23 August 2013 in Social Science Computer Review
Electoral prediction from Twitter data is an appealing research topic. It seems relatively straightforward and the prevailing view is overly optimistic. This is problematic because while simple approaches are assumed to be good enough, core problems are not addressed. Thus, this article aims to (1) provide a balanced and critical review of the state of the art; (2) cast light on the presume predictive power of Twitter data; and (3) propose some considerations to push forward the field. Hence, a scheme to characterize Twitter prediction methods is proposed. It covers every aspect from data collection to performance evaluation, through data processing and vote inference. Using that scheme, prior research is analyzed and organized to explain the main approaches taken up to date but also their weaknesses. This is the first meta-analysis of the whole body of research regarding electoral prediction from Twitter data. It reveals that its presumed predictive power regarding electoral prediction has been somewhat exaggerated: Social media may provide a glimpse on electoral outcomes but, up to now, research has not provided strong evidence to support it can currently replace traditional polls. Nevertheless, there are some reasons for optimism and, hence, further work on this topic is required, along with tighter integration with traditional electoral forecasting research.
14 October 2013 in Internet Research
PMPanagiotis Takis Metaxas
Purpose – Social media provide an impressive amount of data about users and their interactions, thereby offering computer and social scientists, economists, and statisticians – among others – new opportunities for research. Arguably, one of the most interesting lines of work is that of predicting future events and developments from social media data. However, current work is fragmented and lacks of widely accepted evaluation approaches. Moreover, since the first techniques emerged rather recently, little is known about their overall potential, limitations and general applicability to different domains. Therefore, better understanding the predictive power and limitations of social media is of utmost importance. Design/methodology/approach – Different types of forecasting models and their adaptation to the special circumstances of social media are analyzed and the most representative research conducted up to date is surveyed. Presentations of current research on techniques, methods, and empirical studies ai...
16 November 2012 in IEEE Internet Computing
Predicting X from Twitter is a popular fad within the Twitter research subculture. It seems both appealing and relatively easy. Among such studies, electoral prediction is maybe the most attractive, and a growing body of literature exists on this topic. This research problem isn't only interesting, but is also extremely difficult. However, most authors seem to be more interested in claiming positive results than in providing sound and reproducible methods. It's also especially worrisome that recent papers seem to only acknowledge those studies supporting the idea that Twitter can predict elections. This is all problematic because while simple approaches are purported to be good enough, the predictive power of Twitter regarding elections has been greatly exaggerated, and difficult research problems still lie ahead.
1 October 2011 in Communications of the ACM
The power to predict outcomes based on Twitter data is greatly exaggerated, especially for political elections.
30 May 2009 in Information Sciences
Search engine logs provide a highly detailed insight of users’ interactions. Hence, they are both extremely useful and sensitive. The datasets publicly available to scholars are, unfortunately, too few, too dated and too small. There are few because search engine companies are reluctant to release such data; they are dated because they were collected in late 1990s or early 2000s; and they are small because they comprise data for at most one day and just a few hundreds of thousands of users. Even worse, the large query log disclosed by AOL in 2006 caused more harm than good because of a big privacy flaw. In this paper the author provides an overall view of the possible applications of query logs, the privacy concerns researchers must face when working on such datasets, and several ways in which query logs can be easily sanitized. One of such measures consists of segmenting the logs into short topical sessions. Therefore, the author offers a comprehensive survey of session detection methods, as well as a thorough description of a new evaluation framework with performance results for each of the different methods. Additionally, a new, simple, but outperforming session detection method is proposed. It is a heuristic-based technique which works on the basis of a geometric interpretation of both the time gap between queries and the similarity between them in order to flag a topic shift.