Publications
Explore the publications from our RecSys laboratory.
21.
Sertkan, Mete; Althammer, Sophia; Hofstätter, Sebastian; Neidhardt, Julia
Diversifying Sentiments in News Recommendation Proceedings Article
In: 2022.
@inproceedings{ 20.500.12708_175977,
title = {Diversifying Sentiments in News Recommendation},
author = {Mete Sertkan and Sophia Althammer and Sebastian Hofstätter and Julia Neidhardt},
doi = {10.34726/3903},
year = {2022},
date = {2022-01-01},
volume = {3228},
series = {CEUR Workshop Proceedings},
abstract = {Personalized news recommender systems are widely deployed to filter the information overload caused by the sheer amount of news produced daily. Recommended news articles usually have a sentiment similar to the sentiment orientation of the previously consumed news, creating a self-reinforcing cycle of sentiment chambers around people. Wu et al. introduced SentiRec – a sentiment diversity-aware neural news recommendation model to counter this lack of diversity.
In this work, we reproduce SentiRec without access to the original source code and data sample. We re-implement SentiRec from scratch and use the Microsoft MIND dataset (same source but different subset as in the original work) for our experiments. We evaluate and discuss our reproduction from different perspectives. While the original paper mainly has a user-centric perspective on sentiment diversity by comparing the recommendation list to the user’s interaction history, we also analyze the intra-list sentiment diversity of the recommendation list. Additionally, we study the effect of sentiment diversification on topical diversity. Our results suggest that SentiRec does not generalize well to other data since the compared baselines already perform well, opposing the original work’s findings. While the original SentiRec utilizes a rule-based sentiment analyzer, we also study a pre-trained neural sentiment analyzer. However, we observe no improvements in effectiveness nor in sentiment diversity. To foster reproducibility, we make our source code publicly available.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Personalized news recommender systems are widely deployed to filter the information overload caused by the sheer amount of news produced daily. Recommended news articles usually have a sentiment similar to the sentiment orientation of the previously consumed news, creating a self-reinforcing cycle of sentiment chambers around people. Wu et al. introduced SentiRec – a sentiment diversity-aware neural news recommendation model to counter this lack of diversity.
In this work, we reproduce SentiRec without access to the original source code and data sample. We re-implement SentiRec from scratch and use the Microsoft MIND dataset (same source but different subset as in the original work) for our experiments. We evaluate and discuss our reproduction from different perspectives. While the original paper mainly has a user-centric perspective on sentiment diversity by comparing the recommendation list to the user’s interaction history, we also analyze the intra-list sentiment diversity of the recommendation list. Additionally, we study the effect of sentiment diversification on topical diversity. Our results suggest that SentiRec does not generalize well to other data since the compared baselines already perform well, opposing the original work’s findings. While the original SentiRec utilizes a rule-based sentiment analyzer, we also study a pre-trained neural sentiment analyzer. However, we observe no improvements in effectiveness nor in sentiment diversity. To foster reproducibility, we make our source code publicly available.
In this work, we reproduce SentiRec without access to the original source code and data sample. We re-implement SentiRec from scratch and use the Microsoft MIND dataset (same source but different subset as in the original work) for our experiments. We evaluate and discuss our reproduction from different perspectives. While the original paper mainly has a user-centric perspective on sentiment diversity by comparing the recommendation list to the user’s interaction history, we also analyze the intra-list sentiment diversity of the recommendation list. Additionally, we study the effect of sentiment diversification on topical diversity. Our results suggest that SentiRec does not generalize well to other data since the compared baselines already perform well, opposing the original work’s findings. While the original SentiRec utilizes a rule-based sentiment analyzer, we also study a pre-trained neural sentiment analyzer. However, we observe no improvements in effectiveness nor in sentiment diversity. To foster reproducibility, we make our source code publicly available.
22.
Kolb, Thomas Elmar; Nalis, Irina; Sertkan, Mete; Neidhardt, Julia
The Role of Bias in News Recommendation in the Perception of the Covid-19 Pandemic Proceedings Article
In: Thomas, Kolb (Ed.): 2022.
