binarypoet
December 17, 2017, 6:04pm
1
It is probably fairly well known, but new to me, so sharing. Going through my twitter settings, I stumbled across the option to pull your personal tweet archive. They put together an archive with a nice HTML interface to browse them. More interesting is that each day’s worth of tweets is an very parseable JS file (2008_05.js) with entries like this:
{
"source": "<a href=\"http://twitter.com\" rel=\"nofollow\">Twitter Web Client</a>",
"entities": {
"user_mentions": [],
"media": [],
"hashtags": [],
"urls": []
},
"geo": {},
"id_str": "820985823",
"text": "break from ontology class...nothing new so far :(",
"id": 820985823,
"created_at": "2008-05-27 00:00:00 +0000",
"user": {
"name": "James Birchfield",
"screen_name": "birchsport",
"protected": false,
"id_str": "14919086",
"profile_image_url_https": "https://pbs.twimg.com/profile_images/942449706436911105/5-DiYQxG_normal.jpg",
"id": 14919086,
"verified": false
}
}
2 Likes
jeremy
December 17, 2017, 6:32pm
2
3 Likes
binarypoet
December 17, 2017, 8:31pm
3
TY @jeremy for sharing the link…that would have been useful for me to do in the first place.
groverpr
December 18, 2017, 2:51am
4
Another guide to extract tweets and do sentimental analysis on those. We can use tweepy wrapper around twitter API.
# Twitter sentiment analysis
The goal of this project is to learn how to pull twitter data, using the [tweepy](http://www.tweepy.org/) wrapper around the twitter API, and how to perform simple sentiment analysis using the [vaderSentiment](https://github.com/cjhutto/vaderSentiment) library. The tweepy library hides all of the complexity necessary to handshake with Twitter's server for a secure connection.
As you did in the recommendation engine project, you will also produce a web server running at AWS to display the most recent 100 tweets from a given user and the list of users followed by a given user. For example, in response to URL `/the_antlr_guy` (`http://localhost/the_antlr_guy` when tested on your laptop), your web server should respond with a tweet list color-coded by sentiment, using a red to green gradient:
<img src=figures/parrt-tweets.png width=800>
As another example URL `/realdonaldtrump` yields:
<img src=figures/trump-tweets.png width=750>
Next you will create a page responding to URLs, such as `/following/the_antlr_guy`, that displays the list of users followed by a given user:
<img src=figures/parrt-follows.png width=320>
Or:
<img src=figures/trump-follows.png width=350>
show original
1 Like
