseblock_3feaf4ee98c8c94e836964cb72dafb74\" class=\"collapse\" aria-labelledby=\"headingblock_3feaf4ee98c8c94e836964cb72dafb74\">\n    <div class=\"accordion-body p-1.5 md:p-3 \">\n      <p>Modern, deep-learning-based language models are pretrained in a self-supervised way on extensive text databases such as BookCorpus. A very common approach is so-called masked language modelling, where random parts of sentences (e.g. words) are blacked out and the model tries to refill the gap text as close as possible to the original text. For the model to build up a good understanding of natural language structures, millions of examples and many iterations of this guessing game are necessary. Since this process is very resource-intensive (high computing power and costs), these are usually pre-trained by large organisations such as Google or Facebook and \u2013 thankfully \u2013 made publicly available to other developers.<\/p>\n    <\/div>\n  <\/div>\n<\/div>\n\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div class=\"accordion\">\n  <div class=\"accordion-header p-1.5 md:px-3 md:py-2 flex items-center justify-between \" id=\"headingblock_d27b6f2ba00e8c4cdc1a12c2606b9914\">\n    <h3 class=\" inline-block \">\n        Finetuning    <\/h3>\n    <span class=\"accordion-icon-toggle inline-block\"><\/span>\n  <\/div>\n  <div id=\"collapseblock_d27b6f2ba00e8c4cdc1a12c2606b9914\" class=\"collapse\" aria-labelledby=\"headingblock_d27b6f2ba00e8c4cdc1a12c2606b9914\">\n    <div class=\"accordion-body p-1.5 md:p-3 \">\n      <p>Using the principle of so-called transfer learning, pre-trained models can now use their language understanding to learn the solution of concrete tasks (such as NER, text classification or sentiment analysis) with smaller amounts of data. Depending on the complexity of the task, hundreds to thousands of sample data are necessary for this fine-tuning.<\/p>\n    <\/div>\n  <\/div>\n<\/div>\n\n\n\n<div class=\"accordion\">\n  <div class=\"accordion-header p-1.5 md:px-3 md:py-2 flex items-center justify-between \" id=\"headingblock_2d7d7e8d21dbd8ef8483fb273652ca90\">\n    <h3 class=\" inline-block \">\n        Evaluation    <\/h3>\n    <span class=\"accordion-icon-toggle inline-block\"><\/span>\n  <\/div>\n  <div id=\"collapseblock_2d7d7e8d21dbd8ef8483fb273652ca90\" class=\"collapse\" aria-labelledby=\"headingblock_2d7d7e8d21dbd8ef8483fb273652ca90\">\n    <div class=\"accordion-body p-1.5 md:p-3 \">\n      <p>The quality of these models is then quantitatively evaluated using test or validation data provided. Depending on the task and goal, different metrics are used. It may therefore be necessary to evaluate and compare models on the basis of several metrics.<\/p>\n    <\/div>\n  <\/div>\n<\/div>\n\n\n\n<div class=\"accordion\">\n  <div class=\"accordion-header p-1.5 md:px-3 md:py-2 flex items-center justify-between \" id=\"headingblock_c4c6c347a8b2e84c704863432783531c\">\n    <h3 class=\" inline-block \">\n        Productive use    <\/h3>\n    <span class=\"accordion-icon-toggle inline-block\"><\/span>\n  <\/div>\n  <div id=\"collapseblock_c4c6c347a8b2e84c704863432783531c\" class=\"collapse\" aria-labelledby=\"headingblock_c4c6c347a8b2e84c704863432783531c\">\n    <div class=\"accordion-body p-1.5 md:p-3 \">\n      <p>The predictions of the models on new data (inference) provide results according to the structure from the sample data and can thus be integrated into the company workflow.<\/p>\n    <\/div>\n  <\/div>\n<\/div>\n\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:66.66%\">\n<h3 class=\"wp-block-heading\">Current trends and challenges: When AI&#8217;s learn to write, draw and communicate like humans do<\/h3>\n\n\n\n<p>In recent years, almost everything in the NLU field has revolved around the so-called Transformer models. These are a special architecture of artificial neural networks that is particularly suitable for dealing with text data (see also [7]). Google&#8217;s Language Model for Dialogue Applications &#8211; LaMDA for short &#8211; has attracted particular attention in recent months (see [8]). This model is trained to behave as humanly as possible in dialogue, and the model has already been able to prove this ability in several &#8220;interviews&#8221; (see [9]). The DALL-E models developed by OpenAI (see [10]) can also (among other things) generate images that match an input text. The model is based on the GPT-3 architecture (see [11]), which has previously been able to convince with its ability to generate new texts in previously unattained quality. A simplified model based on DALL-E is publicly available at <a href=\"https:\/\/www.craiyon.com\/\" target=\"_blank\" rel=\"noreferrer noopener\">craiyon.com<\/a>: What is a fun gimmick for average internet users can also find numerous productive applications.