Know your customer from Twitter contacts: automatic discrimination of peers contacts from news sources

Know your customer is a core element of any customer relationship management system for mass service organizations. The emergence of social networking services has provided a radically new dimension, creating a more personalized, deeper, ubiquitous and almost real time relation with customers. At the same time, some of the more widespread social network platforms seem to be evolving not only as social networks between individuals but also as mass information distribution media. When knowing your customer through social networking services, it may be of interest to disambiguate which part of the customer context in the network relates to his peers from other sources. In this paper we present an algorithmic approach to disambiguate one aspect of such relation, as expressed in the nature of the contacts established in the social network: with peers or with organizations, news media or influencers. We focus in the case of Twitter where a simple supervised linear regression can provide a ranking score, effectively discriminating and ordering by closeness peer and other types of contacts (mass media or influencers). Such discrimination can serve as a preliminary step for deeper analysis or privacy protection of customer interaction and is suitable for implementation in automated Big Data systems.

Big Data Matching Using the Identity Correlation Approach

The Identity Correlation Approach (ICA) is a statistical technique developed for matching big data where a unique identifier does not exist. This technique was developed to match the Irish Census 2011 dataset to Central Government Administrative Datasets in order to attach a unique identifier to each individual person in the Census dataset (McCormack & Smyth, 20151). The unique identifier attached is the PPS No. (Personal Public Service No.2). By attaching the PPS No. to the Census dataset, each individual can be linked to datasets held centrally by Public Sector Organisations. This expands the range of variables for statistical analysis at individual level. Statistical techniques developed here were undertaken for a major European Structure of Earnings Survey (SES) compiled by the CSO using administrative data only, and thus eliminating the need for an expensive business survey to be conducted (NES, 20073,4,5). A description of how the Identity Correlation Approach was developed is given in this paper. Data matching results and conclusions are presented here in relation to the Structure of Earnings Survey (SES)6 results for 2011.

A bibliometric overview of the Journal of Business Research

The Journal of Business Research is a leading international journal in business research dating back to 1973. This study analyzes all the publications in the journal since its creation by using a bibliometric approach. The objective is to provide a complete overview of the main factors that affect the journal. This analysis includes key issues such as the publication and citation structure of the journal, the most cited articles, and the leading authors, institutions and countries in the journal. Unsurprisingly, the USA is the leading region in the journal although a considerable dispersion exists, especially during the last years when European and Asian universities are taking a more significant position.

Public opinion mining on Sochi-2014 Olympics

The requirements of evidence-based policymaking promote interest to realtime monitoring of public’s opinions on policy-relevant topics, and social media data mining allows diversification of information portfolio used by public administrators. This study discusses issues in public opinion mining with respect to extraction and analysis of information posted on Twitter about Sochi-2014 Olympic. It focuses on topics discussed on Twitter and sentiment analysis of tweets about the Games. Final database contained 613,333 tweets covering time span from November 1, 2013 until March 31, 2014. Using hash tags the data were classified into the following categories: Events (21%); News (14%); Sports (12%); Anticipation of the Games (12%); Cheering of the teams (6%) and Problems & Politics (2%). Research reveals considerable differences in the outcomes of machine sentiment classifiers: Deeply Moving, Pattern, and SentiStrength. SentiStrength produced the most suitable results in terms of minimization of incorrectly classified tweets. Methodological implications and directions for future research are discussed.

Identification of Influencers in eWord-of-Mouth communities using their Online Participation Features

The identification of influencers in any type of online social network is of paramount importance, as they can significantly affect consumers’ purchasing decisions. This paper proposes the utilization of a self-designed web scraper to extract meaningful information for the identification of influencers and the analysis of how this new set of variables can be used to predict them. The experimental results from the Ciao UK website will be used to illustrate the proposed approach and to provide new insights in the identification of influencers. Obtained results show the importance of the trust network, but considering the intensity and the quality of both trustors and trustees.

Is content king? Job seekers’ engagement with social media employer branding content.

Increasing digitization and the emergence of social media have radically changed the recruitment landscape adding interactive digital platforms to traditional means of employer communication. Removing barriers of distance and timing, social media enable firms to continue their efforts of promoting their employment brand online. However, social media employer communication and employer brand building remains woefully understudied. Our study addresses this gap by investigating how firms use social media to promote their employer brand. We analyze employer branding communication in a sample of N = 216,828 human resources (HR) related Tweets from N = 166 Fortune 500 companies. Using supervised machine learning we classify the Tweet content according to its informational and inspirational nature, identifying five categories of employer branding social media communication on Twitter.

Quantifying and comparing web news portals’ article salience using the VoxPopuli tool

VoxPopuli tool enables quantification of absolute and relative salience of news articles published on daily news web portals. Obtained numerical values for the two types of salience enable direct comparison of audience impact of different news articles in specified time period. Absolute salience of a news article in a specified time period is determined as the total number of distinct readers who commented on the story in that period. Hence, articlesthat appear on web portals with larger audiences will in general be (absolutely) more salient as there are more potential commentators to comment on them. On the other hand, relative salience of a particular article during a particular time period is calculated as the quotient of a number of distinct readers who comented on that particular story and the number of all readers who in the same period commented on any news story published on the same news portal. As such relative salience will always be a number between 0 and 1, irrespective of the popularity of particular news portal, the (relative) salience of news stories on different news portals can be compared.