When is it biased?: assessing the representativeness of twitter's streaming API

By Fred Morstatter, Jürgen Pfeffer, Huan Liu. 2014

130 citations recorded by the source corpus.

Twitter shares a free 1% sample of its tweets through the "Streaming API". Recently, research has pointed to evidence of bias in this source. The methodologies proposed in previous work rely on the restrictive and expensive Firehose to find the bias in the Streaming API data. We tackle the problem of finding sample bias without costly and restrictive Firehose data. We propose a solution that focuses on using an open data source to find bias in the Streaming API.

Subfields: Computer Science Applications, Information Systems, Statistical and Nonlinear Physics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Science Applications

In Information Systems

In Statistical and Nonlinear Physics

Citations in the corpus

Social Data: Biases, Methodological Pitfalls, and Ethical Boundaries (2019, background): “Further, in the case of Twitter, much research relies on APIs that give access to at most 1% of the data (González-Bailón et al., 2014a; Joseph et al., 2014; Morstatter et al., 2014).”
Demographics of News Sharing in the U.S. Twittersphere (2017, background): “Although the 1% random sample is not the best data to capture all the dynamics happening in Twitter, its limitations are known [28] and it is the best available option at our disposal.”
Population Bias in Geotagged Tweets (2015, background): “As Morstatter et al. (2013) report from the Twitter Firehose, about 1.4% of tweets are geotagged; and elsewhere (Morstatter, Pfeffer, and Liu 2014) they report the Streaming API is more likely to be biased when the response to a query exceeds 1% of the total volume of tweets.”
Tampering with Twitter’s Sample API (2018, methodology): “It was suggested in the past that Sample API data can be used to estimate the quality of Streaming API data [51].”
What to Expect When the Unexpected Happens: Social Media Communications Across Crises (2015, methodology): “The quality of Twitter data samples acquired via the publicly available APIs that offer limited access to the full Twitter stream has been studied extensively, to understand the nature of the biases of such data samples [18, 19, 25, 31, 32].”
Identifying Platform Effects in Social Media Data (2016, background): “…data (Chang et al. 2010; Mislove et al. 2011; Sloan et al. 2013; Hecht and Stephens 2014; Longley, Adnan, and Lansley 2015; Malik et al. 2015) and to identify biases in public APIs (Morstatter et al. 2013; Morstatter, Pfeffer, and Liu 2014), but platform effects remain empirically unaddressed.”
Surpassing the Limit: Keyword Clustering to Improve Twitter Sample Coverage (2015, background): “These differences indicate that the streaming API introduces some form of bias when determining how to limit the results sent to a stream [18].”
Can One Tamper with the Sample API?: Toward Neutralizing Bias from Spam and Bot Content (2016, result): “By empirically studying the data, one study [3] found that the results were not statistically signiﬁcantly diﬀerent than the results of the Firehose.”
Twitter data: What do they represent? (2014, background): “In this context, then, and given the uneven distribution of participation in specific social media platforms as well as the comparative novelty and variety of the APIs which provide access to data on how participants are utilising these platforms [23, 24], it becomes crucially important to reflect o”
A Measurement Study of Hate Speech in Social Media (2017, background): “Moreover, our work may suffer from the biases that any work that rely on gathering online social media data currently suffers [21].”