<\/p>\n\n\n\n<p>The biggest challenge in using these new models in innovative research projects is the data available for the task at hand. For successful fine-tuning of a pre-trained model to a new task, appropriate data is needed to show the model what to do. This data must also be available in sufficient quantity and meet the specified data quality criteria. Furthermore, the selection of the pre-trained model is also a challenge. In order to achieve the best results, a literature review and the testing and evaluation of different models is essential.<\/p>\n\n\n\n<p>Keeping track of all these exciting new innovations is not always easy. However, not only the latest trends should always be taken into account here either. Some tasks can also be solved with older methods or (in combination) with sophisticated rule-based systems, some of which are more efficient to use and also enable the traceability of model decisions on an ad-hoc basis. It is therefore definitely worthwhile to test out methods that have already been established in the long term for a first prototype.<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:33.33%\">\n<figure class=\"wp-block-image size-large is-style-rounded\"><img decoding=\"async\" width=\"1024\" height=\"683\" sizes=\"(max-width: 1024px) 100vw, 1024px\" src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-1024x683.jpg\" alt=\"NLU\" class=\"wp-image-3314\" srcset=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-1024x683.jpg 1024w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-300x200.jpg 300w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-768x512.jpg 768w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-1536x1024.jpg 1536w, https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-2048x1365.jpg 2048w\" \/><\/figure>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<h3 class=\"wp-block-heading\">Conclusion<\/h3>\n\n\n\n<p>Human language is amazingly complex and versatile. NLU solutions understand and interpret linguistically conveyed content better and better, and the rapid progress is becoming more and more impressive. Almost daily, the number of publicly available models is increasing, and at the same time it is becoming apparent how diversely they can already be used. With increasing digitalisation as well as the amount of routine processes, there is still a lot of untapped potential in the unstructured text data of companies to take their processes and products to the next level with NLP solutions. If you are also interested in using such technologies in your company, we would be happy to support you in planning and implementing NLP projects (<a href=\"https:\/\/www.risc-software.at\/annalyze-nlp\/\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/www.risc-software.at\/annalyze-nlp\/<\/a>).<\/p>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<h3 class=\"wp-block-heading\">Sources<\/h3>\n\n\n\n<p>[1] Wartner, Sandra (2021):&nbsp;<em>\u201eOK Google: What is Natural Language Processing?&#8221; \u2013 How machines reas, decode and undertsnad human language&nbsp;<\/em>(<a href=\"http:\/\/ris.w4.at\/en\/technical-article-natural-language-processing-1\/\" target=\"_blank\" rel=\"noreferrer noopener\">ris.w4.at\/en\/technical-article-natural-language-processing-1\/<\/a>)<\/p>\n\n\n\n<p>[2] Wartner, Sandra (2021):<em>&nbsp;Data quality: From information flow to information content &#8211; why clean data (quality) management pays off <\/em>(<a href=\"http:\/\/ris.w4.at\/en\/technical-article-data-quality\/\" target=\"_blank\" rel=\"noreferrer noopener\">ris.w4.at\/en\/technical-article-data-quality\/<\/a>)<\/p>\n\n\n\n<p>[3] Hochleitner, Christina (2021):&nbsp;<em>F\u00f6rderungen mit laufender Einreichm\u00f6glichkeit&nbsp;<\/em>(<a href=\"https:\/\/www.risc-software.at\/foerderungen-mit-laufender-einreichmoeglichkeit\/\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/www.risc-software.at\/foerderungen-mit-laufender-einreichmoeglichkeit\/<\/a>)<\/p>\n\n\n\n<p>[4] Wartner, Sandra (2021):&nbsp;<em>Why even an good idea needs a feasibility study <\/em>(<a href=\"http:\/\/ris.w4.at\/en\/technical-article-why-even-a-good-idea-needs-a-feasibility-study\/\" target=\"_blank\" rel=\"noreferrer noopener\">ris.w4.at\/en\/technical-article-why-even-a-good-idea-needs-a-feasibility-study\/<\/a>)<\/p>\n\n\n\n<p>[5] Wartner, Sandra (2021):&nbsp;<em>Trust in Artificial Intelligence &#8211; how we create and use trustworthy AI systems<\/em> (<a href=\"http:\/\/ris.w4.at\/en\/technical-article-trust-in-artificial-intelligence\/\" target=\"_blank\" rel=\"noreferrer noopener\">ris.w4.at\/en\/technical-article-trust-in-artificial-intelligence\/<\/a>)<\/p>\n\n\n\n<p>[6] Jaeger, Anna-Sophie (2022):&nbsp;<em>Explainable Artificial Intelligence (XAI)&nbsp;\u2013 How machine learning predictions become interpretable<\/em> (<a href=\"http:\/\/ris.w4.at\/en\/technical-article-explainable-artificial-intelligence\/\" target=\"_blank\" rel=\"noreferrer noopener\">ris.w4.at\/en\/technical-article-explainable-artificial-intelligence\/<\/a>)<\/p>\n\n\n\n<p>[7] Wartner, Sandra (2022):&nbsp;<em>Transformer models conquer Natural Language Processing<\/em> (<a href=\"http:\/\/ris.w4.at\/en\/technical-article-transformer-models-conquer-natural-language-processing\/\" target=\"_blank\" rel=\"noreferrer noopener\">ris.w4.at\/en\/technical-article-transformer-models-conquer-natural-language-processing\/<\/a>)<\/p>\n\n\n\n<p>[8]&nbsp;Thoppilan, Romal, et al. \u201cLamda: Language models for dialog applications.\u201d&nbsp;<em>arXiv preprint arXiv:2201.08239<\/em>&nbsp;(2022).<\/p>\n\n\n\n<p>[9] Lemoine, Blake (2022): \u201cIs LaMDA Sentient?&nbsp;\u2013&nbsp;an Interview\u201d (<a href=\"https:\/\/cajundiscordian.medium.com\/is-lamda-sentient-an-interview-ea64d916d917\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/cajundiscordian.medium.com\/is-lamda-sentient-an-interview-ea64d916d917<\/a>)<\/p>\n\n\n\n<p>[10] OpenAI (2022):&nbsp;<a href=\"https:\/\/openai.com\/dall-e-2\/\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/openai.com\/dall-e-2\/<\/a><\/p>\n\n\n\n<p>[11] Brown, Tom, et al. \u201cLanguage models are few-shot learners.\u201d&nbsp;<em>Advances in neural information processing systems<\/em>&nbsp;33 (2020): 1877-1901.<\/p>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignfull is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-28f84493 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:66.66%\">\n<h3 class=\"wp-block-heading\">Contact<\/h3>\n\n\n\n<div class=\"wp-block-contact-form-7-contact-form-selector\">\n<div class=\"wpcf7 no-js\" id=\"wpcf7-f663-o1\" lang=\"en-US\" dir=\"ltr\" data-wpcf7-id=\"663\">\n<div class=\"screen-reader-response\"><p role=\"status\" aria-live=\"polite\" aria-atomic=\"true\"><\/p> <ul><\/ul><\/div>\n<form action=\"\/en\/wp-json\/wp\/v2\/publication\/3338#wpcf7-f663-o1\" method=\"post\" class=\"wpcf7-form init\" aria-label=\"Contact form\" novalidate=\"novalidate\" data-status=\"init\">\n<fieldset class=\"hidden-fields-container\"><input type=\"hidden\" name=\"_wpcf7\" value=\"663\" \/><input type=\"hidden\" name=\"_wpcf7_version\" value=\"6.1.5\" \/><input type=\"hidden\" name=\"_wpcf7_locale\" value=\"en_US\" \/><input type=\"hidden\" name=\"_wpcf7_unit_tag\" value=\"wpcf7-f663-o1\" \/><input type=\"hidden\" name=\"_wpcf7_container_post\" value=\"0\" \/><input type=\"hidden\" name=\"_wpcf7_posted_data_hash\" value=\"\" \/>\n<\/fieldset>\n<div class=\"form-row\">\n\t<div class=\"form-input\">\n\t\t<p><label class=\"sr-only\" for=\"your-name\">Your name <\/label><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"your-name\"><input size=\"40\" maxlength=\"400\" class=\"wpcf7-form-control wpcf7-text wpcf7-validates-as-required\" id=\"your-name\" aria-required=\"true\" aria-invalid=\"false\" placeholder=\"Name\" value=\"\" type=\"text\" name=\"your-name\" \/><\/span>\n\t\t<\/p>\n\t<\/div>\n\t<div class=\"form-input\">\n\t\t<p><label class=\"sr-only\" for=\"your-email\">Your email<\/label><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"your-email\"><input size=\"40\" maxlength=\"400\" class=\"wpcf7-form-control wpcf7-email wpcf7-validates-as-required wpcf7-text