Soon-Gyo Jung

Persona Generation from Aggregated Social Media Data

We develop a methodology for persona generation using real time social media data for the distribution of products via online platforms. From a large social media account containing more than 30 million interactions from users from 181 countries engaging with more than 4,200 digital products produced by a global media corporation, we demonstrate that our methodology can first identify both distinct and impactful user segments and then create persona descriptions by automatically adding pertinent features, such as names, photos, and personal attributes. We validate our approach by implementing the methodology into an actual working system that leverages large scale online user data for generation of persona descriptions. We present the overall methodological approach, data analysis process, and system development. Findings show this method can develop believable personas representing real groups of people using real-time online user data. Results have implications for those distributing products via online platforms.

Findings of a User Study of Automatically Generated Personas

We report findings and implications from a semi-naturalistic user study of a system for Automatic Persona Generation (APG) using large-scale audience data of an organizations social media channels conducted at the workplace of a major international corporation. Thirteen participants from a range of positions within the company engaged with the system in a use case scenario. We employed a variety of data collection methods, including mouse tracking and survey data, analyzing the data with a mixed method approach. Results show that having an interactive system may aid in keeping personas at the forefront while making customer-centric decisions and indicate that data-driven personas fulfill information needs of decision makers by mixing personas and numerical data. The findings have implications for the design of persona systems and the use of online analytics data to better understand users and customers.

Viewed by too many or viewed too little: Using information dissemination for audience segmentation: Viewed by Too Many or Viewed Too Little: Using Information Dissemination for Audience Segmentation

The identification of meaningful audience segments, such as groups of users, consumers, readers, audience, etc., has important applicability in a variety of domains, including for content publishing. In this research, we seek to develop a technique for determining both information dissemination and information discrimination of online content in order to isolate audience segments. The benefits of the technique include identification of the most impactful content for analysis. With 4,320 online videos from a major news organization, a set of audience attributes, and more than 58 million interactions from hundreds of thousands of users, we isolate the key pieces of content in terms of identifying audience segments that are both (a) least and most discriminating in terms of audience segments and (b) the least and most impactful. By empirical methods, we show that 25.3 percent of the videos are so widely disseminated (i.e., viewed by so many different segments) that they are non‐discriminatory, while 29.7 percent of the videos are very discriminatory (i.e., can clearly identify one or more audience segments) but their impact is marginal, as the user base is small. Implications are that there are critical values that can be identified to isolate the set of both distinct and impactful content in a given data set of online content. We demonstrate the utility of this line of analysis by using the approach to identify critical cut‐off values for dynamic persona generation.

What We Read, What We Search: Media Attention and Public Attention among 193 Countries

We investigate the alignment of international attention of news media organizations within 193 countries with the expressed international interests of the public within those same countries from March 7, 2016 to April 14, 2017. We collect fourteen months of longitudinal data of online news from Unfiltered News and web search volume data from Google Trends and build a multiplex network of media attention and public attention in order to study its structural and dynamic properties. Structurally, the media attention and the public attention are both similar and different depending on the resolution of the analysis. For example, we find that 63.2% of the country-specific media and the public pay attention to different countries, but local attention flow patterns, which are measured by network motifs, are very similar. We also show that there are strong regional similarities with both media and public attention that is only disrupted by significantly major worldwide incidents (e.g., Brexit). Using Granger causality, we show that there are a substantial number of countries where media attention and public attention are dissimilar by topical interest. Our findings show that the media and public attention toward specific countries are often at odds, indicating that the public within these countries may be ignoring their country-specific news outlets and seeking other online sources to address their media needs and desires.

Persona Perception Scale: Developing and Validating an Instrument for Human-Like Representations of Data

Personas are widely used in software development, system design, and HCI studies. Yet, their evaluation is difficult, and there are no recognized and validated measurement scales to date. To improve this condition, this research develops a persona perception scale based on reviewing relevant literature. We validate the scale through a pilot study with 19 participants, each evaluating three personas (57 evaluations in total). This is the first reported effort to systematically develop and validate an instrument for persona perception measurement. We find the constructs and items of the scale perform well, with factor loadings ranging between 0.60 and 0.95. Reliability, measured as Cronbachs Alpha, is also satisfactory, encouraging us to pursue the use of the scale with a larger sample in future work.

