Decoding Philippine Grocery Receipt Item Names Using K-Nearest Neighbor Search and Long Short-Term Memory Network

Item names found in grocery receipts are often abbreviated due to the receipt paper character limit. This makes the item names difficult to read and decode which can cause confusion to customers. This study looked into using K-Nearest Neighbor (KNN) and Long short-term memory (LSTM) methods of decoding grocery receipt item names. A total of 17 receipt images from four different grocery stores were used as data. A total of 95 unique grocery item names were identified in the receipt images. A mobile application was developed as the interface for users. The KNN search achieved an average similarity ratio of 92.63% while LSTM achieved 76.55%. Through convenience sampling of 25 grocery shoppers, the application was evaluated using the System Usability Scale (SUS) and got a mean SUS score of 85.00 indicating acceptability of the application.