The Swiss Foundation Code (SFC) has become a highly respected and often applied tool for practitioners, experts, and civil servants in the philanthropic sector of Switzerland. To date, the SFC was available as textbook and digitally as PDF. This limited the possibilities to find answers on concrete questions and to search specific topics. Together with SwssFoundations, the national association for grant-making foundations, and funded by Gebert Rüf Foundation, a project team of CEPS created Smart SFC, a  web-based tool with the ability to provide semantic similarity search for the SFC. The Smart SFC searches the book’s headers and sentences not only using letters similarity, but also it considers the full meaning of the book’s lines and if it is relevant to the search keyword.

It supports the four languages that the book was published in: German, French, Italian and English. The semantic search mechanism is developed using Sentence-BERT, which is an advanced Large Language Model (LLM) tool that supports text encoding and similarity search. The architecture of the Smart SFC is not only bounded by the capabilities of Sentence-Bert. When dealing with PDFs, the knowledge about the layout and the document structure are very essential to export the texts, detect a full paragraph and distinguish between headers and sub-headers. Therefore, AmazonTextract was used in the preprocessing of the book’s text as it gives the capability of detecting the layout in a PDF document and distinguishing between different texts types.

To perfectly grasp the mechanism of this tool, a search keyword like “Financial plan” that does not actually exist in the book will return results containing “Finance”, “Financial planning”, “Financial management”, “Investment strategy”, “Funding” and “Asset management”. Also, the user has the capability of downloading the search results or saving them into a workspace to view them later. Moreover, the developed tool is designed to auto-learn and enhance its results based on the user’s activity to guarantee better user experience over time.

You can search the Smart SFC here.

Project team: Omar Hussein, Nour Shokry, Georg von Schnurbein