wpcf7-validates-as-email\" id=\"your-email\" aria-required=\"true\" aria-invalid=\"false\" placeholder=\"E-Mail\" value=\"\" type=\"email\" name=\"your-email\" \/><\/span>\n\t\t<\/p>\n\t<\/div>\n<\/div>\n<div class=\"form-row\">\n\t<div class=\"form-input\">\n\t\t<p><label class=\"sr-only\" for=\"your-company\">Company <\/label><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"your-company\"><input size=\"40\" maxlength=\"400\" class=\"wpcf7-form-control wpcf7-text\" id=\"your-company\" aria-invalid=\"false\" placeholder=\"Unternehmen\" value=\"\" type=\"text\" name=\"your-company\" \/><\/span>\n\t\t<\/p>\n\t<\/div>\n\t<div class=\"form-input\">\n\t\t<p><label class=\"sr-only\" for=\"your-position\">Position<\/label><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"your-position\"><input size=\"40\" maxlength=\"400\" class=\"wpcf7-form-control wpcf7-text\" aria-invalid=\"false\" placeholder=\"Position\" value=\"\" type=\"text\" name=\"your-position\" \/><\/span>\n\t\t<\/p>\n\t<\/div>\n<\/div>\n<div class=\"form-row\">\n\t<div class=\"form-input\">\n\t\t<p><label class=\"sr-only\" for=\"your-subject\"> Subject <\/label><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"your-subject\"><input size=\"40\" maxlength=\"400\" class=\"wpcf7-form-control wpcf7-text wpcf7-validates-as-required\" id=\"your-subject\" aria-required=\"true\" aria-invalid=\"false\" placeholder=\"Thema\" value=\"\" type=\"text\" name=\"your-subject\" \/><\/span>\n\t\t<\/p>\n\t<\/div>\n<\/div>\n<p><span id=\"wpcf7-69d5e8e56d1cc-wrapper\" class=\"wpcf7-form-control-wrap phone-95-wrap\" style=\"display:none !important; visibility:hidden !important;\"><label for=\"wpcf7-69d5e8e56d1cc-field\" class=\"hp-message\">Please leave this field empty.<\/label><input id=\"wpcf7-69d5e8e56d1cc-field\"  class=\"wpcf7-form-control wpcf7-text\" type=\"text\" name=\"phone-95\" value=\"\" size=\"40\" tabindex=\"-1\" autocomplete=\"new-password\" \/><\/span><br \/>\n<label class=\"sr-only\" for=\"your-message\"> Your message (optional)<\/label><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"your-message\"><textarea cols=\"40\" rows=\"10\" maxlength=\"2000\" class=\"wpcf7-form-control wpcf7-textarea\" id=\"your-message\" aria-invalid=\"false\" placeholder=\"Ihre Nachricht an uns\" name=\"your-message\"><\/textarea><\/span><br \/>\n<span class=\"wpcf7-form-control-wrap\" data-name=\"hcap-cf7\">\t\t<input\n\t\t\t\ttype=\"hidden\"\n\t\t\t\tclass=\"hcaptcha-widget-id\"\n\t\t\t\tname=\"hcaptcha-widget-id\"\n\t\t\t\tvalue=\"eyJzb3VyY2UiOlsiY29udGFjdC1mb3JtLTdcL3dwLWNvbnRhY3QtZm9ybS03LnBocCJdLCJmb3JtX2lkIjo2NjN9-5cf29316f0fc31f5a29d11a228757560\">\n\t\t\t\t<span id=\"hcap_cf7-69d5e8e56d77a4.85756627\" class=\"wpcf7-form-control h-captcha \"\n\t\t\tdata-sitekey=\"3a6a81c1-2b2e-4b2a-b1eb-d9446bc09afb\"\n\t\t\tdata-theme=\"light\"\n\t\t\tdata-size=\"normal\"\n\t\t\tdata-auto=\"false\"\n\t\t\tdata-ajax=\"false\"\n\t\t\tdata-force=\"false\">\n\t\t<\/span>\n\t\t<input type=\"hidden\" id=\"_wpnonce\" name=\"_wpnonce\" value=\"62ded476a0\" \/><input type=\"hidden\" name=\"_wp_http_referer\" value=\"\/en\/wp-json\/wp\/v2\/publication\/3338\" \/><\/span><input class=\"wpcf7-form-control wpcf7-submit has-spinner btn\" type=\"submit\" value=\"Senden\" \/>\n<\/p><div class=\"wpcf7-response-output\" aria-hidden=\"true\"><\/div>\n<\/form>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:33.33%\">\n<h3 class=\"wp-block-heading\">Author<\/h3>\n\n\n<div class=\"contact-person\">\n      <picture>\n      \n      \n      \n      \n      <img decoding=\"async\" data-aos=\"fade-zoom-in\"\n           data-aos-offset=\"0\" class=\"w-full\" width=\"212\" height=\"293\"\n           src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/swartner1-removebg-preview.png\"\n           alt=\"\">\n    <\/picture>\n    \n\n<h5 class=\"wp-block-heading\">Sandra Wartner, MSc<\/h5>\n\n\n\n<p>Data Scientist<\/p>\n\n  <\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n<div class=\"wp-block-group-container alignfull \">\n<div class=\"wp-block-group alignwide is-layout-constrained wp-block-group-is-layout-constrained\"><div class=\"posts-slider-block\" data-aos=\"fade-up\" data-aos-offset=\"0\" data-aos-anchor-placement=\"top-bottom\">\n        <section class=\"splide posts-slider\" aria-label=\"Gallery Slides\">\n            <div class=\"splide__arrows\">\n                <button class=\"splide__arrow splide__arrow--prev\">\n                    <span