[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-124708-en":3,"doc-seo-124708-105":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":20,"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},124708,8796095360427,"Lucas Martin","https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d",8,"Research & Report","Machine learning to detect corporate greenwashing - Master’s Thesis","This master thesis develops an automatic approach for detecting corporate greenwashing by collecting company-related data and applying fact-checking to identify green claims within it. Data is gathered through custom web scraping, including recursive scraping for the Fauna dataset and link-visit scraping for the CICERO Shades of Green dataset. Preprocessing ensures model compatibility: long Fauna texts are summarized with a Transformer and irrelevant items are removed; Cicero texts receive augmentation. Transfer learning with zero-, one-, and two-shot learning is evaluated, and fine-tuned Cicero models are trained for comparison.","| \u003Cbr>Faculty of Science and Technology\u003Cbr>MASTER’S THESIS |  |\n| --- | --- |\n| Study program/ Specialization:\u003Cbr>Data Science | Spring semester, 2023\u003Cbr>Open / Restricted access |\n| Writer:\u003Cbr>Audun Stjernelund Lien | \u003Cbr>…………………………………………\u003Cbr>(Writer’s signature) |\n| Faculty supervisor:\u003Cbr>Vinay Jayarama Setty |  |\n| Thesis title: Machine learning to detect corporate greenwashing. |  |\n| Credits (ECTS): |  |\n| Key words:\u003Cbr>- Web scraping\u003Cbr>- Transformers\u003Cbr>- Natural Language Processing\u003Cbr>- Mul9-class classifica9on\u003Cbr>- Mul9-label classifica9on\u003Cbr>- Transfer-learning | Pages: 54\u003Cbr>+ enclosure: 6\u003Cbr>Stavanger, 13/06/2023 Date/year |\n\nFront page for master thesis  \nFaculty of Science and Technology Decision made by the Dean October 30th 2009  \nFaculty of Science and Technology Department of Electrical Engineering and Computer Science  \nMachine learning to detect corporate  \ngreenwashing  \nMaster’s Thesis in Computer Science  \nby  \nAudun Stjernelund Lien  \nInternal Supervisors  \nVinay Jayarama Setty  \nReviewers  \nReviewer 1  \nReviewer 2  \nJune 13, 2023  \nAbstract  \nThis master thesis focuses on developing an automatic approach to detect corporate greenwashing. To achieve this, data must be collected, and green claims found from this data must be fact checked. The first step is to collect data by scraping. The web scrapers in this thesis were designed to extract comprehensive information about companies from their websites and reports using two datasets as benchmarks. The Fauna dataset was scraped using a recursive web scraper that extracted data from sub-pages linked to each company’s website. The CICERO Shades of Green dataset was scraped using a scraper that visited each link in the dataset to extract the text from each report made by CICERO.  \nThe collected datasets underwent preprocessing to ensure compatibility with machine learning models. The texts scraped from the Fauna dataset were often excessively long due to the abundance of information on the websites. These texts were summarized using a Transformer model, and irrelevant texts were manually removed from the dataset. In the case of the Cicero dataset, text augmentation was applied to expand the dataset and investigate its impact on model performance.  \nTo address the limited data availability, transfer-learning techniques including zero, one, and two-shot learning were applied to both the Fauna and Cicero datasets. These techniques leverage pre-trained models to learn from a small amount of labeled data. Additionally, fine-tuned models were implemented specifically for the Cicero dataset to provide a basis for comparison. The trained models achieved superior performance to the transfer-learning models, suggesting that training large models with limited training data remains an effective approach.  \nAcknowledgements  \nI wish to sincerely thank my supervisor Vinay Setty for his excellent guidance through the assignment. His knowledge of AI models, natural language processing and claim detection helped me through many issues. I would never have gotten as far as I did without his expertise.  \nContents  \nAbstract iii  \nAcknowledgements iv  \n1 Introduction 1  \n1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 1  \n1.2 Objectives .................................... 2  \n1.3 Approach and Contributions .......................... 3  \n1.4 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4  \n2 Related Work 5  \n2.1 Corporate Greenwashing ............................ 