Adrian M.P. Brasoveanu
Short Bio
Adrian joined Modul University as a Researcher in the Department of New Media Technology in November 2011, after an 8-month internship. Adrian's research interests are in the following areas: Semantic Web (Linked Data, OBDA, Reasoning), NLP & Information Extraction (NER, NEL, KBP, AKBC,resources and evaluation techniques), Information Abstraction (summarization, automated ontology construction), Information Visualization (D3, Vega and related tech) and Machine Learning. Between November 2017 and October 2018, Adrian has spent 12 months as Invited Research Fellow at FHG Graubunden in Chur, Switzerland (formerly known as HTW Chur). He has returned to MODUL since November 2018.
Projects
-
A Digital Well-Being Index for Vienna
How can stakeholders get insights into perceived well-being in a timely manner and related to specific aspects of urban environments? Comprehensive national surveys are costly, require significant organizational effort and are a careful balance between knowledge to be gained and burden placed on respondents. To address these challenges, DWBI will use AI-based natural language processing in conjunction with evolving knowledge graphs to extract affective knowledge from user-generated digital content. Automatically mapped to existing indicators, this knowledge will complement and enrich survey-based approaches and traditional assessment methodologies, mitigating some of the associated problems.
Organisations: MODUL University Vienna, Department of Sustainability, Governance, and Methods
Author: Arno Scharl, Ivo Ponocny, Sabine Sedlacek, Adrian Brasoveanu, Agnieszka Skwara
Date: 01.10.2021 - 30.09.2024
Managed By: MODUL University Vienna
-
GENTIO - Generative Learning Networks for Text and Impact Optimization
GENTIO aims for radical innovation in the way we produce, enrich and analyse digital content. The project will develop a flexible Deep Learning Architecture to unify the understanding of text at three fundamental levels: structure, content and context. The first use case targets the marketing domain. It will experiment with new methods for communication experts to maximize the impact of data-driven publishing. The second use case targets the news media sector, automatically correcting and classifying noisy output from Optical Character Recognition (OCR) systems - using topics extracted from the public debate on other microblogging sites to obtain the required context information.
Organisations: MODUL University Vienna, Department of New Media Technology
Author: Lyndon Nixon, Adrian Brasoveanu, Jakob Steixner
Date: 01.01.2020 - 31.12.2022
Managed By: MODUL University Vienna
-
EPOCH - Extracting and Predicting Events from Online Communication and Hybrid Datasets
EPOCH will measure the effects on statistical indicators of events being reported in the news and social media. Innovatively, it will use the measured effects of now past events to predict the future changes expected due to future events detected in the public dialogue. Through the EPOCH dashboard, organizations can identify and thus better prepare for these changes, adapting their communications, marketing and resources accordingly. This will be demonstrated in the domains of purchase price forecasting and public relations.
Organisations: MODUL University Vienna, Department of New Media Technology, Department of Sustainability, Governance, and Methods
Author: Lyndon Nixon, Sabine Sedlacek, Ivo Ponocny, Adrian Brasoveanu, Jakob Steixner
Date: 01.01.2019 - 31.12.2021
Managed By: Modul Technology GmbH
-
ReTV - Enhancing and Repurposing TV Content for Trans-Vector Engagement
ReTV aims to provide broadcasters and content distributors with technologies and insights to leverage the converging digital media landscape. By advancing the state of the art in the analysis of this media landscape and providing novel methods to dynamically re-purpose content for an array of media vectors (= all relevant digital channels), a Trans-Vector Platform (TVP) will provide these stakeholders with the ability to "publish to all media vectors with the effort of one". It will empower broadcasters and brands to measure and predict the success of their content and advertisments in terms of reach and audience engagement across vectors.
Organisations: MODUL University Vienna, Department of New Media Technology, Modul Technology GmbH
Author: Lyndon Nixon, Adrian Brasoveanu, Jakob Steixner, Adriana Bassani, Pavel Filippov, Maximilian Lang, Rod Michael Coronel
Date: 01.01.2018 - 31.12.2020
Managed By: Modul Technology GmbH
-
ETIHQ - Exposing Tourism Indicators as High Quality Linked Data
Although the tourism domain heavily relies on complex decision making, it currently lacks decision support systems with the capability to seamlessly integrate and visualise data from multiple data sources of tourism (and other) indicators. Linked Data technologies, by contrast, especially when adopted at large scale, greatly facilitate data integration at the syntactic and semantic level alike by providing a uniform data encoding format. Such technologies also help to clearly specify the meaning of the data and to establish links between various datasets. In this project we will use Linked Data technologies to create a reference repository of tourism indicators (ETIHQ) by exposing the content of TourMIS, a major source of European tourism statistics, as high-quality Linked Data. We will ensure data quality by providing semantically rich vocabularies that will support (i) the specification of the meaning of tourism statistics and (ii) the provenance of the data items. To demonstrate the benefits of using Linked Data, we will design and implement a decision support system that makes use of the ETIHQ repository and leverages its detailed semantic specifications to provide appropriate access control mechanisms.
