The abundance of data in the information age poses an immense challenge for us: how to perform large-scale inference to understand and utilize this overwhelming amount of information. Such techniques are of tremendous intellectual significance and practical impact. As part of this grand challenge, the goal of my Ph.D. thesis is to develop effective and efficient statistical topic models for massive text collections by incorporating extra information from other modalities in addition to the text itself. Text documents are not just text, and different kinds of additional information are naturally interleaved with text. Most previous work, however, pays attention to only one modality at a time, and ignore the others. In my thesis, I will present a series of probabilistic topic models to show how we can bridge multiple modalities of information, in a united fashion, for various tasks. Interestingly, joint inference over multiple modalities leads to many findings that can not be discovered from just one modality alone, as briefly illustrated below: Email is pervasive nowadays. Much previous work in natural language processing modeled text using latent topics ignoring the social networks. On the other hand, social network research mainly dealt with the existence of links between entities without taking into consideration the language content or topics on those links. The author-recipient-topic (ART) model, by contrast, steers the discovery of topics according to the relationships between people, and learns topic distributions based on the direction-sensitive messages sent between entities. However, the ART model does not explicitly identify groups formed by entities in the network. Previous work in social network analysis ignores the fact that different groupings arise for different topics. The group-topic (GT) model, a probabilistic generative model of entity relationships and textual attributes, simultaneously discovers groups among the entities and topics among the corresponding text. Many of the large datasets do not have static latent structures; they are instead dynamic. The topics over time (TOT) model explicitly models time as an observed continuous variable. This allows TOT to see long-range dependencies in time and also helps avoid a Markov model's risk of inappropriately dividing a topic in two when there is a brief gap in its appearance. By treating time as a continuous variable, we also avoid the difficulties of discretization. Most topic models, including all of the above, rely on the bag of words assumption. However, word order and phrases are often critical to capturing the meaning of text. The topical n -grams (TNG) model discovers topics as well as meaningful, topical phrases simultaneously. In summary, we believe that these models are clear evidence that we can better understand and utilize massive text collections when additional modalities are considered and modeled jointly with text.

Multi -omics data analysis and integration facilitates hypothesis building toward an understanding of genes and pathway responses driven by environments. Methods designed to estimate and analyze gene expression, with regard to treatments or conditions, can be leveraged to understand gene-level responses in the cell. However, genes often interact and signal within larger structures such as pathways and networks. Complex studies guided toward describing dynamic genetic pathways and networks require algorithms or methods designed for inference based on gene interactions and related topologies. Classes of algorithms and methods may be integrated into generalized workflows for comparative genomics studies, as multi -omics data can be standardized between contact points in various software applications. Further, network inference or network comparison algorithmic designs may involve interchangeable operations given the structure of their implementations. Network comparison and inference methods can also guide transfer-of-knowledge between model organisms and those with less knowledge base.

U.S. politics, media, and citizens are highly polarized, stipulating that society is divided between Democrats and Republicans (Hameleers, 2019). The U.S. has seen an increased political polarization over the past 25 years (Heltzel & Laurin, 2020; Westfall, Van Boven, Chambers, & Judd, 2015). Technological development in the campaign environment has fueled this political polarization (Hong & Kim, 2016). In such a polarized technological society, partisan news media cover political issues and events from their ideological perspective (Arceneaux, Johnson, & Murphy, 2012), which may affect the polarized citizens.The Republican Party is conservative, while the Democratic Party is liberal (Westfall, Van Boven, Chambers, & Judd, 2015). Each party has issue agendas that they prioritize during the campaign. When political campaigns post a message on social media, they not only post just one issue but several related issues. These interlinked issues have a networked effect on the partisan news media and the polarized citizens (McCombs, Shaw, & Weaver, 2014). How political campaigns interlinked different issue agendas during campaigns in a polarized environment has not been investigated. This study intends to see the similarities and dissimilarities between the Democratic and Republican Party issue networks using a network agenda setting theory during the 2020 U.S. presidential campaign and how they build and set networked issue agendas in the partisan news media and the polarized public on Facebook. The study uses a hybrid content analysis and network analysis of issue agendas presented by the Biden and Trump campaigns, partisan media (CNN and Fox News), and the Democratic Party and the Republican Party supporters on Facebook. Facebook posts are collected using Facebook’s CrowdTangle Search option from January 1, 2021, to November 3, 2020. This study uses a hybrid content analysis method which engages with both human coders and computational means to analyze big data sets (Guo et al., 2016). The data analysis involves measuring core-periphery block model, clique analysis, network visualization, and Quadratic Assignment Procedures (QAP). A social networking analysis software, UCINET, is used for measuring core-periphery block model, clique analysis, and QAP correlations(Borgatti, Everett, & Johnson, 2018). The scholarship of political campaign communication needs to reconnect to the ideological positions of political campaigns, partisan news media, and party supporters. This holistic study is significant in terms of better understanding the mechanism of networked agenda-setting activities of presidential campaigns in a polarized environment on Facebook. Methodologically, this study offers new techniques for investigating networked issue agendas of campaigns, news media, and citizens. It uses core-periphery block model and clique analysis as indicators of network agenda building and network agenda-setting influences. Social media practitioners like campaign managers can consider the political polarization, fragmented nature of social media, and polarized audience during political campaigning.

