Dynamic Recommender System using Gift-wrap Messages on E-commerce Websites
Publication Date: 2016-Oct-06
The IP.com Prior Art Database
Many E-commerce websites provide users with extra options while placing order to gift wrap their products and attach a card to it with ones personalized message. The aim is to create a greeting message recommender system by learning attributes like Occasion for Gifting, Relationship between Sender and Receiver, Gender of Sender and Receiver, etc. from the gift messages and associating them to the corresponding order.
Page 01 of 4
Dynamic Recommender System using Gift -wrap Messages on E-commerce Websites
Disclosed is a system which recommends greeting messages based on the requirements mentioned by the user like occasion, relationship with the receiver and/or disambiguation on receiver's gender using the gift wrapping messages on products purchased from e-commerce sites. The system involves three major stages:
1 System Bootstrap
This stage involve a series of steps that need to be employed in order to create the system.
1.1 Data Preparation Module
Each Gift message contains the fields: Message, Sender's name and Receiver's name. To create the training data for the system, the messages need to be labeled with occasion, age group and relationship tags. These tags can be either hand labeled or might be assigned computationally based on keywords. The tags based on Sender's and/or Receiver's gender can be easily computed using various online available sociolinguistics databases. For the classifier, training data is created of the form
p of Receiver
Gender of Receiver
Gender of Sender
Here's a book I'm sure you'll enjoy. It's a psychological thriller with some mind-blowing twists. Brutal at times, and touching in others.
As a dad, I suspect you'll shed a few tears (like I always do) when you read the soccer scene in the closing pages (but don't look ahead!).
We love you very much. We thought by giving you a gift card you could get something
you especially desired. Merry Christmas, Sweet boy, We love
Aunt Erica and Micah and Luke
BDay, senior! Loadz of love <3
Table 1: Sample training data for bootstrapping
1.2 Message Feature Extractor Module
Message Feature Extractor Component aims at creation of module which mines features from gift message learnt from labeled message data set. One possible implementation is by use...