5  \n2.1.1 Machine Learning for Corporate Greenwashing Detection ..... 6  \n2.2 Multi-Label Text Classification Using Transformers ............. 7  \n2.3 Transfer Learning ................................ 7  \n2.3.1 Zero-Shot Learning ........................... 8  \n2.3.2 One-Shot Learning ........................... 8  \n2.3.3 Few-Shot Learning ........................... 9  \n2.4 Fact checking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .","cbCailop1m9x3ioy","https://ap.wps.com/l/cbCailop1m9x3ioy","pdf",5671620,1,66,"English","en",105,"# Abstract\n# Acknowledgements\n# 1 Introduction\n## Background and Motivation\n## Objectives\n## Approach and Contributions\n## Outline\n# 2 Related Work\n## Corporate Greenwashing\n### Machine Learning for Corporate Greenwashing Detection\n## Multi-Label Text Classification Using Transformers\n## Transfer Learning\n### Zero-Shot Learning\n### One-Shot Learning\n### Few-Shot Learning\n## Fact checking\n### Fake News Detection\n## Green rating\n### Fauna\n### CICERO\n## Scraping\n# 3 Approach\n## Collecting Data from Organizations With a Custom Web Scraper\n### Porting the Web Scraper to Python\n### Expanding the Scraper\n## Automating Green Impact Identification Using Machine Learning\n### Problem Definition\n### Challenges\n### Categorizing the Dataset\n### Balance the Dataset\n### Summarizing Scraped Text\n### Manually Checking the Texts\n### A Multi-Label, Zero-Shot Learning Approach\n## CICERO Shaded of Green\n### Scraping the Dataset\n### Extracting Relevant Text\n### Balance the Dataset\n### Data Augmentation and Undersampling\n### Correcting Texts\n### Supervised Training of the Dataset\n### Transfer Learning on the Cicero Dataset\n# 4 Experimental Evaluation\n## Experimental Setup\n## Models\n## Trained Models","[{\"question\":\"How does the thesis collect data for corporate greenwashing detection?\",\"answer\":\"It uses custom web scraping. The Fauna dataset is collected with a recursive scraper over company sub-pages, while the CICERO Shades of Green dataset is collected by visiting each report link and extracting the report text.\"},{\"question\":\"What preprocessing steps are applied to scraped texts before modeling?\",\"answer\":\"Fauna texts are often overly long, so they are summarized using a Transformer model and irrelevant texts are manually removed. For the CICERO dataset, text augmentation expands the dataset to study its effect on model performance.\"},{\"question\":\"Which learning strategies are used to handle limited labeled data?\",\"answer\":\"Transfer learning is applied using zero-, one-, and two-shot learning with pre-trained models to learn from small labeled sets. Additionally, fine-tuned supervised models are implemented for the CICERO dataset to provide a comparison baseline.\"}]","Machine learning to detect corporate greenwashing - Master’s Thesis | PDF",1785894027,166,{"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},"machine-learning-to-detect-corporate-greenwashing-masters-thesis","",{"@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/machine-learning-to-detect-corporate-greenwashing-masters-thesis/124708/",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","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"How does the thesis collect data for corporate greenwashing detection?","Question",{"text":75,"@type":76},"It uses custom web scraping. The Fauna dataset is collected with a recursive scraper over company sub-pages, while the CICERO Shades of Green dataset is collected by visiting each report link and extracting the report text.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What preprocessing steps are applied to scraped texts before modeling?",{"text":80,"@type":76},"Fauna texts are often overly long, so they are summarized using a Transformer model and irrelevant texts are manually removed. For the CICERO dataset, text augmentation expands the dataset to study its effect on model performance.",{"name":82,"@type":73,"acceptedAnswer":83},"Which learning strategies are used to handle limited labeled data?",{"text":84,"@type":76},"Transfer learning is applied using zero-, one-, and two-shot learning with pre-trained models to learn from small labeled sets. Additionally, fine-tuned supervised models are implemented for the CICERO dataset to provide a comparison baseline.","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},[93,97,101,105,110,115,120,123,128,131,135],{"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":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]