swissgeol.ch

Did you already use generative AI to unlock new possibilities in geological data management? We discovered a very valuable use case! The challenge ⌨️ Our internal Swissgeol database of geological documents has more than 14000 entries in various languages. Many of these documents were imported from a legacy system, where titles had to be entered in all-uppercase, for example: “N14,BAUGRUNDUNTERSUCHUNGEN FUER REUSSBRUECKE BEI GISIKON”. Fixing the capitalization of these document titles is important, not just to improve the readability, but also to improve the accuracy of full text search by restoring certain accented characters (e.g. writing “für” instead of “FUER”). First solution attempt 🛠️ Manually rewriting thousands of titles is not a productive use of our time. Therefore, we tried to define rules to automate this, but quickly ran into issues. How do we know when “INS” is the regular German word “ins” (all lowercase) or refers to the community of Ins in the canton of Bern (with an uppercase “I”)? How do we decide when “UE” should be converted to “ü” (e.g. “für”, “Brücke”) and when to preserve the “ue” (e.g. in place names such as “Uetliberg” and words such as “Quelle”)? Soon, we found ourselves designing a complex pipeline that involved methods such as part-of-speech tagging and named entity recognition, without making significant progress towards a satisfactory solution. Solution with generative AI 🤖 Then, a new idea popped up: can large language models (LLMs) fix the titles for us? Initially, we did not expect an LLM to be able to understand the subtleties of the task and the many obscure geographical names that occur in the titles. However, we were blown away by the results! After just a few hours of experimentation, the LLM was giving us much better results than the rule-based pipeline that we were already working on for several days. Did you already use generative AI in a similar way? Do you see other use cases for LLMs in the management of geological data? Let us know in the comments!

Introducing boreholes.swissgeol.ch – Now Live!   In addition to assets.swissgeol.ch (https://lnkd.in/ddfBxS2p), our borehole data platform is now publicly available at boreholes.swissgeol.ch!   🔎 What’s Available? The platform currently features data from 15 boreholes. While this may seem like a small start, it’s just the beginning! As we refine our quality control processes, the dataset will grow, and we’re committed to regularly expanding the number of borehole records. However, these datasets are fully accessible - no login required - so you can easily explore and interact with the platform’s features. Additionally, our REST API allows you to integrate borehole data directly into your systems (e.g., https://lnkd.in/d6fQ_Uev , https://lnkd.in/dr6fWFN2).   📈 What’s Coming Next? We’re constantly evolving the platform. Over the next year, we plan to expand the dataset and enhance the functionality, leveraging the latest advancements in data science (https://lnkd.in/duep_C4S). These improvements will streamline how we digitize borehole data, increase data accuracy, and simplify data management processes.   🌍 Why It Matters? Digitalizing borehole data is a critical step toward better understanding the subsurface. This will help decision-making, reduce risks, and improve outcomes in sectors like resource exploration, construction and environmental planning. To ensure data quality, the platform is integrated with our internal production system, where every dataset undergoes quality control before being automatically published.   👥 Join Us on This Journey! We’re committed to continuous improvement and encourage the community to engage with the platform. Your feedback is invaluable as we work to expand the data and functionality.   Explore the platform today at boreholes.swissgeol.ch, and stay tuned for future updates on swissgeol.ch!   #Boreholes #DataManagement #PublicPlatform #Geoscience #DigitalTransformation #MachineLearning #DataScience

🚀 Exciting news for Geologists, Scientists, Urban Planners and Interested Parties: Launch of assets.swissgeol.ch 🌍 assets.swissgeol.ch, a new swissgeol application, is providing free access to geological data from all over Switzerland. Designed to help users find relevant geological information based on keywords or geographic locations, this platform represents a significant step forward in data accessibility. Why assets.swissgeol.ch? assets.swissgeol.ch replaces the previous nine "Geological Documents" layers on map.geo.admin.ch, which were taken offline in June 2022. It is part of the “Digitisation of the Geological Subsurface” action plan approved by the Federal Council in May 2021. A key measure of this initiative is to continuously update and reorganise geological datasets to improve both quality and accessibility. The new datasets focus on freely available geological information and documents of national interest (e.g., studies by federal offices, Swiss Federal Railways, hydrogeological or geothermal studies, etc.), which can be downloaded directly as PDF files. What can you do with assets.swissgeol.ch? As shown in the images below, assets.swissgeol.ch allows you to: 🔍 Search for geological data easily and efficiently 📊 Access detailed subsurface information and reports 🛠️ Support research, project planning, and decision-making with reliable geological data Whether you're a geologist, urban planner, environmental scientist, or just curious about Switzerland's subsurface, assets.swissgeol.ch aims to make your work easier. 👉 Stay tuned for updates on this application and the entire swissgeol suite. Ready to explore? Check it out now! 👉 assets.swissgeol.ch. #SwissGeol #GeologicalData #DataAccess #ResearchTools

