Matches in ScholarlyData for { ?s ?p Online social media such as wikis, blogs or message boardsenable large groups of users to generate and socialize around content.With increasing adoption of such media, the number of users who interactwith user-generated content and perform certain activities on it (suchas authoring content, replying to it or liking it) grows. As a result,the amount of pragmatic metadata - i.e. data about the usage of data- is growing as well. This paper sets out to explore if and how suchmetadata has an eect on the quality of semantic user models that canbe learned from user generated content. To this end, we use data from anonline message board (the Irish boards.ie), incorporate dierent types ofpragmatic metadata into topic modeling algorithms and evaluate themby comparing their predictive performance. Our evaluation is based onthe assumption that "better" user models will be able to predict futurecontent of users more accurately and will need less time and trainingdata. Our results suggest that not all types of pragmatic metadata areequally useful for acquiring accurate semantic user models, and sometypes of pragmatic metadata can even have detrimental eects. One ofour key observations is that semantic models of users which are based onthe content to which they replied (rather than the content which theyauthored) perform better than others. To the best of our knowledge, thisis the rst work that demonstrates an eect between pragmatic metadataon one hand, and the quality of semantic user models based on user-generated content on the other. Our results are relevant for scientistsinterested in topic modeling algorithms and pragmatic metadata, andfor engineers interested in semantic analysis of users and textual contentin social media.. }
Showing items 1 to 1 of
1
with 100 items per page.
- 2 abstract "Online social media such as wikis, blogs or message boardsenable large groups of users to generate and socialize around content.With increasing adoption of such media, the number of users who interactwith user-generated content and perform certain activities on it (suchas authoring content, replying to it or liking it) grows. As a result,the amount of pragmatic metadata - i.e. data about the usage of data- is growing as well. This paper sets out to explore if and how suchmetadata has an eect on the quality of semantic user models that canbe learned from user generated content. To this end, we use data from anonline message board (the Irish boards.ie), incorporate dierent types ofpragmatic metadata into topic modeling algorithms and evaluate themby comparing their predictive performance. Our evaluation is based onthe assumption that "better" user models will be able to predict futurecontent of users more accurately and will need less time and trainingdata. Our results suggest that not all types of pragmatic metadata areequally useful for acquiring accurate semantic user models, and sometypes of pragmatic metadata can even have detrimental eects. One ofour key observations is that semantic models of users which are based onthe content to which they replied (rather than the content which theyauthored) perform better than others. To the best of our knowledge, thisis the rst work that demonstrates an eect between pragmatic metadataon one hand, and the quality of semantic user models based on user-generated content on the other. Our results are relevant for scientistsinterested in topic modeling algorithms and pragmatic metadata, andfor engineers interested in semantic analysis of users and textual contentin social media.".