@inproceedings{ 20.500.12708_150340,
title = {The Role of Bias in News Recommendation in the Perception of the Covid-19 Pandemic},
author = {Thomas Elmar Kolb and Irina Nalis and Mete Sertkan and Julia Neidhardt},
editor = {Kolb Thomas},
doi = {10.48550/ARXIV.2209.07608},
year = {2022},
date = {2022-01-01},
abstract = {News recommender systems (NRs) have been shown to shape public discourse and to enforce behaviors that have a critical, oftentimes detrimental effect on democracies. Earlier research on the impact of media bias has revealed their strong impact on opinions and preferences. Responsible NRs are supposed to have depolarizing capacities, once they go beyond accuracy measures. We performed sequence prediction by using the BERT4Rec algorithm to investigate the interplay of news of coverage and user behavior. Based on live data and training of a large data set from one news outlet "event bursts", "rally around the flag" effect and "filter bubbles" were investigated in our interdisciplinary approach between data science and psychology. Potentials for fair NRs that go beyond accuracy measures are outlined via training of the models with a large data set of articles, keywords, and user behavior. The development of the news coverage and user behavior of the COVID-19 pandemic from primarily medical to broader political content and debates was traced. Our study provides first insights for future development of responsible news recommendation that acknowledges user preferences while stimulating diversity and accountability instead of accuracy, only.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
News recommender systems (NRs) have been shown to shape public discourse and to enforce behaviors that have a critical, oftentimes detrimental effect on democracies. Earlier research on the impact of media bias has revealed their strong impact on opinions and preferences. Responsible NRs are supposed to have depolarizing capacities, once they go beyond accuracy measures. We performed sequence prediction by using the BERT4Rec algorithm to investigate the interplay of news of coverage and user behavior. Based on live data and training of a large data set from one news outlet “event bursts”, “rally around the flag” effect and “filter bubbles” were investigated in our interdisciplinary approach between data science and psychology. Potentials for fair NRs that go beyond accuracy measures are outlined via training of the models with a large data set of articles, keywords, and user behavior. The development of the news coverage and user behavior of the COVID-19 pandemic from primarily medical to broader political content and debates was traced. Our study provides first insights for future development of responsible news recommendation that acknowledges user preferences while stimulating diversity and accountability instead of accuracy, only.
23.
Neidhardt, Julia; Wörndl, Wolfgang; Kuflik, Tsvi; Goldenberg, Dmitri; Zanker, Markus
Workshop on Recommenders in Tourism (RecTour) Proceedings Article
In: Golbeck, Jennifer; Harper, F. Maxwell; Murdock, Vanessa (Ed.): pp. 678–679, Association for Computing Machinery, New York, 2022.
@inproceedings{ 20.500.12708_191201,
title = {Workshop on Recommenders in Tourism (RecTour)},
author = {Julia Neidhardt and Wolfgang Wörndl and Tsvi Kuflik and Dmitri Goldenberg and Markus Zanker},
editor = {Jennifer Golbeck and F. Maxwell Harper and Vanessa Murdock},
doi = {10.1145/3523227.3547416},
year = {2022},
date = {2022-01-01},
pages = {678–679},
publisher = {Association for Computing Machinery},
address = {New York},
abstract = {The Workshop on Recommenders in Tourism (RecTour) 2022, which is held in conjunction with the 16th ACM Conference on Recommender Systems (RecSys), addresses specific challenges for recommender systems in the tourism domain. In this overview paper, we summarize our motivations to organize the RecTour workshop and present the main topic areas of RecTour submissions. These include context-aware recommendations, group recommender systems, recommending composite items, decision making and user interaction issues, different information sources and various application scenarios.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The Workshop on Recommenders in Tourism (RecTour) 2022, which is held in conjunction with the 16th ACM Conference on Recommender Systems (RecSys), addresses specific challenges for recommender systems in the tourism domain. In this overview paper, we summarize our motivations to organize the RecTour workshop and present the main topic areas of RecTour submissions. These include context-aware recommendations, group recommender systems, recommending composite items, decision making and user interaction issues, different information sources and various application scenarios.
24.
Sertkan, Mete; Neidhardt, Julia
Exploring Expressed Emotions for Neural News Recommendation Proceedings Article
In: pp. 22–28, Association for Computing Machinery, New York, NY, United States, 2022.
@inproceedings{ 20.500.12708_153156,
title = {Exploring Expressed Emotions for Neural News Recommendation},
author = {Mete Sertkan and Julia Neidhardt},
doi = {10.1145/3511047.3536414},
year = {2022},
date = {2022-01-01},
pages = {22–28},
publisher = {Association for Computing Machinery},
address = {New York, NY, United States},
abstract = {Due to domain-specific challenges such as short item lifetimes and continuous cold-start issues, news recommender systems rely more on content-based methods to deduce reliable user models and make personalized recommendations. Research has shown that alongside the content of an item, the way it is presented to the users also plays a critical role. In this work, we focus on the effect of incorporating expressed emotions within news articles on recommendation performance. We propose a neural news recommendation model that disentangles semantic and emotional modeling of news articles and users. While we exploit the textual content for the semantic representation, we extract and combine emotions of different information levels for the emotional representation. Offline experiments on a real-world dataset show that our approach outperforms non-emotion-aware solutions significantly. Finally, we provide a future outline, where we plan to investigate a) the online performance and b) the explainability/explorability of our approach.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Due to domain-specific challenges such as short item lifetimes and continuous cold-start issues, news recommender systems rely more on content-based methods to deduce reliable user models and make personalized recommendations. Research has shown that alongside the content of an item, the way it is presented to the users also plays a critical role. In this work, we focus on the effect of incorporating expressed emotions within news articles on recommendation performance. We propose a neural news recommendation model that disentangles semantic and emotional modeling of news articles and users. While we exploit the textual content for the semantic representation, we extract and combine emotions of different information levels for the emotional representation. Offline experiments on a real-world dataset show that our approach outperforms non-emotion-aware solutions significantly. Finally, we provide a future outline, where we plan to investigate a) the online performance and b) the explainability/explorability of our approach.