Take a look! 😊

We are looking for new colleagues for Bundesamt für Landestopografie swisstopo

Last week, Stijn Vermeeren, Project Manager Data Science at the Federal Office of Topography swisstopo and part of the swissgeol.ch team, was invited to present at the prestigious CNO Academy event in Bern 🏛. Together with David Cleres from our partner Visium, we showcased our work on automatic data extraction from borehole profiles. The talk was part of the "AI and Sustainability" track, organized by another swissgeol.ch partner company, EBP. 🤝 We are getting a lot of interest and positive feedback around this project, which is very motivating! 👏🚀

Borehole profiles are a key source of knowledge about the subsurface ⛰. Therefore, swisstopo has already collected tens of thousands of boreholes profiles, mostly as PDF files with free-form layouts. Older documents, scanned from paper archives, are only available as images without any machine-readable text. To make borehole data 🗃🎞 accessible for geological modelling 🗺📈, we need to bring this data into a uniformly structured digital database 📅. Manually entering data from boreholes profiles into a structured database is a very error-prone and costly process. For this reason, we’ve started looking into automating this process as much as possible 🤖. Existing solutions for digitizing documents struggle with correctly parsing the specific layouts of borehole profiles. Particularly challenging is the fact that material descriptions are not always horizontally aligned with the corresponding depths, often requiring you to read diagonally in order to match the descriptions with the correct layer depths. This is why we partnered with Visium to develop our own data extraction solution for borehole profiles. The results so far are very promising, as you can see from the attached images: - A typical borehole profile, where depths and corresponding material descriptions are not always horizontally aligned. 📋 - A visualization of how our automatic data extraction pipeline parses the data and correctly associates each material description with the corresponding depths. 🔀 - The same result represented in a machine-readable JSON format, which can be used for further automatic processing. 💻 The code is published as open source software on Github (https://lnkd.in/euczDusi). We’re looking forward to presenting more details about this project to you in the near future!

We are pleased to announce the publication of a public tender on simap.ch, inviting qualified organizations to participate in the call for tender or share it with their networks. The contract focuses on the implementation, adaptation, and maintenance of a platform for hosting, managing, and publishing controlled vocabularies in the field of geology. For more details and to participate, please visit the tender announcement: https://lnkd.in/e7Jqg8-T. We look forward to your participation! #geology #controlledvocabularies #semanticweb #publictender #cloudbasedsolutions

Remember our "Middle Jura" Chronostratigraphy challenge a few months back - https://lnkd.in/e7maNnb4 ? Your positive feedback was a huge motivator — thank you! We’re excited to share an update on lexic.swissgeol.ch, our pilot platform for building geological vocabularies and testing features like semantic queries. The platform is really taking shape! We're working on vocabularies, data augmentation, and enhancing query capabilities. Once released, lexic.swissgeol.ch will allow exploration through two approaches: the main landing page, where vocabularies will be available, or via a web-map viewer. The viewer is based on the Tectonic Map of Switzerland (1:500,000), and we're planning to integrate GeoCover soon. Our first semantic search feature is live, allowing users to include narrower terms from the Tectonic Units vocabulary. We're also developing this functionality for Chronostratigraphy, which will make it easier to select time intervals or related units. What's next? We're integrating GeoCover into the web-map viewer and finishing up the Chronostratigraphy search feature. With that in place, we want to make lexic.swissgeol.ch available for testing. By the end of the year, we aim to publish the Lithostratigraphy and Lithology vocabularies, too. 𝗗𝗶𝘀𝗰𝗹𝗮𝗶𝗺𝗲𝗿: 𝘗𝘭𝘦𝘢𝘴𝘦 𝘯𝘰𝘵𝘦 𝘵𝘩𝘢𝘵 𝘭𝘦𝘹𝘪𝘤.𝘴𝘸𝘪𝘴𝘴𝘨𝘦𝘰𝘭.𝘤𝘩 𝘪𝘴 𝘴𝘵𝘪𝘭𝘭 𝘶𝘯𝘥𝘦𝘳 𝘥𝘦𝘷𝘦𝘭𝘰𝘱𝘮𝘦𝘯𝘵, 𝘢𝘯𝘥 𝘵𝘩𝘦 𝘜𝘙𝘐 𝘪𝘴 𝘯𝘰𝘵 𝘳𝘦𝘴𝘰𝘭𝘷𝘢𝘣𝘭𝘦 𝘢𝘵 𝘵𝘩𝘦 𝘮𝘰𝘮𝘦𝘯𝘵. Stay tuned for more updates!

Exciting news for the geothermal exploration community! The borehole data for Vinzel-1, Vinzel-1S, and Vinzel-1 Malm has been published on https://map.geo.admin.ch This extraordinary and comprehensive exploration dataset is now freely available on the swisstopo platform: Vizel-1: https://lnkd.in/dQWwiwb7 Vizel-1S: https://lnkd.in/dzrXVQmJ Vizel-1 Malm: https://lnkd.in/dc4CCXv4 This dataset offers invaluable insights for anyone involved in geothermal exploration, drilling, and well analysis. Whether you're interested in mudlogging, MWD, LWD, well geometry, FITs, LOTs, wireline logging (including sonic and borehole image logs), or EOWR, this resource is a must-have. Feel free to explore and share this valuable data from this geothermal exploration borehole in western Switzerland!