Welcome to October’s Monthly Focus! This month is about our dear PDF Importer.
Users just need to drag & drop a PDF term sheet in LexiFi Apropos to automatically capture a given product in the system. This process is instantaneous: product definitions are created in LexiFi at the very moment a user drops the file! So far, the system has reached just over 10,000 unique PDF term sheets since its launch back in 2016.
We asked one of our developers to tell us more, and he said it was too technical and probably a little boring, so we tried to keep it short:
Marc: So the first important info is that this tool has been created to help bankers and brokers speed up the product-booking process and reduce the risk of human errors when entering financial products into their database! The PDF importer scans term sheets and recognizes products characteristics using a combination of layout and keywords analysis.
Meriem: How so? And how does this work?
Marc: Each issuer has specific layouts and jargon. Sometimes, the same characteristic is expressed in various terms, even in term sheets from the same issuer… So we have first identified all important key terms such as dates, payoff formulas, etc.; then, we have coded what I like to describe as a 3-step program which: 1- Reads and understands the payoff 2- Extracts key information to input the product in the database 3- Translates this information into LexiFi parameters
If the program fails to understand the term sheet, we identify the reason and fix it so that it works the next time a similar product is imported! And the user gets a warning. Now if the program understands part of the term sheet, but not everything, it still books the product while producing notifications for the missing info.
Meriem: What were the biggest challenges?
Marc: There were loads of them! Perhaps the main challenge was to include all variations of the same term, even for the same issuer! We also had to account for typos and created a list with the most frequent ones. Some term sheets embed formulas as images, so we had to develop a special OCR (Optical Character Recognition) system in order to read payoff formulas.
The list is long! But the key takeaway is that the more people use the tool and feed it with term sheets, the better it gets!
Meriem: How do you make sure this is actually working well?
Marc: We have a whole testing infrastructure composed of statistical, quantitative and non-regression tests. For example, ALL products are repriced at their issue dates using LexiFi’s pricing libraries on a daily basis, and the results are then compared to the official issue prices. If prices are not consistent, the product is flagged! We also run automated non-regression tests on the entire back log every time we improve the tool! Since we are able to import 2,000 PDF term sheets per minute, we can really monitor closely our algorithm improvements.
Meriem: Thanks Marc! This is a powerful technology, its speed and completeness for product capture are remarkable! I will get back to you with questions from our readers!
We hope you enjoyed reading about the PDF importer! We’ll be glad to chat with you about this tool. We have a webpage listing the eligible types of products and issuers as well as a demo video!
You may find below a couple of questions as well as the tool’s page and video!