This  thesis  contributes  to  the  advancement  on  the  network  view  of  social  capital andentrepreneurship, focusing on science-based innovation by observatory social network research. The study has conducted a survey to construct a network map of and network attitudes in Zulu science park (alias name) located in Sweden. The analysis showed that there were relatively positive expectations to utilize network more effectively and  efficiently in their business activity with respect to sharing of research skills and resources, as well as social exchange in particular. Currently, the science park seems not yet to take advantage of the potential momentum of the respondents or bottom-up initiatives where tenants maintain the environment mutually under trust. I discuss the merits and challenges in such resource and knowledge sharing in the business development support and governance as a new way of unique business incubator and science-park management, focusing on networking.

Introduction: In a mixed-methods study I assessed the role of social networks as predictors and outcomes of the implementation of an intervention to promote evidence-informed decision-making (EIDM) in three public health departments in Ontario, Canada. The quantitative strand included the analysis of the role of staff’s position in networks on the adoption of EIDM, the longitudinal evolution of networks, and the association between the name generators’ position in surveys and respondents’ motivation to answer survey questions. The qualitative strand aimed to explain and contextualize the quantitative findings. Methods: A tailored intervention was implemented in the public health departments, including the mentoring of staff through the EIDM process by a knowledge broker. The staff participated in three online surveys before and after the 22-month intervention, providing the names of peers to whom they turned to seek information, whom they considered as experts, and their friends. I assessed the dynamic evolution of social networks, and the role of local opinion leaders (OL) in promoting the adoption of EIDM. I interviewed key network actors about their interpretation and experience regarding the quantitative findings. Results: Overall, there was no statistically significant impact on EIDM behavior and skill in health departments. However, the analysis of the role of OLs in behaviour change showed that non-engaged staff who were connected to highly engaged OLs, and those OLs who communicated with each other improved their EIDM behavior. Social networks became more centralized around already popular staff due to selective training of recognized experts. Highly engaged staff tended to connect to each other, and to limit their connections within organizational divisions over time. In the department where multiple activities were being implemented to support EIDM, the highly engaged staff became more popular due to department-wise presentations and informal information spread. I also found that when name generator questions are asked later in surveys then respondents are more likely to refuse, indicate they do not know anyone, or provide fewer names than when these questions are asked earlier Conclusion: Social network analysis showed the structure of information-seeking relations, the impact of opinion leaders on the EIDM behavior of their peers, and underlying social changes through implementing an EIDM intervention. These findings can inform the design and tailoring of EIDM interventions in public health organizations. / Thesis / Doctor of Philosophy (PhD) / In three public health departments in Ontario, where we offered an intervention to a group of staff on how to use more research evidence in practice, I studied how the pattern of communication among staff influenced their use of evidence, how those communications changed over time, and how the staff themselves viewed those changes. In the department that largely promoted staff engagement in the intervention, the staff who were engaged became more popular over time. In all departments, already popular staff became more popular. The staff who sought information from popular people engaged in the intervention, and those popular people who communicated with each other used more research evidence over time. Network analysis helped reveal the social structure and identify popular staff and could be used to inform similar interventions. It also showed how selecting and training a group of staff can change the way people communicate in health departments.

This research builds on the social movements theory of networks and coalition building, the theory of digital activism, and the social networks theory of organizations to study the rich case of online mobilization for the 2011 Egyptian revolution. I use the analytical tools of social network analysis to study Twitter networks of activists of the Egyptian revolution in early 2011, when solidarity characterized the movement, and late 2014, when schism spread it apart. In this, I investigate how the repertoire of online activism relates to the on-the-ground movement. The social movements theory of networks states that activists’ ideological congruence, the presence of bridge builders who bring the movement together, and the presence of previous ties among the activists are all factors of coalition building and movement solidarity. This dissertation tested the role of these factors in the Twitter networks of Egyptian activists. The results suggest that digital activism complements rather than mirrors on-the-ground activism. While all three factors influence the network, they manifest somewhat differently than research has suggested they do in offline networks. This dissertation contributes to social movements theory of coalition building through adding validity to its application to digital activism, and suggests modifications to be made while applying this theory to the repertoire of online mobilization. The research has a methodological contribution through using cutting edge techniques of social network analysis to study Twitter networks of activists. Unlike earlier studies on the Egyptian revolution, this methodological approach revealed new findings that could not have been studied through other methods of research. / Dissertation / Doctor of Philosophy (PhD)

