<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@687948cede1548f5ac7829d1ba5bc1a5" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.1 Using vocabularies for bias detection</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@2a2d15e4d9c84418881430f3881dcada"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@2a2d15e4d9c84418881430f3881dcada" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>In the previous chapters you have learned about the importance of co-creating shared vocabularies to raise awareness and put cultural heritage professionals as well as users into the position to identify, contextualise, and - potentially - replace harmful and offensive language. You have been introduced to some examples of such vocabularies, including the DE-BIAS vocabulary, to the various shades and categories of bias, and to tools such as the Wheel of Bias, which can help you and your organisation in approaching the question of debiasing your collections and collection descriptions.</p> <p>But what are the options for applying all of this to the often huge collections held by cultural institutions and the vast amount of descriptions and other metadata used to make those collections findable and accessible?</p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@b146ec6116e64aa5b107fb7122b1822c" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.2 The example of the DE-BIAS tool: how the tool makes use of the vocabulary</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@fcf7646f455f4da0aec29971a1f500c8"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@fcf7646f455f4da0aec29971a1f500c8" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>Making use of a series of Natural Language Processing methods, the DE-BIAS tool&rsquo;s main source for identifying bias is the DE-BIAS vocabulary. It functions as a kind of matrix against which the tool examines the metadata to be analysed.</p> <p>The first step in this is the simple identification as to whether a term defined in the vocabulary can be found in the metadata. For this, a method called <strong>Tokenization</strong> is applied, which splits the plain text into words, phrases or other meaningful elements. Those are then checked against the terms in the DE-BIAS vocabulary.</p> <p>This might include different variations of a word, which already are included in the vocabulary, e.g. singular or plural forms, female or male forms, or words used as nouns and as adjectives and/or verbs. This is supported by a method called <strong>Lemmatization</strong>, which identifies the canonical or dictionary form of a word based on its intended meaning by taking into account their part of speech.</p> <p>Once the vocabulary terms have been identified in the analysed texts, two additional methods are available within the DE-BIAS tool to confirm that the specific context, in which a term is used, indeed renders its use as contentious. This binds the workflows of the tool back to the subtleties of bias, where a term might be considered harmful in one context, but not in another.</p> <p>One of these additional methods is <strong>Named Entity Recognition</strong>. With this, the tool aims to identify whether a contentious term might actually be part of a named entity mentioned within the metadata, which could point to a harmful term having been appropriated by the affected community itself. Examples for such cases are the "Queer Liberation March", an annual LGBT protest march in Manhattan, organised by the Reclaim Pride Coalition, or the name of the musical group "Gypsy Kings".</p> <p>The second of these additional methods is <strong>disambiguation</strong>, for which the tool makes use of <strong>Large Language Models</strong>. These models have been trained on a variety of big data sets in the relevant languages (i.e. the languages the DE-BIAS project has focussed on, namely Dutch, English, French, German, and Italian) providing a robust pool of reference points for the tool to "understand" the "meaning" of words depending on their context. The DE-BIAS vocabulary itself includes an indication, for which terms disambiguation should be applied, and an explanation of the context(s), in which a term is considered harmful and in which it isn&rsquo;t, as part of the contentious issue descriptions.</p> <p><span style="text-decoration: underline;"><strong>Note:</strong></span> While Named Entity Recognition is, by default, applied to all texts to be analysed, disambiguation is only applied for those terms, which the DE-BIAS vocabulary has flagged as ambiguous. Furthermore, disambiguation only works well when there is a detailed enough contentious issue description that helps in distinguishing between problematic and non-problematic uses of a specific term. Especially in the case of subtle bias, such a clear distinction might not always be possible.