Publikationen

2021

  • B. van Aken, J. Papaioannou, M. Mayrdorfer, K. Budde, F. Gers, and A. Loeser, „Clinical Outcome Prediction from Admission Notes using Self-Supervised Knowledge Integration,“ in Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, Online, 2021, p. 881–893.
    [BibTeX] [Abstract] [Download PDF]

    Outcome prediction from clinical text can prevent doctors from overlooking possible risks and help hospitals to plan capacities. We simulate patients at admission time, when decision support can be especially valuable, and contribute a novel *admission to discharge* task with four common outcome prediction targets: Diagnoses at discharge, procedures performed, in-hospital mortality and length-of-stay prediction. The ideal system should infer outcomes based on symptoms, pre-conditions and risk factors of a patient. We evaluate the effectiveness of language models to handle this scenario and propose *clinical outcome pre-training* to integrate knowledge about patient outcomes from multiple public sources. We further present a simple method to incorporate ICD code hierarchy into the models. We show that our approach improves performance on the outcome tasks against several baselines. A detailed analysis reveals further strengths of the model, including transferability, but also weaknesses such as handling of vital values and inconsistencies in the underlying data.

    @inproceedings{van-aken-etal-2021-clinical,
    title = "Clinical Outcome Prediction from Admission Notes using Self-Supervised Knowledge Integration",
    author = "van Aken, Betty and
    Papaioannou, Jens-Michalis and
    Mayrdorfer, Manuel and
    Budde, Klemens and
    Gers, Felix and
    Loeser, Alexander",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2021.eacl-main.75",
    pages = "881--893",
    abstract = "Outcome prediction from clinical text can prevent doctors from overlooking possible risks and help hospitals to plan capacities. We simulate patients at admission time, when decision support can be especially valuable, and contribute a novel *admission to discharge* task with four common outcome prediction targets: Diagnoses at discharge, procedures performed, in-hospital mortality and length-of-stay prediction. The ideal system should infer outcomes based on symptoms, pre-conditions and risk factors of a patient. We evaluate the effectiveness of language models to handle this scenario and propose *clinical outcome pre-training* to integrate knowledge about patient outcomes from multiple public sources. We further present a simple method to incorporate ICD code hierarchy into the models. We show that our approach improves performance on the outcome tasks against several baselines. A detailed analysis reveals further strengths of the model, including transferability, but also weaknesses such as handling of vital values and inconsistencies in the underlying data.",
    }

  • S. Castle, R. Schwarzenberg, and M. Pourvali, „Detecting Covariate Drift with Explanations,“ in Natural Language Processing and Chinese Computing, Cham, 2021, p. 317–322.
    [BibTeX] [Abstract]

    Detecting when there is a domain drift between training and inference data is important for any model evaluated on data collected in real time. Many current data drift detection methods only utilize input features to detect domain drift. While effective, these methods disregard the model’s evaluation of the data, which may be a significant source of information about the data domain. We propose to use information from the model in the form of explanations, specifically gradient times input, in order to utilize this information. Following the framework of Rabanser et al. [11], we combine these explanations with two-sample tests in order to detect a shift in distribution between training and evaluation data. Promising initial experiments show that explanations provide useful information for detecting shift, which potentially improves upon the current state-of-the-art.

    @InProceedings{10.1007/978-3-030-88483-3_24,
    author="Castle, Steffen
    and Schwarzenberg, Robert
    and Pourvali, Mohsen",
    editor="Wang, Lu
    and Feng, Yansong
    and Hong, Yu
    and He, Ruifang",
    title="Detecting Covariate Drift with Explanations",
    booktitle="Natural Language Processing and Chinese Computing",
    year="2021",
    publisher="Springer International Publishing",
    address="Cham",
    pages="317--322",
    abstract="Detecting when there is a domain drift between training and inference data is important for any model evaluated on data collected in real time. Many current data drift detection methods only utilize input features to detect domain drift. While effective, these methods disregard the model's evaluation of the data, which may be a significant source of information about the data domain. We propose to use information from the model in the form of explanations, specifically gradient times input, in order to utilize this information. Following the framework of Rabanser et al. [11], we combine these explanations with two-sample tests in order to detect a shift in distribution between training and evaluation data. Promising initial experiments show that explanations provide useful information for detecting shift, which potentially improves upon the current state-of-the-art.",
    isbn="978-3-030-88483-3"
    }