The United States (U.S.) military withdrawal from Afghanistan in August 2021 was met with turmoil as Taliban regained control of most of the country, including Kabul. These events have affected many and were widely discussed on social media, especially in the U.S. In this work, we focus on Twitter discourse regarding these events, especially potential opinion shifts over time and the effect social media posts by established U.S. legislators might have had on online public perception. To this end, we investigate two datasets on the war in Afghanistan, consisting of Twitter posts by self-identified U.S. accounts and conversation threads initiated by U.S. politicians. We find that Twitter users' discussions revolve around the Kabul airport event, President Biden's handling of the situation, and people affected by the U.S. withdrawal. Microframe analysis indicates that discourse centers the humanitarianism underlying these occurrences and politically leans liberal, focusing on care and fairness. Lastly, network analysis shows that Republicans are far more active on Twitter compared to Democrats and there is more positive sentiment than negative in their conversations. / Master of Science / The United States (U.S.) military withdrawal from Afghanistan in August 2021 was met with turmoil as Taliban regained control of most of the country, including Kabul. These events have affected many and were widely discussed on social media, especially in the U.S. In this work, we focus on Twitter regarding these events, and study if public's opinion change over time especially by the posts of legislators. Therefore, we used two datasets about unrest in Afghanistan after the Taliban takeover. One datasets consists of of Twitter posts by self-identified U.S. accounts and the other one are the conversation threads initiated by U.S. politicians. We find that Twitter users' discussions revolve around the Kabul airport event, President Biden's handling of the situation, and people affected by the U.S. withdrawal. According to our findings based on several methods analyzing the content of the posts of Twitter users, the pressing issues are the humanitarian concerns for the people who could be the target of Taliban. Last but not least, we also studied the relationship between legislators and twitter users along with the dominant sentiment about the topic. Our analysis shows that Republicans are far more active on Twitter compare to Democrats and there is more positive sentiment than negative in their conversations.

Rising childhood obesity with its detrimental health consequences poses a challenge to the health care system. Community-based, multi-setting interventions with the participatory involvement of relevant stakeholders are emerging as promising. To gain insights into the structural and processual characteristics of stakeholder networks, conducting a network analysis (NA) is advisable. Within the program “Family+—Healthy Living Together in Families and Schools”, a network analysis was conducted in two rural model regions and one urban model region. Relevant stakeholders were identified in 2020–2021 through expert interviews and interviewed by telephone to elicit key variables such as frequency of contact and intensity of collaboration. Throughout the NA, characteristics such as density, centrality, and connectedness were analyzed and are presented graphically. Due to the differences in the number of inhabitants and the rural or urban structure of the model regions, the three networks (network#1, network#2, and network#3) included 20, 14, and 12 stakeholders, respectively. All networks had similar densities (network#1, 48%; network#2, 52%; network#3, 42%), whereas the degree centrality of network#1 (0.57) and network#3 (0.58) was one-third higher compared with network#2 (0.39). All three networks differed in the distribution of stakeholders in terms of field of expertise and structural orientation. On average, stakeholders exchanged information quarterly and were connected on an informal level. Based on the results of the NA, it appears to be useful to initialize a community health facilitator to involve relevant stakeholders from the education, sports, and health systems in projects and to strive for the goal of sustainable health promotion, regardless of the rural or urban structure of the region. Participatory involvement of relevant stakeholders can have a positive influence on the effective dissemination of information and networking with other stakeholders.

学位プログラム名: 京都大学大学院思修館 / 京都大学 / 新制・課程博士 / 博士(総合学術) / 甲第24949号 / 総総博第31号 / 新制||総総||5(附属図書館) / 京都大学大学院総合生存学館総合生存学専攻 / (主査)教授 池田 裕一, 教授 IALNAZOV Dimiter Savov, 准教授 関山 健, 安橋 正人 (奈良女子大学) / 学位規則第4条第1項該当 / Doctor of Philosophy / Kyoto University / DFAM

The aim of this thesis was to study co-authorship in the constraint programming research community. This was done by conducting social network analysis (SNA) based on published scientific papers from the proceedings of the International Conference on Principles and Practice of Constraint Programming. Bibliographic data of the scientific literature was collected for the years 2018–2022 of the annual conference. For quantitative analysis, graph metrics were computed to study the properties and structure of the overall network, and also to study the attributes and characteristics of individual authors to be able to identify central actors of the community. Furthermore, graph layout algorithms were used for visualisation of the network. The computed metrics and the graphical visualisations enabled identifying collaboration patterns and behaviours within the studied field. The results of this study show that the most central actors of the community are mainly male and dominated by white organisations and countries. The results of the study also show that the vast majority of authors of the community collaborate with others in writing papers. However, due to the low density of the network there is opportunity and room for new collaboration patterns to take place within the research community.