class=\"sr-only\">Previous<\/span>\n                    <img decoding=\"async\" loading=\"lazy\" width=\"25\" height=\"21\" src=\"https:\/\/www.risc-software.at\/app\/themes\/risc-theme\/public\/images\/icon-arrow.35d2ec.svg\"\n                         alt=\"Previous\">\n                <\/button>\n                <button class=\"splide__arrow splide__arrow--next\">\n                    <span class=\"sr-only\">Next<\/span>\n                    <img decoding=\"async\" loading=\"lazy\" width=\"25\" height=\"21\" src=\"https:\/\/www.risc-software.at\/app\/themes\/risc-theme\/public\/images\/icon-arrow.35d2ec.svg\"\n                         alt=\"Next\">\n                <\/button>\n            <\/div>\n            <div class=\"inner\">\n                <div class=\"splide__track\">\n                    <div class=\"splide__list\">\n\n                                                    <a href=\"https:\/\/www.risc-software.at\/en\/technicalarticles\/technical-article-natural-language-processing-1\/\" class=\"splide__slide blog-post-teaser mb-1 lg:mb-3\">\n                                <div class=\"blog-image\">\n                                                                                                                                <picture>\n                                                                                        <img decoding=\"async\" src=\"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/shutterstock_1348453082-360x214.jpg\"\n                                                 alt=\"\u201cOK Google: What is Natural Language Processing?\u201d\">\n                                        <\/picture>\n                                                                    <\/div>\n                                <div class=\"blog-content px-2 py-3 xl:px-4 xl:py-5\">\n                                    <h3>\u201cOK Google: What is Natural Language Processing?\u201d<\/h3>\n                                    <div class=\"blog-post-excerpt mt-2\">\n                                        Natural Language Processing makes it possible to read, decode and understand human language by machine. Speech assistants, spelling correctors, email spam filters &#8211; NLP as a technology is omnipresent and already hides behind many processes and software applications deeply embedded in our everyday lives.\n                                    <\/div>\n                                    <span class=\"inline-block mt-2 more\">mehr erfahren <span class=\"ml-1 icon-more\"><\/span><\/span>\n\n                                <\/div>\n                            <\/a>\n                                            <\/div>\n                <\/div>\n            <\/div>\n        <\/section>\n    <\/div>\n<\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Regardless of whether fault messages in manufacturing processes are to be analyzed, doctor&#8217;s letters are to be filed in a structured manner, or products are to be suggested automatically, Natural Language Understanding (NLU) offers a broad spectrum of industry-specific and cross-industry applications for gaining knowledge from unstructured text data.<\/p>\n","protected":false},"featured_media":3315,"template":"","publication-category":[50],"class_list":["post-3338","publication","type-publication","status-publish","has-post-thumbnail","hentry","publication-category-data-science-and-a-i"],"acf":[],"portrait_thumb_url":"https:\/\/www.risc-software.at\/app\/uploads\/2023\/06\/2022-11-30-NLU-360x214.jpg","_links":{"self":[{"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/publication\/3338","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/publication"}],"about":[{"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/types\/publication"}],"version-history":[{"count":20,"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/publication\/3338\/revisions"}],"predecessor-version":[{"id":5045,"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/publication\/3338\/revisions\/5045"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/media\/3315"}],"wp:attachment":[{"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/media?parent=3338"}],"wp:term":[{"taxonomy":"publication-category","embeddable":true,"href":"https:\/\/www.risc-software.at\/en\/wp-json\/wp\/v2\/publication-category?post=3338"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}