Mining cross-domain rating datasets from structured data on twitter

Abstract

While rating data is essential for all recommender systems research, there are only a few public rating datasets available, most of them years old and limited to the movie domain. With this work, we aim to end the lack of rating data by illustrating how vast amounts of ratings can be unambiguously collected from Twitter. We validate our approach by mining ratings from four major online websites focusing on movies, books, music and video clips. In a short mining period of 2 weeks, close to 3 million ratings were collected. Since some users turned up in more than one dataset, we believe this work to be amongst the first to provide a true cross-domain rating dataset.

Utilizing emotion scores for improving classifier performance for predicting customer's intended ratings from social media posts

“…The combination of sentiment and emotion scores yielded an increase in accuracy of 19.41%, 27.83% and 40.20% in cases of e-learning, OFD and OTA services, respectively, over only sentiment-based classification technique. The prediction of more accurate data sets from recent data can help to solve the issues related to limited vast data set containing recent data as pointed by Dooms et al (2013) and Dooms et al (2014). Additionally, accurate prediction of ratings will help both prospective customers and service providers in saving the time spent in reading the post since users prefer to spend very less time reading a post (Patel, 2016; Read, 2016).…”

“…However, establishing ratings of social media comments can help both service providers as well as the customers. Second, predicting the ratings will help organizations develop a data set of ratings from the most recent posts rather than depending on older data sets like MovieLens for developing recommender systems (Dooms et al , 2013). This makes detection of ratings for social media feeds a relevant study.…”

“…However, the authors have proposed the need for checking content polarity for improving the results further. Dooms et al (2013) prepared a data set “MovieTweeting,” which contained ratings extracted from structured tweets to build a better data set than the “MovieLens” data set, where the ratings do not contain recent reviews and ratings. Dooms et al (2014) extended their work to test it on three more cases (Youtube, Pandora and Goodreads).…”

“…Dooms et al (2013) prepared a data set “MovieTweeting,” which contained ratings extracted from structured tweets to build a better data set than the “MovieLens” data set, where the ratings do not contain recent reviews and ratings. Dooms et al (2014) extended their work to test it on three more cases (Youtube, Pandora and Goodreads). However, the authors have noted high sparsity values.…”

“…From a business perspective, the ratings are of importance to organizations since checking customer's ratings provide an easy overall view of what the customer wants to convey. Additionally, Dooms et al (2014) have voiced the need for large data sets containing recent data from building better recommender systems. This makes the detection of ratings for social media feeds a relevant study.…”