Implementation of web product search engine

• Main Author: Fu, Zixiang.
• Other Authors: Hoi Chu Hong
• Format: Final Year Project
• Language: English
• Published: 2011
• Subjects: DRNTU::Engineering::Computer science and engineering::Information systems::Information storage and retrieval
• Online Access: http://hdl.handle.net/10356/46342
• Institution: Nanyang Technological University

Online shopping is gaining popularity in recent years due to the convenience of purchasing products through the Internet without the need to be physically present at the shop. One challenge that online shops (e-Commerce) are facing today is allowing its consumer to look for the product they are searching within a short period of time. This in term leads to the popularity of search engine, which aims to answer a query in the shortest possible time with the most relevant results. Traditional search engine requires its user to formulate a query that describes the information that he/she needs. This is especially useful when each record in the database of the search engine contains plentiful text description. However, in the context of an e-Commerce, the capability of the search engine to allow image as a query is an additional yet useful option for users. This function minus the need for a user to formulate a query that may be subjective or unclear, it instead searches for products that look similar to the image submitted based on the images‟ features. In this project, a backend text search engine and a backend image search engine were implemented to support an e-Commerce website known as Visual Image Search Engine (VISE). The text search engine is fully based on the inverted index implemented by Lucene, which is one of the most popular open source text-based search engine used today. The image search engine incorporated two types of search services. The first service extracts global features from an image and stores them in a data structure using Multi-Probe Locality Hashing. The second service implemented the bag-of-words model based on SIFT features, where visual words are used as a representation of an image and stored in an inverted index for searching. The text and image search engine implemented in this project support an online shopping website which takes certain real-time issues (heavy incoming traffic, large dataset etc) into concern. Code elegance, feasibility and a well-designed system are also a priority to ensure that new modules can be easily integrated into the current solution in the future.