See all upcoming seminars in LäsIT and seminar web pages at the homepage for the PhD studentseminars, TDB, Vi2, Theory and Applications Seminars (TAS) @ UpMARC., Department of Mathematics and The Stockholm Logic Seminar.

Tomorrow (21 Mar)
Dave Zachariah: Reliable Semi-Supervised Learning when Labels are Missing at Random
Location: ITC 2215, Time: 13:15

Semi-supervised learning methods are motivated by the availability of large datasets with unlabeled features in addition to labeled data. Unlabeled data is, however, not guaranteed to improve classification performance and has in fact been reported to impair the performance in certain cases. In this talk we discuss some fundamental limitations to semi-supervised learning and restrictive assumptions which result in unreliable classifiers. We also propose a learning approach that relaxes such assumptions and is capable of providing classifiers that reliably quantify the label uncertainty.

Licentiatseminarium | Licentia
Friday 22 Mar
Diane Golay: An Invisible Burden: An Experience-Based Approach to Nurses' Daily Work Life with Healthcare Information Technology
Location: Fakultetsrummet, Ångström, Time: 13:15

Information and Communication Technology (ICT) has been an increasingly pervasive component of most workplaces throughout the past half century. In healthcare, the turn to the digital has resulted into the broad implementation of Healthcare Information Technology (HIT). The impacts of ICT on work life have been investigated predominantly through surveys, although some researchers have advocated for the use of a qualitative, experience-based approach. Meanwhile, the existing body of research on the impacts of HIT on clinicians has painted a mixed picture of digitalisation. Despite some clear benefits, HIT has indeed been found to have unexpected, unintended adverse consequences for hospital staff. Typical issues include loss in efficiency, extra effort to carry out routine tasks, and the creation of new, HIT-induced work activities. Simultaneously, research outside of the healthcare domain has shown that ICT could require extra effort from some users in order for the socio-technical system to function properly - extra work often invisible to developers. Based on observation, interview and focus group data collected at a large Swedish hospital, this thesis set out to investigate the impact of HIT on hospital nurses from an experience-based perspective, resulting in four main contributions. First, a method supporting experience-based data analysis, the HolisticUX method, is introduced. Second, 13 forms of HIT-induced additional tasks in nurses' workload are identified, five of which are not acknowledged in previous research. Third, task avoidance is identified as a consequence of nurses' increased workload, negatively affecting patient safety, care quality and nurses' professional satisfaction. Finally, four factors are argued to contribute to a suggested invisibility of the HIT-induced time burden in nurses' work life to management and developers: 1) lack of a holistic perspective, 2) the hidden cost of a single click, 3) the invisibility of nursing work, and 4) visible data, invisible work.

Monday 25 Mar
Moses Boudourides, Robert K. Merton Visiting Research Fellow, Institute for Analytical Sociology, Linköping University: Text Network Analysis in Twitter: Multilayer Networks of Co-Occurrent Hashtags, Mentioning Tweeple and Topic Modeling Terms
Location: Mötesrum 19204, Time: 11:00-12:00

Nowadays, Twitter data can be easily mined or even be downloaded from publicly available datasets, which have been already retrieved from the Twitter and being distributed in existing open repositories of datasets. Typically, the network analysis of Twitter data proceeds towards the construction of (at least) two important networks: (1) graphs of hashtags co-occurring (in tweets) and (2) graphs of mentions or, better said, graphs among mentioning/mentioned tweeple, i.e., among mentioning-tweeple, who are senders of tweets, and mentioned-tweeple, who are simply mentioned in the contents of tweets sent by the former; notice that a mentioned-tweeple is not necessarily a sender-tweeple, although for the graph of mentions to be a nontrivial directed graph it is necessary that some of the mentioned-tweeple are themselves mentioning-tweeple. Furthermore, there exist various techniques of text analysis, which aim towards the extraction and classification of keywords from a corpus of textual data. Here, we are going to apply it in the corpus of the tweet contents of the analyzed Twitter dataset in order to be able to extract certain salient words from this corpus. Apparently, these salient words are called “TM-terms” or simply “terms,” since each one of them belongs to at least one of the “topics” that TM yields. Thus, given a Twitter dataset, we are going to extract from it three constituting a multi-digraph, i.e., a weighted directed network. After constructing the three-layer Twitter network of hashtags, tweeple and TM-terms, our aim is to examine how an existing partition in the subgraph of one layer may induce a corresponding grouping in the subgraphs of the remaining layers. There are three cases to be considered: 1. The community partition of hashtags in the layer of co-occurrent hashtags: Then each community of hashtags may induce two groupings: a grouping of tweeple consisting of those tweeple who co-occur with (at least) one hashtag of the given community (of hashtags) and a grouping of TM-terms each one of which similarly co-occurs with hashtags of that community. Therefore, the community partition of hashtags induces a grouping of the set of tweeple and a grouping of the set of terms. Apparently, these groupings do not constitute partitions of the sets of tweeple/terms (respectively), since they might be overlapping, due to the fact that the co-occurrence relation is not necessarily one-to-one (evidently, in general, it is one-to-many). 2. The decomposition of the layer of terms into topics (this is certainly not a partition because topics are overlapping): Again each topic may induce two groupings: a grouping of tweeple consisting of those tweeple who co-occur (inside tweet contents) with (at least) one term of the given topic and a grouping of hashtags each one of which similarly co-occurs with terms in that topic. Thus, the structure of topics (obtained by TM) induces a corresponding decomposition of tweeple and hashtags (respectively). 3. Similarly, the label propagation algorithm for community detection in the layer of mentioning tweeple may induce two corresponding decompositions on hashtags and terms (respectively). In this seminar, I am going to present a framework of computations (implemented in Python) which uphold the above described construction of a 3-layer Twitter network and the generation of induced decompositions in each layer from existing partitions or decompositions in another layer.

See also the list of all upcoming seminars.

Internal seminars. Lecturers may be either internal or external.