# LeNet Document Recognition **Entity class:** Convolutional neural network lineage ## Definition and history LeNet names the convolutional recognition lineage associated with Yann LeCun and colleagues' work on handwritten ZIP codes and later document reading. [[articles/Intelligence Inside|Intelligence Inside]] separates the research input of postal handwriting from the commercial path into NCR bank-check readers. The 1998 document-recognition paper describes LeNet-5 and applications in check processing. Recognition of handwritten marks at scale was a deployed financial operation before conversational AI became the public reference point. ## Mechanism and limits A convolutional network learns feature detectors from image data and uses them to classify visual input. A training corpus from postal ZIP codes does not mean the later deployed NCR system was the USPS mail sorter. The distinction between research data provenance and product deployment is the central historical edge here. ## Relationships **Research lineage:** Yann LeCun and collaborators at [[wiki/Bell Labs|Bell Labs]]. **Commercial application:** NCR check readers. **Parallel postal systems:** [[wiki/CEDAR Document Recognition|CEDAR Document Recognition]] and [[wiki/SRI Postal Address Recognition|SRI Postal Address Recognition]]. **Article route:** [[articles/Intelligence Inside|Intelligence Inside]]. ## Sources [LeCun et al., *Gradient-Based Learning Applied to Document Recognition* (1998)](http://vision.stanford.edu/cs598_spring07/papers/Lecun98.pdf); [[articles/Intelligence Inside|Intelligence Inside]].