Most product documentation is stored in PDF files. This means that you are dealing with unstructured data such as images, tables and text all in the same file. You may need to know a particular specification such as volts or charger type, or check compatibility with another product, or see if there's a technical drawing present. Manually opening the PDF file and going through the contents for that exact bit of detail you want can take anywhere from a few seconds to a few hours.
Q: Does this product come in black?
Model scans all product documents, understands that obsidian and charcoal are variants of black, and returns:
A: Yes. The product comes in obsidian and charcoal.
Q: What is the type of switch of [enter MPN here]?
Model goes through the table containing products and extracts the type of switch for the MPN you asked for.
A: The switch type of [MPN] is an integral diaphragm.
Q: What are the applications?
If there’s an Applications section in the PDF, the model will find it, and return:
A: Basement sumps, dewatering, and water transfer.
With extensive experience working with machine learning models, including LLMs, our chat bots do not hallucinate. We have deployed B2B product data solutions across industries and domains, and our algorithms can tell a lifestyle image from a primary product image, a user guide from a brochure, an infographic from an engineering drawing, a safety specification from a product feature - and everything in between.