Automatic Persona Generation (APG): A Rationale and Demonstration

We present Automatic Persona Generation (APG), a methodology and system for quantitative persona generation using large amounts of online social media data. The system is operational, beta deployed with several client organizations in multiple industry verticals and ranging from small-to-medium sized enterprises to large multi-national corporations. Using a robust web framework and stable back-end database, APG is currently processing tens of millions of user interactions with thousands of online digital products on multiple social media platforms, such as Facebook and YouTube. APG identifies both distinct and impactful user segments and then creates persona descriptions by automatically adding pertinent features, such as names, photos, and personal attributes. We present the overall methodological approach, architecture development, and main system features. APG has a potential value for organizations distributing content via online platforms and is unique in its approach to persona generation. APG can be found online at https://persona.qcri.org.

Fixation and Confusion: Investigating Eye-tracking Participants' Exposure to Information in Personas

To more effectively convey relevant information to end users of persona profiles, we conducted a user study consisting of 29 participants engaging with three persona layout treatments. We were interested in confusion engendered by the treatments on the participants, and conducted a within-subjects study in the actual work environment, using eye-tracking and talk-aloud data collection. We coded the verbal data into classes of informativeness and confusion and correlated it with fixations and durations on the Areas of Interests recorded by the eye-tracking device. We used various analysis techniques, including Mann-Whitney, regression, and Levenshtein distance, to investigate how confused users differed from non-confused users, what information of the personas caused confusion, and what were the predictors of confusion of end users of personas. We consolidate our various findings into a confusion ratio measure, which highlights in a succinct manner the most confusing elements of the personas. Findings show that inconsistencies among the informational elements of the persona generate the most confusion, especially with the elements of images and social media quotes. The research has implications for the design of personas and related information products, such as user profiling and customer segmentation.

Customer segmentation using online platforms: isolating behavioral and demographic segments for persona creation via aggregated user data

We propose a novel approach for isolating customer segments using online customer data for products that are distributed via online social media platforms. We use non-negative matrix factorization to first identify behavioral customer segments and then to identify demographic customer segments. We employ a methodology for linking the two segments to present integrated and holistic customer segments, also known as personas. Behavioral segments are generated from customer interactions with online content. Demographic segments are generated using the gender, age, and location of these customers. In addition to evaluating our approach, we demonstrate its practicality via a system leveraging these customer segments to automatically generate personas, which are fictional but accurate representations of each integrated behavioral and demographic segment. Results show that this approach can accurately identify both behavioral and demographical customer segments using actual online customer data from which we can generate personas representing real groups of people.

Combining Behaviors and Demographics to Segment Online Audiences: Experiments with a YouTube Channel

Social media channels with audiences in the millions are increasingly common. Efforts at segmenting audiences for populations of these sizes can result in hundreds of audience segments, as the compositions of the overall audiences tend to be complex. Although understanding audience segments is important for strategic planning, tactical decision making, and content creation, it is unrealistic for human decision makers to effectively utilize hundreds of audience segments in these tasks. In this research, we present efforts at simplifying the segmentation of audience populations to increase their practical utility. Using millions of interactions with hundreds of thousands of viewers with an organization’s online content collection, we first isolate the maximum number of audience segments, based on behavioral profiling, and then demonstrate a computational approach of using non-negative matrix factorization to reduce this number to 42 segments that are both impactful and representative segments of the overall population. Initial results are promising, and we present avenues for future research leveraging our approach.

Neural Network Hate Deletion: Developing a Machine Learning Model to Eliminate Hate from Online Comments

We propose a method for modifying hateful online comments to non-hateful comments without losing the understandability and original meaning of the comments. To accomplish this, we retrieve and classify 301,153 hateful and 1,041,490 non-hateful comments from Facebook and YouTube channels of a large international media organization that is a target of considerable online hate. We supplement this dataset by 10,000 Reddit comments manually labeled for hatefulness. Using these two datasets, we train a neural network to distinguish linguistic patterns. The model we develop, Neural Network Hate Deletion (NNHD), computes how hateful the sentences of a social media comment are and if they are above a given threshold, it deletes them using a language dependency tree. We evaluate the results by comparing crowd workers’ perceptions of hatefulness and understandability before and after transformation and find that our method reduces hatefulness without resulting in a significant loss of understandability. In some cases, removing hateful elements improves understandability by reducing the linguistic complexity of the comment. In addition, we find that NNHD can satisfactorily retain the original meaning on average but is not perfect in this regard. In terms of practical implications, NNHD could be used in social media platforms to suggest more neutral use of language to agitated online users.