Tweet
Here are the steps I followed to obtain my results
1) Obtaining Data - Signed up as Developer in "http://www.regulations.gov/" and using an API Key, I downloaded JSON Data through an app. We can download upto 1000 comments with one REST API Call. Doing it recursively by changing request parameters, we can obtain all the other comments. These comments can be stored in a database for future use.
2) Sentiment Analysis - A comment's sentiment can be found out by using software. Expected results from such analysis are generally positive, negative, neutral and unclassified. My original plan was to use SAS Sentiment Analysis Studio to find out the sentiment. But, having no access to the software made me search for other ways to do it.
3) Free Sentiment Analysis API - My next stop was API, I made an application which downloads the comments from regulations.gov website, and then using a REST API call, find out about the sentiment of each comment from its reponse. Sentiment Analysis API. But, there are some comments, which the API was not able to categorize. I also used a dictionary to filter positive and negative comments based on words such as "strongly support", "support", "oppose", "strongly oppose".
Based on the above process, here are my results
Positive Comments = 12267
Negative Comments = 4856
Neutral = 93
Unclassified/Errors = 4968
P.S - I am still trying to get better results and also confidence intervals for the above results. Help me out, if you know any other way to get better results.
Thank you,
Cheers
Dheeraj Jami
1) Obtaining Data - Signed up as Developer in "http://www.regulations.gov/" and using an API Key, I downloaded JSON Data through an app. We can download upto 1000 comments with one REST API Call. Doing it recursively by changing request parameters, we can obtain all the other comments. These comments can be stored in a database for future use.
2) Sentiment Analysis - A comment's sentiment can be found out by using software. Expected results from such analysis are generally positive, negative, neutral and unclassified. My original plan was to use SAS Sentiment Analysis Studio to find out the sentiment. But, having no access to the software made me search for other ways to do it.
3) Free Sentiment Analysis API - My next stop was API, I made an application which downloads the comments from regulations.gov website, and then using a REST API call, find out about the sentiment of each comment from its reponse. Sentiment Analysis API. But, there are some comments, which the API was not able to categorize. I also used a dictionary to filter positive and negative comments based on words such as "strongly support", "support", "oppose", "strongly oppose".
Based on the above process, here are my results
Positive Comments = 12267
Negative Comments = 4856
Neutral = 93
Unclassified/Errors = 4968
P.S - I am still trying to get better results and also confidence intervals for the above results. Help me out, if you know any other way to get better results.
Thank you,
Cheers
Dheeraj Jami
Comment