25.
Neidhardt, Julia; Sertkan, Mete
Towards an Approach for Analyzing Dynamic Aspects of Bias and Beyond-Accuracy Measures Proceedings Article
In: Boratto, Ludovico; Faralli, Stefano; Marras, Mirko; stilo, (Ed.): pp. 35–42, Springer Cham, 2022.
@inproceedings{ 20.500.12708_153166,
title = {Towards an Approach for Analyzing Dynamic Aspects of Bias and Beyond-Accuracy Measures},
author = {Julia Neidhardt and Mete Sertkan},
editor = {Ludovico Boratto and Stefano Faralli and Mirko Marras and stilo},
doi = {10.1007/978-3-031-09316-6_4},
year = {2022},
date = {2022-01-01},
volume = {1610},
pages = {35–42},
publisher = {Springer Cham},
abstract = {The quality of recommender systems has traditionally only been assessed using accuracy measures. Research has shown that accuracy is only one side of the medallion and that we should also consider quality features that go beyond accuracy. Recently, also fairness-related aspects and bias have increasingly been considered as outcome dimensions in this context. While beyond-accuracy measures including diversity, novelty and serendipity and bias in recommendation have been subject to the research discourse, their interrelation and temporal and group dynamics are clearly under-explored. In this position paper, we propose an approach that groups users based on their behaviors and preferences and that addresses beyond-accuracy needs of those groups while controlling for bias. Further, we consider the analysis of long-term dynamics of different interrelated beyond-accuracy measures and bias as crucial research direction since it helps to advance the field and to address societal issues related to recommender systems and personalization.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
The quality of recommender systems has traditionally only been assessed using accuracy measures. Research has shown that accuracy is only one side of the medallion and that we should also consider quality features that go beyond accuracy. Recently, also fairness-related aspects and bias have increasingly been considered as outcome dimensions in this context. While beyond-accuracy measures including diversity, novelty and serendipity and bias in recommendation have been subject to the research discourse, their interrelation and temporal and group dynamics are clearly under-explored. In this position paper, we propose an approach that groups users based on their behaviors and preferences and that addresses beyond-accuracy needs of those groups while controlling for bias. Further, we consider the analysis of long-term dynamics of different interrelated beyond-accuracy measures and bias as crucial research direction since it helps to advance the field and to address societal issues related to recommender systems and personalization.
26.
Dietz, Linus W.; Sertkan, Mete; Myftija, Saadi; Palage, Sameera Thimbiri; Neidhardt, Julia; Wörndl, Wolfgang
A Comparative Study of Data-Driven Models for Travel Destination Characterization Journal Article
In: Frontiers in Big Data, vol. 5, 2022.
@article{ 20.500.12708_139725,
title = {A Comparative Study of Data-Driven Models for Travel Destination Characterization},
author = {Linus W. Dietz and Mete Sertkan and Saadi Myftija and Sameera Thimbiri Palage and Julia Neidhardt and Wolfgang Wörndl},
url = {https://api.elsevier.com/content/abstract/scopus_id/85128638005},
doi = {10.3389/fdata.2022.829939},
year = {2022},
date = {2022-01-01},
journal = {Frontiers in Big Data},
volume = {5},
publisher = {Frontiers Media},
abstract = {Characterizing items for content-based recommender systems is a challenging task in complex domains such as travel and tourism. In the case of destination recommendation, no feature set can be readily used as a similarity ground truth, which makes it hard to evaluate the quality of destination characterization approaches. Furthermore, the process should scale well for many items, be cost-efficient, and most importantly correct. To evaluate which data sources are most suitable, we investigate 18 characterization methods that fall into three categories: venue data, textual data, and factual data. We make these data models comparable using rank agreement metrics and reveal which data sources capture similar underlying concepts. To support choosing more suitable data models, we capture a desired concept using an expert survey and evaluate our characterization methods toward it. We find that the textual models to characterize cities perform best overall, with data models based on factual and venue data being less competitive. However, we show that data models with explicit features can be optimized by learning weights for their features.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Characterizing items for content-based recommender systems is a challenging task in complex domains such as travel and tourism. In the case of destination recommendation, no feature set can be readily used as a similarity ground truth, which makes it hard to evaluate the quality of destination characterization approaches. Furthermore, the process should scale well for many items, be cost-efficient, and most importantly correct. To evaluate which data sources are most suitable, we investigate 18 characterization methods that fall into three categories: venue data, textual data, and factual data. We make these data models comparable using rank agreement metrics and reveal which data sources capture similar underlying concepts. To support choosing more suitable data models, we capture a desired concept using an expert survey and evaluate our characterization methods toward it. We find that the textual models to characterize cities perform best overall, with data models based on factual and venue data being less competitive. However, we show that data models with explicit features can be optimized by learning weights for their features.