</p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@be8102fff4444f40a47811a0e7ab3672" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.5 How to use the DE-BIAS tool</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@4f426c1d6c8c47548bf3230a0765ecdb"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@4f426c1d6c8c47548bf3230a0765ecdb" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>The DE-BIAS tool is available in two main formats: as an API and as a standalone version. This part of the course can be used by people who are <strong>planning to use the tool</strong>, or people wanting to <strong>see how the tool works</strong>. This part is <strong>optional and not required to complete</strong> to receive a certificate for this course.</p> <p><span style="text-decoration: underline;"><strong>Note:</strong></span> The API can be used to include a debiasing step into one&rsquo;s own data workflows and processes, but requires the availability of a technical team or partner for the integration and potential adaptation. This is therefore not described in any further detail here, but instead the use of the DE-BIAS API as integrated into the Metis Sandbox for aggregators and data providers to Europeana.eu is explained. More information about the API can be found on GitHub.</p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@922ca78c2a264eee8dbab3d51dfe13cf" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.5.1 The standalone version of the DE-BIAS tool: Insert texts</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@8db5c9cedb544cce9818462e8890f4db"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@8db5c9cedb544cce9818462e8890f4db" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>The tool is available at&nbsp;<a href="https://debias-tool.ails.ece.ntua.gr/" target="_blank">https://debias-tool.ails.ece.ntua.gr/</a> and offers two input options:</p> <ul> <li>"Insert texts" and</li> <li>"Upload a file".</li> </ul> <p><img height="560" width="500" src="/assets/courseware/v1/42a3c346c58b9e9bdd2667e879be66fc/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_1.png" alt="DeBias tool 1" /></p> <p><em>Image: User interface of the standalone version of the DE-BIAS tool, open on the default tab "Insert texts".</em></p> <p>If users want to analyse a small amount of data for potential bias, the option of direct data input is their best choice. To run the tool over a selected text, choose the language, in which the metadata is written, from the drop-down menu under (1) and enter or copy your text into the free text field under (3). Using the option to "Add text" below the text field allows you to add more than one text and have them analysed together. To initiate the analysis simply click "Run".</p> <p><img height="155" width="501" src="/assets/courseware/v1/a6963c95d12f2144aefaaf4bdf7eb3f8/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_2.png" alt="DeBias tool 2" /></p> <p><em>Image: Section 2 of the user interface: Operations.</em></p> <p>Under (2) you also have the option to enable Named Entity Recognition (NER) and/or disambiguation. By default, both of these options are turned off. You can activate them independently by clicking the slider.</p> <p>Once you have run the tool you will be presented with an analysis report directly in the user interface of the standalone version: the problematic terms found by the tool will be highlighted in yellow. When you click on a highlighted term, a pop-up will appear including the URI of that term in the DE-BIAS vocabulary. Following this link will lead you to the complete description of the term, including - if available - suggestions for alternative terms to use.</p> <p><img height="462" width="500" src="/assets/courseware/v1/485ff9f22bbb6ce976e8a912b5811f6c/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_3.png" alt="DeBias tool 3" /></p> <p><em>Image: Analysis report for inserted texts.</em></p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@76d00a4d4c554fa094ac268150c87e64" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.5.2 The standalone version of the DE-BIAS tool: Upload files</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@3d3886eacdd04b17887677635155651c"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@3d3886eacdd04b17887677635155651c" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>Alternatively to entering or copying texts in the free text field, the standalone tool allows you to complete upload files to be analysed. Please note that, at the time of writing, the tool only supports the upload of plain .txt files provided in a .zip file and that you will need to ensure that the text encoding is set to UTF-8, so that the tool can recognise the language used in your files correctly.</p> <p><img height="488" width="500" src="/assets/courseware/v1/0e7f6dec89b8f7e779bac671553f023e/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_4.png" alt="DeBias tool 4" /></p> <p><em>Image: User interface of the standalone version of the DE-BIAS tool, open on the tab "Upload a file"</em></p> <p>You will again start by indicating the language under (1) and deciding to enable - or to disable - Named Entity Recognition and/or disambiguation under (2). Afterwards, click the "Upload .zip" button, which will lead you to your local file manager, from where you can select the file you want to upload.