[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-122695-en":3,"doc-seo-122695-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},122695,962075006959,"Anda","https://ap-avatar.wpscdn.com/avatar/e0002397efbe92a78e?_k=1776741047341049297",8,"Research & Report","Natural Language Explanations for Machine Learning Classification Decisions","Natural language explanations for machine learning classification decisions are developed to address stakeholder demand for understandability of both predictions and their underlying reasons. A dataset of expert-written textual explanations paired with numerical explanations is introduced as a data-to-text generation task. BART and T5 are fine-tuned to linearize information from explainable output graphs and generate fluent, largely accurate narratives. Experiments show reduced error with dataset augmentation, and question answering on numerical explanations reaches 91% accuracy with T5.","Natural Language Explanations for Machine Learning Classification Decisions  \nJames Burton  \nDepartment of Computer Science Durham University  \nDurham, UK  \n[james.burton@durham.ac.uk](james.burton@durham.ac.uk)  \nNoura Al Moubayed  \nDepartment of Computer Science Durham University  \nDurham, UK  \n[noura.al-moubayed@durham.ac.uk](noura.al-moubayed@durham.ac.uk)  \nAmir Enshaei  \nFaculty of Medical Sciences Newcastle University  \nNewcastle, UK  \n[amir.enshaei@newcastle.ac.uk](amir.enshaei@newcastle.ac.uk)  \nAbstract—This paper addresses the challenge of providing understandable explanations for machine learning classification decisions. To do this, we introduce a dataset of expert-written textual explanations paired with numerical explanations, forming a data-to-text generation task. We fine-tune BART and T5 language models on this dataset to generate natural language explanations by linearizing the information represented by explainable output graphs. We find that the models can produce fluent and largely accurate textual explanations. We experiment with various configurations and see that an augmented dataset leads to a reduced error rate. Additionally, we probe the numerical explanations more directly by fine-tuning BART and T5 on a question-answer task and achieved an accuracy of 91% with T5.  \nIndex Terms—explainability, data-to-text, natural language generation  \nI. INTRODUCTION  \nDespite the advantages of using machine learning (ML) techniques to solve complex problems, there are particular application areas, such as finance, health, and criminal justice, that have been hesitant to adopt ML approaches [1]–[3] with stakeholders concerned about the consequences a wrong decision could have. For these areas in particular, the notion of explainability is critical. Stakeholders not only want to know what the model is predicting but also why. Understanding the factors influencing an ML model’s prediction enables actionable business choices, transparency, and confidence. Furthermore, this aligns them with the recently proposed EU Artificial Intelligence Act [4] .  \nSimple predictive algorithms such as linear models, generalized additive models, and shallow decision trees are inherently explainable since they are easy to understand, and sourcing the reason for classification output decisions is simple [3], [5] . However, for complicated architectures such as deep neural networks, it is challenging to trace which features were relied upon most for making the decision [6], [7] as these black-box models employ billions of parameters to make predictions, making them difficult to troubleshoot and trust.  \nIn recent years, there has been an effort to increase transparency in the decision-making process of black-box models used for predictions and incorporate eXplainable AI (XAI) techniques. In a typical explainability pipeline, as shown in Fig. 1 (left), a trained classifier will make a prediction. To make a local-level explanation, the XAI technique will utilize  \nthe prediction and the classifier to yield importance scores for each input feature.  \nFour commonly used eXplainable AI (XAI) techniques include Local Interpretable Model-Agnostic Explanations (LIME) [8], SHapley Additive exPlanations (SHAP) [9], Integrated Gradients (IG) [10], and Layer-wise Relevance Propagation (LRP) [11] . Although distinct, these techniques all produce feature importance values that quantify the contribution of each feature to the prediction.  \nGraphs and figures are commonly used to communicate the contributions of each variable used to arrive at a given prediction. For example, Fig. 3 shows a graph generated using LIME to explain why a given “wine” was labeled as “high quality”. These graphs produced by XAI techniques indicate which features are positive (supporting the prediction output), negative (contradicting the prediction output), and neutral (having a negligible influence on the prediction decision) . However, for non-experts, it can be challeng","cbCaipHGd12IWhx6","https://ap.wps.com/l/cbCaipHGd12IWhx6","pdf",1111921,1,10,"English","en",105,"# Introduction\n## Explainability needs for high-stakes domains\n## From interpretable models to black-box neural networks\n## XAI techniques and feature importance graphs\n## Narrative generation with TEXEN dataset\n## Numerical explanations as question answering","[{\"question\":\"What problem does the paper address for machine learning classification?\",\"answer\":\"It tackles the difficulty of providing understandable, natural-language explanations for why a model makes a specific classification decision.\"},{\"question\":\"How is the TEXtual Explanation Narratives (TEXEN) dataset constructed?\",\"answer\":\"It pairs local-level explainability outputs with expert-written textual narratives, enabling a data-to-text generation formulation.\"},{\"question\":\"Which models and methods are used to generate explanations, and how effective are they?\",\"answer\":\"BART and T5 are fine-tuned to produce narratives from linearized information in explanation graphs, yielding fluent and largely accurate text; dataset augmentation reduces error and a QA probe achieves 91% accuracy with T5.\"}]","Natural Language Explanations for Machine Learning Classification Decisions | PDF",1785812298,25,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"natural-language-explanations-for-machine-learning-classification-decisions","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/natural-language-explanations-for-machine-learning-classification-decisions/122695/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-04",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address for machine learning classification?","Question",{"text":75,"@type":76},"It tackles the difficulty of providing understandable, natural-language explanations for why a model makes a specific classification decision.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How is the TEXtual Explanation Narratives (TEXEN) dataset constructed?",{"text":80,"@type":76},"It pairs local-level explainability outputs with expert-written textual narratives, enabling a data-to-text generation formulation.",{"name":82,"@type":73,"acceptedAnswer":83},"Which models and methods are used to generate explanations, and how effective are they?",{"text":84,"@type":76},"BART and T5 are fine-tuned to produce narratives from linearized information in explanation graphs, yielding fluent and largely accurate text; dataset augmentation reduces error and a QA probe achieves 91% accuracy with T5.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":21,"slug":133},"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]