  • J. Frey and S. Hellmann, „FAIR Linked Data – Towards a Linked Data Backbone for Users and Machines,“ in Companion of The Web Conference 2021, Virtual Event / Ljubljana, Slovenia, April 19-23, 2021, 2021, p. 431–435. doi:10.1145/3442442.3451364
    [BibTeX] [Download PDF]
    @inproceedings{DBLP:conf/www/FreyH21,
    author = {Johannes Frey and
    Sebastian Hellmann},
    editor = {Jure Leskovec and
    Marko Grobelnik and
    Marc Najork and
    Jie Tang and
    Leila Zia},
    title = {{FAIR} Linked Data - Towards a Linked Data Backbone for Users and
    Machines},
    booktitle = {Companion of The Web Conference 2021, Virtual Event / Ljubljana, Slovenia,
    April 19-23, 2021},
    pages = {431--435},
    publisher = {{ACM} / {IW3C2}},
    year = {2021},
    url = {https://doi.org/10.1145/3442442.3451364},
    doi = {10.1145/3442442.3451364},
    timestamp = {Mon, 07 Jun 2021 14:34:36 +0200},
    biburl = {https://dblp.org/rec/conf/www/FreyH21.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}
    }

  • S. Hellmann, J. Frey, M. Hofer, M. Dojchinovski, K. Wecel, and W. Lewoniewski, „Towards a Systematic Approach to Sync Factual Data across Wikipedia, Wikidata and External Data Sources,“ in Proceedings of the Conference on Digital Curation Technologies (Qurator 2021), Berlin, Germany, February 8th – to – 12th, 2021, 2021.
    [BibTeX] [Download PDF]
    @inproceedings{DBLP:conf/qurator/HellmannFHDWL21,
    author = {Sebastian Hellmann and
    Johannes Frey and
    Marvin Hofer and
    Milan Dojchinovski and
    Krzysztof Wecel and
    Wlodzimierz Lewoniewski},
    editor = {Adrian Paschke and
    Georg Rehm and
    Jamal Al Qundus and
    Clemens Neudecker and
    Lydia Pintscher},
    title = {Towards a Systematic Approach to Sync Factual Data across Wikipedia,
    Wikidata and External Data Sources},
    booktitle = {Proceedings of the Conference on Digital Curation Technologies (Qurator
    2021), Berlin, Germany, February 8th - to - 12th, 2021},
    series = {{CEUR} Workshop Proceedings},
    volume = {2836},
    publisher = {CEUR-WS.org},
    year = {2021},
    url = {http://ceur-ws.org/Vol-2836/qurator2021\_paper\_18.pdf},
    timestamp = {Wed, 28 Apr 2021 17:11:52 +0200},
    biburl = {https://dblp.org/rec/conf/qurator/HellmannFHDWL21.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}
    }

  • L. Hennig and H. Uszkoreit, „Semantische Technologien für Enterprise Intelligence am Beispiel von Lieferkettenbeobachtung,“ in Semantische Datenintelligenz im Einsatz, B. Ege and A. Paschke, Eds., Springer Vieweg, 2021, p. 87–126.
    [BibTeX]
    @incollection{hennig_2021_semantische_technologien,
    author = {Hennig, Leonhard and Uszkoreit, Hans},
    editor = {Ege, B\"{o}rte\c{c}in and Paschke, Adrian},
    title = {Semantische Technologien für Enterprise Intelligence am Beispiel von Lieferkettenbeobachtung},
    booktitle = {Semantische Datenintelligenz im Einsatz},
    year = {2021},
    pages = {87--126},
    publisher = {Springer Vieweg},
    isbn = {978-3-658-31938-0}
    }

  • J. Papaioannou, M. Mayrdorfer, S. Arnold, F. A. Gers, K. Budde, and A. Löser, „Aspect-Based Passage Retrieval with Contextualized Discourse Vectors,“ in Advances in Information Retrieval, Cham, 2021, p. 537–542.
    [BibTeX] [Abstract]

    Passage retrieval is the task of retrieving only the portions of a document that are relevant to a particular information need. One application medical doctors and researchers face is the challenge of reading a large amount of novel literature. For example, since the outbreak of Coronavirus disease 2019 (COVID-19), tens of thousands of papers have been published each month about the disease. We demonstrate how we can support healthcare professionals in this exploratory research task with our neural passage retrieval system based on Contextualized Discourse Vectors (CDV). CDV captures the discourse of long documents on sentence level and allows to query a large corpus with medical entities and aspects. Our demonstration covers over 27,000 diseases and 14,000 clinical aspects including symptoms, diagnostics, treatments and medications. It returns passages and highlights sentences to effectively answer clinical queries with up to 65{\%} Recall@1. We showcase our system on the COVID-19 Open Research Dataset (CORD-19), Orphanet and Wikipedia diseases corpora.

