intelligence on
intelligent assistants
Methodology
For this project, Python Twitter Scraping, Tableau, and Voyant tool were used.
Python Twitter scraping was based off the chapter one in Mining the Social Web, but additional lines of code were included to set the time frame of data, show full tweet content, and exclude tweets that include certain words.
Tableau was used for data visualization to compare the compound values of the negative, positive, and neutral sentiments among the four different devices.
Most frequently used words were identified through Voyant tool and the content of each individual tweet was carefully assessed. Two most frequently repeated words for each device were closely identified, filtered data to display only the tweets that include those two words, and the tweets were individually analyzed through close reading. Promotional tweets by companies or sponsors were excluded; tweets that reflect users’ opinions and suggestions were included for close-reading.
Work Plan
The data was scraped using Python and visualized with Tableau and Voyant Tools. The data was first scrapped over the course of two weeks through the use of Python, which was written to only retrieve English tweets, meaning that the content of the tweet was only English, not the user set language was English. The code that was written to retrieve the tweets that included references to the device name, as in the case of Alexa and Siri, as opposed to the potential reference of a human with the same name.
Because scraping with Python retrieved relatively relevant results, due to the strategically added conditional statements, cleaning the data was simpler yet also more complicated. Since the data was already relatively well-refined, using data-cleaning tools to clean the data, or in our specific case, select data that referred to specific themes of our these we wanted to explore, such as privacy, yielded insufficient results when we tried to use tools similar to OpenRefine. This issue is rooted in our inability to retrieve tweets from more than about two weeks, and the general public’s frequency of tweeting activity surrounding these topics.