27.
Yim, Seung-Bin; Wünsche, Katharina; Cetin, Asil; Neidhardt, Julia; Baumann, Andreas; Wissik, Tanja
Visualizing Parliamentary Speeches as Networks: the DYLEN Tool Proceedings Article
In: Fišer, Darja; Eskevich, Maria; Lenardič, Jakob; Jong, Franciska (Ed.): pp. 56–60, European Language Resources Association (ELRA), 2022.
@inproceedings{ 20.500.12708_177100,
title = {Visualizing Parliamentary Speeches as Networks: the DYLEN Tool},
author = {Seung-Bin Yim and Katharina Wünsche and Asil Cetin and Julia Neidhardt and Andreas Baumann and Tanja Wissik},
editor = {Darja Fišer and Maria Eskevich and Jakob Lenardič and Franciska Jong},
doi = {10.34726/4105},
year = {2022},
date = {2022-01-01},
pages = {56–60},
publisher = {European Language Resources Association (ELRA)},
abstract = {In this paper, we present a web based interactive visualization tool for lexical networks based on the utterances of Austrian Members of Parliament. The tool is designed to compare two networks in parallel and is composed of graph visualization, node-metrics comparison and time-series comparison components that are interconnected with each other.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, we present a web based interactive visualization tool for lexical networks based on the utterances of Austrian Members of Parliament. The tool is designed to compare two networks in parallel and is composed of graph visualization, node-metrics comparison and time-series comparison components that are interconnected with each other.
28.
Kolb, Thomas; Sekanina, Katharina; Kern, Bettina Manuela Johanna; Neidhardt, Julia; Wissik, Tanja; Baumann, Andreas
The ALPIN Sentiment Dictionary: Austrian Language Polarity in Newspapers Proceedings Article
In: pp. 4708–4716, European Language Resources Association, Marseille, France, 2022.
@inproceedings{ 20.500.12708_177094,
title = {The ALPIN Sentiment Dictionary: Austrian Language Polarity in Newspapers},
author = {Thomas Kolb and Katharina Sekanina and Bettina Manuela Johanna Kern and Julia Neidhardt and Tanja Wissik and Andreas Baumann},
doi = {10.34726/4101},
year = {2022},
date = {2022-01-01},
pages = {4708–4716},
publisher = {European Language Resources Association},
address = {Marseille, France},
abstract = {This paper introduces the Austrian German sentiment dictionary ALPIN to account for the lack of resources for dictionary-based sentiment analysis in this specific variety of German, which is characterized by lexical idiosyncrasies that also affect word sentiment. The proposed language resource is based on Austrian news media in the field of politics, an austriacism list based on different resources and a posting data set based on a popular Austrian news media. Different resources are used to increase the diversity of the resulting language resource. Extensive crowd-sourcing is performed followed by evaluation and automatic conversion into sentiment scores. We show that crowd-sourcing enables the creation of a sentiment dictionary for the Austrian German domain. Additionally, the different parts of the sentiment dictionary are evaluated to show their impact on the resulting resource. Furthermore, the proposed dictionary is utilized in a web application and available for future research and free to use for anyone.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper introduces the Austrian German sentiment dictionary ALPIN to account for the lack of resources for dictionary-based sentiment analysis in this specific variety of German, which is characterized by lexical idiosyncrasies that also affect word sentiment. The proposed language resource is based on Austrian news media in the field of politics, an austriacism list based on different resources and a posting data set based on a popular Austrian news media. Different resources are used to increase the diversity of the resulting language resource. Extensive crowd-sourcing is performed followed by evaluation and automatic conversion into sentiment scores. We show that crowd-sourcing enables the creation of a sentiment dictionary for the Austrian German domain. Additionally, the different parts of the sentiment dictionary are evaluated to show their impact on the resulting resource. Furthermore, the proposed dictionary is utilized in a web application and available for future research and free to use for anyone.
29.
Neidhardt, Julia; Werthner, Hannes; Woltran, Stefan
It Is Simple, It Is Complicated. Perspectives on Digital Humanism. Book Section
In: Werthner, Hannes; Prem, Erich; Lee, Edward A.; Ghezzi, Carlo (Ed.): Perspectives on Digital Humanism, pp. 335–342, Springer, 2022.