    @InProceedings{10.1007/978-3-030-72240-1_61,
    author="Papaioannou, Jens-Michalis
    and Mayrdorfer, Manuel
    and Arnold, Sebastian
    and Gers, Felix A.
    and Budde, Klemens
    and L{\"o}ser, Alexander",
    editor="Hiemstra, Djoerd
    and Moens, Marie-Francine
    and Mothe, Josiane
    and Perego, Raffaele
    and Potthast, Martin
    and Sebastiani, Fabrizio",
    title="Aspect-Based Passage Retrieval with Contextualized Discourse Vectors",
    booktitle="Advances in Information Retrieval",
    year="2021",
    publisher="Springer International Publishing",
    address="Cham",
    pages="537--542",
    abstract="Passage retrieval is the task of retrieving only the portions of a document that are relevant to a particular information need. One application medical doctors and researchers face is the challenge of reading a large amount of novel literature. For example, since the outbreak of Coronavirus disease 2019 (COVID-19), tens of thousands of papers have been published each month about the disease. We demonstrate how we can support healthcare professionals in this exploratory research task with our neural passage retrieval system based on Contextualized Discourse Vectors (CDV). CDV captures the discourse of long documents on sentence level and allows to query a large corpus with medical entities and aspects. Our demonstration covers over 27,000 diseases and 14,000 clinical aspects including symptoms, diagnostics, treatments and medications. It returns passages and highlights sentences to effectively answer clinical queries with up to 65{\%} Recall@1. We showcase our system on the COVID-19 Open Research Dataset (CORD-19), Orphanet and Wikipedia diseases corpora.",
    isbn="978-3-030-72240-1"
    }

  • U. Yaseen and S. Langer, „Neural Text Classification and Stacked Heterogeneous Embeddings for Named Entity Recognition in SMM4H 2021,“ in Proceedings of the Sixth Social Media Mining for Health (\#SMM4H) Workshop and Shared Task, Mexico City, Mexico, 2021, p. 83–87. doi:10.18653/v1/2021.smm4h-1.14
    [BibTeX] [Abstract]

    This paper presents our findings from participating in the SMM4H Shared Task 2021. We addressed Named Entity Recognition (NER) and Text Classification. To address NER we explored BiLSTM-CRF with Stacked Heterogeneous embeddings and linguistic features. We investigated various machine learning algorithms (logistic regression, SVM and Neural Networks) to address text classification. Our proposed approaches can be generalized to different languages and we have shown its effectiveness for English and Spanish. Our text classification submissions have achieved competitive performance with F1-score of 0.46 and 0.90 on ADE Classification (Task 1a) and Profession Classification (Task 7a) respectively. In the case of NER, our submissions scored F1-score of 0.50 and 0.82 on ADE Span Detection (Task 1b) and Profession span detection (Task 7b) respectively.

    @inproceedings{yaseen-langer-2021-neural,
    title = "Neural Text Classification and Stacked Heterogeneous Embeddings for Named Entity Recognition in {SMM}4{H} 2021",
    author = "Yaseen, Usama and Langer, Stefan",
    booktitle = "Proceedings of the Sixth Social Media Mining for Health ({\#}SMM4H) Workshop and Shared Task",
    month = jun,
    year = "2021",
    address = "Mexico City, Mexico",
    publisher = "Association for Computational Linguistics",
    url = https://www.aclweb.org/anthology/2021.smm4h-1.14,
    doi = "10.18653/v1/2021.smm4h-1.14",
    pages = "83--87",
    abstract = "This paper presents our findings from participating in the SMM4H Shared Task 2021. We addressed Named Entity Recognition (NER) and Text Classification. To address NER we explored BiLSTM-CRF with Stacked Heterogeneous embeddings and linguistic features. We investigated various machine learning algorithms (logistic regression, SVM and Neural Networks) to address text classification. Our proposed approaches can be generalized to different languages and we have shown its effectiveness for English and Spanish. Our text classification submissions have achieved competitive performance with F1-score of 0.46 and 0.90 on ADE Classification (Task 1a) and Profession Classification (Task 7a) respectively. In the case of NER, our submissions scored F1-score of 0.50 and 0.82 on ADE Span Detection (Task 1b) and Profession span detection (Task 7b) respectively.",
    }