</p> <p><img height="115" width="499" src="/assets/courseware/v1/22e94f6eb4ee0ff0c3e88af76e53fb89/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_5.png" alt="DeBias tool 5" /></p> <p><em>Image: File uploaded and ready for processing</em></p> <p>A successful upload is indicated by the name of the uploaded file appearing next to the button. There will now also be a bin icon via which you can remove the uploaded file again. Under (4) you will then be asked to provide an email address, to which the analysis report will be sent. The report consists of a PDF file with some general statistics including:</p> <ul> <li>The total number of annotations, i.e. cases of contentious terms detected;</li> <li>The absolute number of files ("values") processed containing at least 1 detected term;</li> <li>The average number of terms detected per file;</li> <li>A distribution grouping the analysed files by the number of terms detected per file.</li> </ul> <p>Additionally, you will be sent a file in .json format, which lists the text values ("literal") and the annotations ("tags"), including their position ("start", "end", "length") per file ("id").</p> <p><img height="453" width="499" src="/assets/courseware/v1/ea4a5be51b4cd4ef5b7f10c7568b66f9/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_6.png" alt="DeBias tool 6" /></p> <p><em>Image: .json file included in the report sent when uploading a file to the DE-BIAS standalone tool</em></p> <p>This .json file can be used for further processing with the help of a technical team, e.g. to implement a user interface that supports the annotation of detected bias terms.</p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@a7065402c0f64e759bff4433d1fd151c" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.5.3 The DE-BIAS tool integrated into the Metis Sandbox</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@d4b8b713a32045d39272561392e94991"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@d4b8b713a32045d39272561392e94991" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>For those using the Metis Sandbox to pre-process and check their data before submitting them to Europeana.eu, there also is an option to request a bias report via the Sandbox. Once you have processed your data following <a href="https://europeana.atlassian.net/wiki/spaces/EF/pages/2104295432/Metis+Sandbox+User+Guide" target="_blank">the usual workflow</a>, there will be a link at the top right of your screen to "run report DE-BIAS".</p> <p><img height="214" width="500" src="/assets/courseware/v1/dbbe2b060f5a8a7a428168ac43048727/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_7.png" alt="DeBias tool 7" /></p> <p><em>Image: The header section of the Metis Sandbox with the option to run a bias report </em></p> <p>Click this link to start the analysis. A successful run of the DE-BIAS tool will be indicated by the link changing to "view report DE-BIAS". Another click will open the result of the bias detection in a pop-up, which will remain available as long as your dataset remains available in the Sandbox.</p> <p><img height="401" width="500" src="/assets/courseware/v1/b7d1a626e13bd5d8dd31a29df14e1d19/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_8.png" alt="DeBias tool 8" /></p> <p><em>Image: Bias report as shown in the Metis Sandbox</em></p> <p>At the top, you will have an indication in how many items biassed terms have been detected in total. For each item you will then see:</p> <ul> <ul> <li>Its Europeana ID (linked to the preview function of the Sandbox);</li> <li>The metadata field (from the Europeana Data Model), in which the biassed term(s) was/were found;</li> <li>The language detected for that metadata field;</li> <li>The literal value including the detected bias term(s).</li> </ul> </ul> <p>Similar to the "Insert text" option of the standalone tool, each detected bias term is highlighted and linked to the term&rsquo;s URI in the DE-BIAS vocabulary. Following this link will lead you to the complete description of the term, including - if available - suggestions for alternative terms to use.</p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@0452052a02bc4da889aad05773f36c8c" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.6 What to do with the DE-BIAS tool’s output</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@58eac84d891a44f4a6efd54106bd78a6"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@58eac84d891a44f4a6efd54106bd78a6" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>The DE-BIAS tool, in its variations as described above, first of all flags the detected bias terms and provides you as the user of the tool with some additional information, which differ from one version of the tool to the other.