This repo is a collection of codes our group wrote for the project of DSO560 NLP. Our client is Behold. Behold would like to partner with USC’s DSO560 NLP class to develop a recommendation system that is based on the retail fashion brand’s brand descriptive information.
Professor: Yu Chen
Team Black:
- Yui Li
- Jiaxin Lu
- Guoqiong Sun
- Kaiming Yu
- Jinnan Xue
There are mainly two problems we need to solve:
- A common problem for Behold is the extremely manual classification of new products into their specific brands. Behold has supplier relationships from many global clothing vendors and marketplaces, but has found that the bottleneck for scaling out the number of products they can sell is identifying the brands associated with each new product. Use the brands’ biographical description (behold_brands.csv), product descriptions (products.csv) and any externally collected metadata, build an NLP classification model to predict which brand a new product should be assigned.
- For this problem, you can find our solution here.
- Create a brand recommender algorithm that would recommend an outfit given a customer’s search query.
- For this problem, you can find our solution here.