2020

  • C. Alt, A. Gabryszak, and L. Hennig, „Probing Linguistic Features of Sentence-Level Representations in Neural Relation Extraction,“ in Proceedings of ACL 2020, 2020.
    [BibTeX] [Abstract] [Download PDF]

    Despite the recent progress, little is known about the features captured by state-of-the-art neural relation extraction (RE) models. Common methods encode the source sentence, conditioned on the entity mentions, before classifying the relation. However, the complexity of the task makes it difficult to understand how encoder architecture and supporting linguistic knowledge affect the features learned by the encoder. We introduce 14 probing tasks targeting linguistic properties relevant to RE, and we use them to study representations learned by more than 40 different encoder architecture and linguistic feature combinations trained on two datasets, TACRED and SemEval 2010 Task 8. We find that the bias induced by the architecture and the inclusion of linguistic features are clearly expressed in the probing task performance. For example, adding contextualized word representations greatly increases performance on probing tasks with a focus on named entity and part-of-speech information, and yields better results in RE. In contrast, entity masking improves RE, but considerably lowers performance on entity type related probing tasks.

    @inproceedings{Alt2020ProbingLF,
    title={Probing Linguistic Features of Sentence-Level Representations in Neural Relation Extraction},
    author={Christoph Alt and Aleksandra Gabryszak and Leonhard Hennig},
    abstract={Despite the recent progress, little is known about the features captured by state-of-the-art neural relation extraction (RE) models. Common methods encode the source sentence, conditioned on the entity mentions, before classifying the relation. However, the complexity of the task makes it difficult to understand how encoder architecture and supporting linguistic knowledge affect the features learned by the encoder. We introduce 14 probing tasks targeting linguistic properties relevant to RE, and we use them to study representations learned by more than 40 different encoder architecture and linguistic feature combinations trained on two datasets, TACRED and SemEval 2010 Task 8. We find that the bias induced by the architecture and the inclusion of linguistic features are clearly expressed in the probing task performance. For example, adding contextualized word representations greatly increases performance on probing tasks with a focus on named entity and part-of-speech information, and yields better results in RE. In contrast, entity masking improves RE, but considerably lowers performance on entity type related probing tasks.},
    booktitle={Proceedings of ACL 2020},
    year={2020},
    url={https://arxiv.org/abs/2004.08134},
    publisher = {Association for Computational Linguistics},
    }

  • C. Alt, A. Gabryszak, and L. Hennig, „TACRED Revisited: A Thorough Evaluation of the TACRED Relation Extraction Task,“ in Proceedings of ACL 2020, 2020.
    [BibTeX] [Abstract] [Download PDF]

    TACRED (Zhang et al., 2017) is one of the largest, most widely used crowdsourced datasets in Relation Extraction (RE). But, even with recent advances in unsupervised pre-training and knowledge enhanced neural RE, models still show a high error rate. In this paper, we investigate the questions: Have we reached a performance ceiling or is there still room for improvement? And how do crowd annotations, dataset, and models contribute to this error rate? To answer these questions, we first validate the most challenging 5K examples in the development and test sets using trained annotators. We find that label errors account for 8% absolute F1 test error, and that more than 50% of the examples need to be relabeled. On the relabeled test set the average F1 score of a large baseline model set improves from 62.1 to 70.1. After validation, we analyze misclassifications on the challenging instances, categorize them into linguistically motivated error groups, and verify the resulting error hypotheses on three state-of-the-art RE models. We show that two groups of ambiguous relations are responsible for most of the remaining errors and that models may adopt shallow heuristics on the dataset when entities are not masked.

    @inproceedings{Alt2020TACREDRA,
    title={TACRED Revisited: A Thorough Evaluation of the TACRED Relation Extraction Task},
    author={Christoph Alt and Aleksandra Gabryszak and Leonhard Hennig},
    booktitle={Proceedings of ACL 2020},
    abstract={TACRED (Zhang et al., 2017) is one of the largest, most widely used crowdsourced datasets in Relation Extraction (RE). But, even with recent advances in unsupervised pre-training and knowledge enhanced neural RE, models still show a high error rate. In this paper, we investigate the questions: Have we reached a performance ceiling or is there still room for improvement? And how do crowd annotations, dataset, and models contribute to this error rate? To answer these questions, we first validate the most challenging 5K examples in the development and test sets using trained annotators. We find that label errors account for 8% absolute F1 test error, and that more than 50% of the examples need to be relabeled. On the relabeled test set the average F1 score of a large baseline model set improves from 62.1 to 70.1. After validation, we analyze misclassifications on the challenging instances, categorize them into linguistically motivated error groups, and verify the resulting error hypotheses on three state-of-the-art RE models. We show that two groups of ambiguous relations are responsible for most of the remaining errors and that models may adopt shallow heuristics on the dataset when entities are not masked. },
    year={2020},
    url={https://arxiv.org/abs/2004.14855},
    publisher = {Association for Computational Linguistics},
    }

