Visual sentiment analysis on twitter data streams

Abstract

Twitter currently receives about 190 million tweets (small textbased Web posts) a day, in which people share their comments regarding a wide range of topics. A large number of tweets include opinions about products and services. However, with Twitter being a relatively new phenomenon, these tweets are underutilized as a source for evaluating customer sentiment. To explore high-volume twitter data, we introduce three novel timebased visual sentiment analysis techniques: (1) topic-based sentiment analysis that extracts, maps, and measures customer opinions; (2) stream analysis that identifies interesting tweets based on their density, negativity, and influence characteristics; and (3) pixel cell-based sentiment calendars and high density geo maps that visualize large volumes of data in a single view. We applied these techniques to a variety of twitter data, (e.g., movies, amusement parks, and hotels) to show their distribution and patterns, and to identify influential opinions.

Why e-government projects fail? An analysis of the Healthcare.gov website

“…More specifically, Twitter data analysis was performed in order to demonstrate the occurrence of the previously defined failure and corresponding timeline, as well as the impact of failure to the end-users, as expressed in social media discussions. Twitter was selected for the following reasons: it is based on short text messages that can be easily analyzed; it is highly popular, and it demonstrates users and issues of high impact on public opinion (Hao et al, 2011).…”

Twitter Sentiment in Data Streams with Perceptron

“…One popular approach is to analyze tweets according to the words which they contain. This is known as the lexicon based approach [1,[3][4][5][6][7]. This method uses a dictionary of words or n-grams labeled as positive or negative to determine a weight according to the frequency of these key words or n-grams.…”

“…Because of the high volume of tweets, the rate at which they are created, and the constantly changing nature of the tweets, twitter data are best modeled as a data stream [5,10]. This presents a number of different challenges since many of the traditional batch learning methods fail when applied to a data stream since we have a number of limitations which are not faced in batch learning [11].…”

Quantity estimation in visualizations of tagged text

“…The dimensions of these vectors are often specific words or tokens in the corpus, creating a one to one link between individual words and high level statistical patterns (such as topic membership, rhetorical difference, or entity possession). Other text visualization work, especially work dealing with streaming text data, has also incorporated annotated raw text or annotated word clouds into their visualizations [7,17]. Since the analyses behind these visualizations operate at the level of individual tokens, tagging can be used to connect higher level properties to specific passages of text.…”