</p> <p>If you are mainly interested in an initial sample evaluation of your collections or want to look at a small(er) collection to start with, the standalone tool with its option to insert texts and get a direct analysis result provides you with the necessary basics. You might e.g. want to use its analysis report and the information from the linked vocabulary entries in combination with an exercise of applying the Wheel of Bias as a conversation starter. While the tool points you to cases of interest, the Wheel of Bias allows you and your team to discuss about the classification and severity of the detected bias and the vocabulary entries can support you in evaluating what options there are to contextualise or potentially replace biassed terms.</p> <p>A similar approach could be taken when you are a user of the Metis Sandbox and have requested the DE-BIAS report in this context. The links to the detected bias term&rsquo;s entry in the DE-BIAS vocabulary provide you with the necessary information about why a term has been flagged, which you can then countercheck by reviewing the complete metadata information of the item via the links to the item&rsquo;s preview in the Europeana.eu test environment. Furthermore, this report points you directly to the metadata field where the biassed term(s) was/were found, which might give you an indication whether there are certain metadata fields that require more or more immediate attention when debiasing your collections.</p> <p>You will also get an indication as to what to prioritise from the general statistics report included in the analysis of the standalone tool when using the option to upload files. Using this with data from several collections might point to collections with higher total numbers of files where biassed terms have been detected, i.e. collections where mitigation is needed more immediately. In addition, the .json file included in this option also gives you a direct way into technically processing and reacting to the detected bias.</p> <p>In the DE-BIAS project, for example, the information included in these .json files forms the basis for the annotations, which will also be visible to users of the Europeana.eu website, raising awareness amongst them about the issue of bias and helping them to understand the context(s), in which specific terms are contentious, respectively suggesting alternative terms that they can apply to their searches and when reusing the material they have found.</p> <p><img height="386" width="499" src="/assets/courseware/v1/1654ee1a5ad751fb99a57ad2d7593326/asset-v1:europeana+BIAS_01+2024Q4+type@asset+block/DE-BIAS_InteractiveLecture_Tool_9.png" alt="DeBias tool 9" /></p> <p><em> Image: Mock-up of the planned user interface extension on Europeana.eu with the detected bias term underlined and the pop-up with information from the DE-BIAS vocabulary open</em></p> </div> </div> </div> </div>

<div class="xblock xblock-public_view xblock-public_view-vertical" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="VerticalStudentView" data-course-id="course-v1:europeana+BIAS_01+2024Q4" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@vertical+block@cac8e0ad28884cef8f6eeb0741edb768" data-block-type="vertical"> <h2 class="hd hd-2 unit-title">5.7 Exercise 14: Using the DE-BIAS tool on your own</h2> <div class="vert-mod"> <div class="vert vert-0" data-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@8201b1d9cb6b4572b34f1c68116d4ece"> <div class="xblock xblock-public_view xblock-public_view-html xmodule_display xmodule_HtmlModule" data-graded="True" data-has-score="False" data-runtime-class="LmsRuntime" data-init="XBlockToXModuleShim" data-block-type="html" data-request-token="241cb6fa47f711f0b4770e526f83b440" data-runtime-version="1" data-usage-id="block-v1:europeana+BIAS_01+2024Q4+type@html+block@8201b1d9cb6b4572b34f1c68116d4ece" data-type="HTMLModule" data-course-id="course-v1:europeana+BIAS_01+2024Q4"> <script type="json/xblock-args" class="xblock-json-init-args"> {"xmodule-type": "HTMLModule"} </script> <p>Give the standalone version of the DE-BIAS tool a try yourself. Go to <a href="https://debias-tool.ails.ece.ntua.gr/" target="_blank">https://debias-tool.ails.ece.ntua.gr</a>/ and add texts to be analysed via the "Insert texts" tab.</p> <p>You can choose examples from this&nbsp;<a href="https://docs.google.com/document/d/1Yuy9CwDHshSPxxHENgQA5iPzuX0-MAlj/edit" target="_blank">collection of texts</a> or you can use texts from your own website or from Europeana.eu. Don&rsquo;t forget to indicate the language of the selected texts before running the tool.</p> <p>Play around with enabling or disabling the options for Named Entity Recognition and disambiguation to see how this might affect the analysis report.</p> <p>Should you come across any terms that are flagged as part of a named entity, but aren&rsquo;t, or terms where you think disambiguation might not work precisely enough, please reach out to the DE-BIAS team via <a href="mailto:project.debias@gmail.com" target="_blank">project.debias@gmail.com</a>.</p> </div> </div> </div> </div>