@incollection{ 20.500.12708_30705,
title = {It Is Simple, It Is Complicated. Perspectives on Digital Humanism.},
author = {Julia Neidhardt and Hannes Werthner and Stefan Woltran},
editor = {Hannes Werthner and Erich Prem and Edward A. Lee and Carlo Ghezzi},
year = {2022},
date = {2022-01-01},
booktitle = {Perspectives on Digital Humanism},
pages = {335–342},
publisher = {Springer},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
30.
Kolb, Thomas Elmar
Dynamic sentiment analysis for measuring media bias Masters Thesis
Technische Universität Wien, Wien, 2022.
@mastersthesis{ 20.500.12708_19920,
title = {Dynamic sentiment analysis for measuring media bias},
author = {Thomas Elmar Kolb},
doi = {10.34726/hss.2022.73300},
year = {2022},
date = {2022-01-01},
address = {Wien},
school = {Technische Universität Wien},
abstract = {Analyzing the sentiments of texts in the field of social news and news media is a big area of interest for many researchers around the world. It is a well-known problem to “teach” machines to understand the sentiments of texts e.g. news media. This master thesis aims to unveil the sentiments towards persons of public interests, who are often presented in emotionally charged contexts, for different media and over time with a focus on Vienna. Although sentiment analysis has been widely applied for analysing news and social media content, there are still many challenges unsolved. This is particularly true for the area of news media as it is not as well researched in this context as the area of social media (e.g., Twitter). Sentiment analysis is strongly language dependent. Sentiment analysis of German texts, however, hardly considers the specifics of Austrian German. To tackle these research gaps, this research performs a supervised machine learning (SML) analysis for analyzing the sentiment towards politicians over the time, and different media of the Austrian Media Corpus (AMC) different methods were compared. Sentiment analysis has been shown to be very challenging in this narrow domain. Nevertheless, results show that it is possible to predict the polarity of politicians over time. Modern state-of-the-art approaches such as BERT based models outperform traditional approaches but are not transparent, which is important when it comes to explainability and fairness. To overcome this lack of transparency, a lexical-based method was used, resulting in a new sentiment dictionary. This sentiment dictionary can be further used for research in this field and is called “Austrian Language Polarity in Newspapers (ALPIN)”. The developed models form the basis of a web application to explore media coverage of Viennese politicians and related sentiment dynamics.},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
Analyzing the sentiments of texts in the field of social news and news media is a big area of interest for many researchers around the world. It is a well-known problem to “teach” machines to understand the sentiments of texts e.g. news media. This master thesis aims to unveil the sentiments towards persons of public interests, who are often presented in emotionally charged contexts, for different media and over time with a focus on Vienna. Although sentiment analysis has been widely applied for analysing news and social media content, there are still many challenges unsolved. This is particularly true for the area of news media as it is not as well researched in this context as the area of social media (e.g., Twitter). Sentiment analysis is strongly language dependent. Sentiment analysis of German texts, however, hardly considers the specifics of Austrian German. To tackle these research gaps, this research performs a supervised machine learning (SML) analysis for analyzing the sentiment towards politicians over the time, and different media of the Austrian Media Corpus (AMC) different methods were compared. Sentiment analysis has been shown to be very challenging in this narrow domain. Nevertheless, results show that it is possible to predict the polarity of politicians over time. Modern state-of-the-art approaches such as BERT based models outperform traditional approaches but are not transparent, which is important when it comes to explainability and fairness. To overcome this lack of transparency, a lexical-based method was used, resulting in a new sentiment dictionary. This sentiment dictionary can be further used for research in this field and is called “Austrian Language Polarity in Newspapers (ALPIN)”. The developed models form the basis of a web application to explore media coverage of Viennese politicians and related sentiment dynamics.
31.
Hofstätter, Sebastian; Khattab, Omar; Althammer, Sophia; Sertkan, Mete; Hanbury, Allan
Introducing Neural Bag of Whole-Words with ColBERTer: Contextualized Late Interactions using Enhanced Reduction Proceedings Article
In: pp. 737–747, Association for Computing Machinery (ACM), New York, US, 2022.