Organisations: Department of New Media Technology, Department of Tourism and Service Management, MODUL University Vienna
Author: Marta Sabou, Adrian Brasoveanu, Irem Önder, Arno Scharl, Karl Wöber
Date: 01.10.2013 - 01.10.2014
Managed By: Department of New Media Technology
Research Output
- 2020
-
"In Media Res: A Corpus for Evaluating Named Entity Linking with Creative Works"2020 Pages: 355-364
Annotation styles express guidelines that direct human annotators in what rules to follow when creating gold standard annotations of text corpora. These guidelines not only shape the gold standards they help create, but also influence the training and evaluation of Named Entity Linking (NEL) tools, since different annotation styles correspond to divergent views on the entities present in the same texts. Such divergence is particularly present in texts from the media domain that contain references to creative works. In this work we present a corpus of 1000 annotated documents selected from the media domain. Each document is presented with multiple gold standard annotations representing various annotation styles. This corpus is used to evaluate a series of Named Entity Linking tools in order to understand the impact of the differences in annotation styles on the reported accuracy when processing highly ambiguous entities such as names of creative works. Relaxed annotation guidelines that include overlap styles lead to better results across all tools.
Author(s): Lyndon Nixon, Adrian Brasoveanu, Albert Weichselbraun
Publication date: 19. 11. 2020
Pages: 355-364
-
"Automatic Expansion of Domain-Specific Affective Models for Web Intelligence Applications"2020 in: Cognitive Computation.
Background. Sentic computing relies on welldefined affective models of different complexity - polarity to distinguish positive and negative sentiment,
for example, or more nuanced models to capture expressions of human emotions. When used to measure
communication success, even the most granular affective model combined with sophisticated machine learning approaches may not fully capture an organisation’s
strategic positioning goals. Such goals often deviate from
the assumptions of standardised affective models. While
certain emotions such as Joy and Trust typically represent desirable brand associations, specific communication goals formulated by marketing professionals often
go beyond such standard dimensions. For instance, the
brand manager of a television show may consider fear
or sadness to be desired emotions for its audience.
Method. This article introduces expansion techniques
for affective models, combining common and commonsense knowledge available in knowledge graphs with
language models and affective reasoning, improving coverage and consistency as well as supporting domainspecific interpretations of emotions.
Results and Conclusions. An extensive evaluation
compares the performance of different expansion techniques: (i) a quantitative evaluation based on the revisited Hourglass of Emotions model to assess perfor
mance on complex models that cover multiple affective categories, using manually compiled gold standard
data, and (ii) a qualitative evaluation of a domainspecific affective model for television programme brands.
The results of these evaluations demonstrate that the
introduced techniques support a variety of embeddings
and pre-trained models. The paper concludes with a
discussion on applying this approach to other scenarios
where affective model resources are scarce.Author(s): Albert Weichselbraun, Jakob Steixner, Adrian Brasoveanu, Arno Scharl, Max Göbel, Lyndon Nixon
Publication date: 2020
Electronic version(s), related files and links: http://dx.doi.org/10.1007/s12559-021-09839-4
- 2019
-
"Introducing orbis: An extendable evaluation pipeline for named entity linking performance drill‐down analyses"2019 in: Proceedings of the Association for Information Science and Technology . Volume: 56. Issue number: 1 Pages: 468-471
Most current evaluation tools are focused solely on benchmarking and comparative evaluations thus only provide aggregated statistics such as precision, recall and F1‐measure to assess overall system performance. They do not offer comprehensive analyses up to the level of individual annotations. This paper introduces Orbis, an extendable evaluation pipeline framework developed to allow visual drill‐down analyses of individual entities, computed by annotation services, in the context of the text they appear in, in reference to the entities specified in the gold standard.
Author(s): Fabian Odoni, Adrian Brasoveanu, Philip Kuntschik, Albert Weichselbraun
Publication date: 18. 10. 2019
Volume: 56
Issue number: 1
Pages: 468-471
Electronic version(s), related files and links: http://dx.doi.org/10.1002/pra2.49
-
"Semantic fake news detection: a machine learning perspective"2019 Pages: 656-667
Fake news detection is a difficult problem due to the nuances of language. Understanding the reasoning behind certain fake items implies inferring a lot of details about the various actors involved. We believe that the solution to this problem should be a hybrid one, combining machine learning, semantics and natural language processing. We introduce a new semantic fake news detection method built around relational features like sentiment, entities or facts extracted directly from text. Our experiments show that by adding semantic features the accuracy of fake news classification improves significantly.
Author(s): Adrian Brasoveanu, Razvan Andonie
Publication date: 6. 2019
Publisher: Springer
Pages: 656-667
Electronic version(s), related files and links: http://dx.doi.org/https://doi.org/10.1007/978-3-030-20521-8_54
-
"Name variants for improving entity discovery and linking"2019 Pages: 14:1-14:15
Identifying all names that refer to a particular set of named entities is a challenging task, as quite often we need to consider many features that include a lot of variation like abbreviations, aliases, hypocorism, multilingualism or partial matches. Each entity type can also have specific rules for
name variances: people names can include titles, country and branch names are sometimes removed from organization names, while locations are often plagued by the issue of nested entities. The lack of a clear strategy for collecting, processing and computing name variants significantly lowers the
recall of tasks such as Named Entity Linking and Knowledge Base Population since name variances are frequently used in all kind of textual content.
This paper proposes several strategies to address these issues. Recall can be improved by combining knowledge repositories and by computing additional variances based on algorithmic approaches. Heuristics and machine learning methods then analyze the generated name variances and mark ambiguous names to increase precision. An extensive evaluation demonstrates the effects
of integrating these methods into a new Named Entity Linking framework and confirms that systematically considering name variances yields significant performance improvements.Author(s): Albert Weichselbraun, Philip Kuntschik, Adrian Brasoveanu
Publication date: 2019
Pages: 14:1-14:15
Load more results
