Social Media Sentiment Analysis - graphlab.sentiment_analysis

User 5244 | 7/12/2016, 7:37:03 AM


I have two questions about: graphlab.sentiment_analysis module.

I'm working now on my project work during my Master and it's about Sentiment Analysis about Refugees on Facebook and Twitter in German Community.

  1. Since my data is unlabeled Data (Comments and Tweets), I wanted to ask how I should work with such data on Graphlab.

  2. Is there any option for the languages except English, in this case: German?

I really appreciate any help you can provide.

Best, Vahid


User 5371 | 7/15/2016, 9:29:23 PM

Hi Vahid,

For second question, we do not currently support languages other than English.

For first question, GraphLab Create does ship with built-in model for sentiment analysis, so you do not need have labeled data at all. Here is a very simple example from user user guide here

>>> import graphlab as gl
>>> data = gl.SFrame({'text': ['hate it', 'love it']})
>>> m = gl.sentiment_analysis.create(data, features=['text'])
>>> m.predict(data)
dtype: float
Rows: 2
[0.5565335646003744, 0.7532539958283951]

You can refer to the user guide above for more detailed usage information.