@inproceedings{ 20.500.12708_152296,
title = {Introducing Neural Bag of Whole-Words with ColBERTer: Contextualized Late Interactions using Enhanced Reduction},
author = {Sebastian Hofstätter and Omar Khattab and Sophia Althammer and Mete Sertkan and Allan Hanbury},
doi = {10.1145/3511808.3557367},
year = {2022},
date = {2022-01-01},
pages = {737–747},
publisher = {Association for Computing Machinery (ACM)},
address = {New York, US},
abstract = {Recent progress in neural information retrieval has demonstrated large gains in quality, while often sacrificing efficiency and interpretability compared to classical approaches. We propose ColBERTer, a neural retrieval model using contextualized late interaction (ColBERT) with enhanced reduction. Along the effectiveness Pareto frontier, ColBERTer dramatically lowers ColBERT's storage requirements while simultaneously improving the interpretability of its token-matching scores. To this end, ColBERTer fuses single-vector retrieval, multi-vector refinement, and optional lexical matching components into one model. For its multi-vector component, ColBERTer reduces the number of stored vectors by learning unique whole-word representations and learning to identify and remove word representations that are not essential to effective scoring. We employ an explicit multi-task, multi-stage training to facilitate using very small vector dimensions. Results on the MS MARCO and TREC-DL collection show that ColBERTer reduces the storage footprint by up to 2.5x, while maintaining effectiveness. With just one dimension per token in its smallest setting, ColBERTer achieves index storage parity with the plaintext size, with very strong effectiveness results. Finally, we demonstrate ColBERTer's robustness on seven high-quality out-of-domain collections, yielding statistically significant gains over traditional retrieval baselines.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Recent progress in neural information retrieval has demonstrated large gains in quality, while often sacrificing efficiency and interpretability compared to classical approaches. We propose ColBERTer, a neural retrieval model using contextualized late interaction (ColBERT) with enhanced reduction. Along the effectiveness Pareto frontier, ColBERTer dramatically lowers ColBERT’s storage requirements while simultaneously improving the interpretability of its token-matching scores. To this end, ColBERTer fuses single-vector retrieval, multi-vector refinement, and optional lexical matching components into one model. For its multi-vector component, ColBERTer reduces the number of stored vectors by learning unique whole-word representations and learning to identify and remove word representations that are not essential to effective scoring. We employ an explicit multi-task, multi-stage training to facilitate using very small vector dimensions. Results on the MS MARCO and TREC-DL collection show that ColBERTer reduces the storage footprint by up to 2.5x, while maintaining effectiveness. With just one dimension per token in its smallest setting, ColBERTer achieves index storage parity with the plaintext size, with very strong effectiveness results. Finally, we demonstrate ColBERTer’s robustness on seven high-quality out-of-domain collections, yielding statistically significant gains over traditional retrieval baselines.
32.
Hofstätter, Sebastian; Althammer, Sophia; Sertkan, Mete; Hanbury, Allan
A Time-Optimized Content Creation Workflow for Remote Teaching Proceedings Article
In: pp. 731–737, Association for Computing Machinery, New York, NY, United States, 2022.
@inproceedings{ 20.500.12708_193896,
title = {A Time-Optimized Content Creation Workflow for Remote Teaching},
author = {Sebastian Hofstätter and Sophia Althammer and Mete Sertkan and Allan Hanbury},
doi = {10.1145/3478431.3499421},
year = {2022},
date = {2022-01-01},
volume = {1},
pages = {731–737},
publisher = {Association for Computing Machinery},
address = {New York, NY, United States},
abstract = {We describe our workflow to create an engaging remote learning experience for a university course, while minimizing the post-production time of the educators. We make use of ubiquitous and commonly free services and platforms, so that our workflow is inclusive for all educators and provides polished experiences for students. Our learning materials provide for each lecture: 1) a recorded video, uploaded on YouTube, with exact slide timestamp indices, which enables an enhanced navigation UI; and 2) a high-quality flow-text automated transcript of the narration with proper punctuation and capitalization, improved with a student participation workflow on GitHub. All these results could be created by hand in a time consuming and costly way. However, this would generally exceed the time available for creating course materials. Our main contribution is to automate the transformation and post-production between raw narrated slides and our published materials with a custom toolchain. Furthermore, we describe our complete workflow: from content creation to transformation and distribution. Our students gave us overwhelmingly positive feedback and especially liked our use of ubiquitous platforms. The most used feature was YouTube's chapter UI enabled through our automatically generated timestamps. The majority of students, who started using the transcripts, continued to do so. Every single transcript was corrected by students, with an average word-change of 6%. We conclude with the positive feedback that our enhanced content formats are much appreciated and utilized. Important for educators is how our low overhead production workflow was sustainable throughout a busy semester.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We describe our workflow to create an engaging remote learning experience for a university course, while minimizing the post-production time of the educators. We make use of ubiquitous and commonly free services and platforms, so that our workflow is inclusive for all educators and provides polished experiences for students. Our learning materials provide for each lecture: 1) a recorded video, uploaded on YouTube, with exact slide timestamp indices, which enables an enhanced navigation UI; and 2) a high-quality flow-text automated transcript of the narration with proper punctuation and capitalization, improved with a student participation workflow on GitHub. All these results could be created by hand in a time consuming and costly way. However, this would generally exceed the time available for creating course materials. Our main contribution is to automate the transformation and post-production between raw narrated slides and our published materials with a custom toolchain. Furthermore, we describe our complete workflow: from content creation to transformation and distribution. Our students gave us overwhelmingly positive feedback and especially liked our use of ubiquitous platforms. The most used feature was YouTube’s chapter UI enabled through our automatically generated timestamps. The majority of students, who started using the transcripts, continued to do so. Every single transcript was corrected by students, with an average word-change of 6%. We conclude with the positive feedback that our enhanced content formats are much appreciated and utilized. Important for educators is how our low overhead production workflow was sustainable throughout a busy semester.
33.
Althammer, Sophia; Hofstätter, Sebastian; Sertkan, Mete; Verberne, Suzan; Hanbury, Allan
PARM: A Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval Proceedings Article
In: pp. 19–34, Springer, 2022.
@inproceedings{ 20.500.12708_137002,
title = {PARM: A Paragraph Aggregation Retrieval Model for Dense Document-to-Document Retrieval},
author = {Sophia Althammer and Sebastian Hofstätter and Mete Sertkan and Suzan Verberne and Allan Hanbury},
doi = {10.1007/978-3-030-99736-6_2},
year = {2022},
date = {2022-01-01},
pages = {19–34},
publisher = {Springer},
abstract = {Dense passage retrieval (DPR) models show great effectiveness gains in first stage retrieval for the web domain. However in the web domain we are in a setting with large amounts of training data and a query-to-passage or a query-to-document retrieval task. We investigate in this paper dense document-to-document retrieval with limited labelled target data for training, in particular legal case retrieval. In order to use DPR models for document-to-document retrieval, we propose a Paragraph Aggregation Retrieval Model (PARM) which liberates DPR models from their limited input length. PARM retrieves documents on the paragraph-level: for each query paragraph, relevant documents are retrieved based on their paragraphs. Then the relevant results per query paragraph are aggregated into one ranked list for the whole query document. For the aggregation we propose vector-based aggregation with reciprocal rank fusion (VRRF) weighting, which combines the advantages of rank-based aggregation and topical aggregation based on the dense embeddings. Experimental results show that VRRF outperforms rank-based aggregation strategies for dense document-to-document retrieval with PARM. We compare PARM to document-level retrieval and demonstrate higher retrieval effectiveness of PARM for lexical and dense first-stage retrieval on two different legal case retrieval collections. We investigate how to train the dense retrieval model for PARM on limited target data with labels on the paragraph or the document-level. In addition, we analyze the differences of the retrieved results of lexical and dense retrieval with PARM.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
Dense passage retrieval (DPR) models show great effectiveness gains in first stage retrieval for the web domain. However in the web domain we are in a setting with large amounts of training data and a query-to-passage or a query-to-document retrieval task. We investigate in this paper dense document-to-document retrieval with limited labelled target data for training, in particular legal case retrieval. In order to use DPR models for document-to-document retrieval, we propose a Paragraph Aggregation Retrieval Model (PARM) which liberates DPR models from their limited input length. PARM retrieves documents on the paragraph-level: for each query paragraph, relevant documents are retrieved based on their paragraphs. Then the relevant results per query paragraph are aggregated into one ranked list for the whole query document. For the aggregation we propose vector-based aggregation with reciprocal rank fusion (VRRF) weighting, which combines the advantages of rank-based aggregation and topical aggregation based on the dense embeddings. Experimental results show that VRRF outperforms rank-based aggregation strategies for dense document-to-document retrieval with PARM. We compare PARM to document-level retrieval and demonstrate higher retrieval effectiveness of PARM for lexical and dense first-stage retrieval on two different legal case retrieval collections. We investigate how to train the dense retrieval model for PARM on limited target data with labels on the paragraph or the document-level. In addition, we analyze the differences of the retrieved results of lexical and dense retrieval with PARM.
34.
Aayesha,; Qureshi, Muhammad Bilal; Afzaal, Muhammad; Qureshi, Muhammad Shuaib; Gwak, Jeonghwan; others,
Fuzzy-Based Automatic Epileptic Seizure Detection Framework. Journal Article
In: Computers, Materials & Continua, vol. 70, no. 3, 2022.
@article{qureshi2022fuzzy,
title = {Fuzzy-Based Automatic Epileptic Seizure Detection Framework.},
author = {Aayesha and Muhammad Bilal Qureshi and Muhammad Afzaal and Muhammad Shuaib Qureshi and Jeonghwan Gwak and others},
year = {2022},
date = {2022-01-01},
journal = {Computers, Materials & Continua},
volume = {70},
number = {3},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
35.
Neidhardt, Julia
Network Science and e-Tourism Book Section
In: Xiang, Zheng; Fuchs, Matthias; Gretzel, Ulrike; Höpken, Wolfram (Ed.): Springer Cham, 2021.
@incollection{ 20.500.12708_152206,
title = {Network Science and e-Tourism},
author = {Julia Neidhardt},
editor = {Zheng Xiang and Matthias Fuchs and Ulrike Gretzel and Wolfram Höpken},
doi = {10.1007/978-3-030-05324-6_33-1},
year = {2021},
date = {2021-01-01},
publisher = {Springer Cham},
abstract = {This chapter provides an introduction to network science and its applications within e-tourism research. In the first part, an overview of network science as a continuously growing scientific field is given. Network science provides various concepts and methods for the analysis of the structure and dynamics of all kinds of networks such as social networks, information networks, and economic networks. Afterward, popular software and tools to model, analyze, and visualize network data are briefly discussed. In the third part, an overview of research in e-tourism that utilized network science methods is provided. In existing studies, different types of networks were constructed and analyzed, in particular networks of travelers, networks of tourism websites, networks capturing behavioral patterns of travelers, or text networks of travel-related posts. Furthermore, it is briefly discussed, which data sources are typically used in the literature. Finally, the main points are summarized and conclusions are drawn.},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
This chapter provides an introduction to network science and its applications within e-tourism research. In the first part, an overview of network science as a continuously growing scientific field is given. Network science provides various concepts and methods for the analysis of the structure and dynamics of all kinds of networks such as social networks, information networks, and economic networks. Afterward, popular software and tools to model, analyze, and visualize network data are briefly discussed. In the third part, an overview of research in e-tourism that utilized network science methods is provided. In existing studies, different types of networks were constructed and analyzed, in particular networks of travelers, networks of tourism websites, networks capturing behavioral patterns of travelers, or text networks of travel-related posts. Furthermore, it is briefly discussed, which data sources are typically used in the literature. Finally, the main points are summarized and conclusions are drawn.
36.
Kolb, Thomas Elmar; Sekanina, Katharina; Kern, Bettina Manuela Johanna; Neidhardt, Julia; Baumann, Andreas; Wissik, Tanja
Creating an Austrian language polarity dictionary with the crowd Miscellaneous
2021.
@misc{ 20.500.12708_153831,
title = {Creating an Austrian language polarity dictionary with the crowd},
author = {Thomas Elmar Kolb and Katharina Sekanina and Bettina Manuela Johanna Kern and Julia Neidhardt and Andreas Baumann and Tanja Wissik},
year = {2021},
date = {2021-01-01},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
37.
Neidhardt, Julia; Wörndl, Wolfgang; Kuflik, Tsvi; Zanker, Markus
Workshop on Recommenders in Tourism (RecTour) Proceedings Article
In: Fifteenth ACM Conference on Recommender Systems, ACM, 2021.
@inproceedings{ 20.500.12708_58556,
title = {Workshop on Recommenders in Tourism (RecTour)},
author = {Julia Neidhardt and Wolfgang Wörndl and Tsvi Kuflik and Markus Zanker},
doi = {10.1145/3460231.3470930},
year = {2021},
date = {2021-01-01},
booktitle = {Fifteenth ACM Conference on Recommender Systems},
publisher = {ACM},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
38.
Kuflik, Tsvi; Barbu, Catalin Mihai; Delic, Amra; Goldenberg, Dmitri; Neidhardt, Julia; Coba, Ludovik; Zanker, Markus
WebTour 2021 Workshop on Web Tourism Proceedings Article
In: Proceedings of the 14th ACM International Conference on Web Search and Data Mining, ACM, 2021.
@inproceedings{ 20.500.12708_58574,
title = {WebTour 2021 Workshop on Web Tourism},
author = {Tsvi Kuflik and Catalin Mihai Barbu and Amra Delic and Dmitri Goldenberg and Julia Neidhardt and Ludovik Coba and Markus Zanker},
doi = {10.1145/3437963.3441836},
year = {2021},
date = {2021-01-01},
booktitle = {Proceedings of the 14th ACM International Conference on Web Search and Data Mining},
publisher = {ACM},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
39.
Neidhardt, Julia
Digital Humanism Ringvorlesung “Introduction to Digital Humanities” Miscellaneous
2021.
@misc{ 20.500.12708_87290,
title = {Digital Humanism Ringvorlesung "Introduction to Digital Humanities"},
author = {Julia Neidhardt},
year = {2021},
date = {2021-01-01},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
40.
Neidhardt, Julia
Digital Humanism. Digital Humanism – AI and Ethics Miscellaneous
2021.
@misc{ 20.500.12708_87289,
title = {Digital Humanism. Digital Humanism - AI and Ethics},
author = {Julia Neidhardt},
year = {2021},
date = {2021-01-01},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}