  • S. Arnold, B. van Aken, P. Grundmann, F. A. Gers, and A. Löser, „Learning Contextualized Document Representations for Healthcare Answer Retrieval,“ in WWW ’20: The Web Conference 2020, Taipei, Taiwan, April 20-24, 2020, 2020, p. 1332–1343. doi:10.1145/3366423.3380208
    [BibTeX] [Download PDF]
    @inproceedings{DBLP:conf/www/0001AGGL20,
    author = {Sebastian Arnold and
    Betty van Aken and
    Paul Grundmann and
    Felix A. Gers and
    Alexander L{\"{o}}ser},
    editor = {Yennun Huang and
    Irwin King and
    Tie{-}Yan Liu and
    Maarten van Steen},
    title = {Learning Contextualized Document Representations for Healthcare Answer
    Retrieval},
    booktitle = {{WWW} '20: The Web Conference 2020, Taipei, Taiwan, April 20-24, 2020},
    pages = {1332--1343},
    publisher = {{ACM} / {IW3C2}},
    year = {2020},
    url = {https://doi.org/10.1145/3366423.3380208},
    doi = {10.1145/3366423.3380208},
    timestamp = {Wed, 06 May 2020 12:56:16 +0200},
    biburl = {https://dblp.org/rec/conf/www/0001AGGL20.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}
    }

  • Y. Chaudhary, H. Schütze, and P. Gupta, „Explainable and Discourse Topic-aware Neural Language Understanding,“ in Proceedings of the 37th International Conference on Machine Learning, ICML, virtual, 2020.
    [BibTeX]
    @inproceedings{chaudhary-etal-2020-nclm-icml,
    title = "Explainable and Discourse Topic-aware Neural Language Understanding",
    author = {Chaudhary, Yatin and Sch{\"u}tze, Hinrich and Gupta, Pankaj},
    booktitle = "Proceedings of the 37th International Conference on Machine Learning, ICML",
    month = jul,
    year = "2020",
    address = "virtual"
    }

  • J. Frey, D. Streitmatter, F. Götz, and S. Hellmann, „DBpedia Archivo – A Web-Scale Interface for Ontology Archiving under Consumer-oriented Aspects,“ submitted to Semantics 2020, 2020.
    [BibTeX] [Download PDF]
    @article{freydbpedia,
    title = {DBpedia Archivo - A Web-Scale Interface for Ontology Archiving under Consumer-oriented Aspects},
    added-at = {2020-06-29T15:55:45.000+0200},
    author = {Frey, Johannes and Streitmatter, Denis and G{\"o}tz, Fabian and Hellmann, Sebastian},
    biburl = {https://www.bibsonomy.org/bibtex/25f8565541c6412590594b898a632fc0a/aksw},
    interhash = {2836caa18d470561ec3169b3f59aaf21},
    intrahash = {5f8565541c6412590594b898a632fc0a},
    journal = {submitted to Semantics 2020},
    keywords = {arndt frey group_aksw hellmann kilt streitmatter},
    timestamp = {2020-06-29T15:55:45.000+0200},
    url = {https://svn.aksw.org/papers/2020/semantics_archivo/public.pdf},
    year = 2020
    }

  • P. Gupta, Y. Chaudhary, T. Runkler, and H. Schütze, „Neural Topic Modeling with Continual Lifelong Learning,“ in Proceedings of the 37th International Conference on Machine Learning, ICML, virtual, 2020.
    [BibTeX]
    @inproceedings{gupta-etal-2020-lntm-icml,
    title = "Neural Topic Modeling with Continual Lifelong Learning",
    author = {Gupta, Pankaj and Chaudhary, Yatin and Runkler, Thomas and Sch{\"u}tze, Hinrich},
    booktitle = "Proceedings of the 37th International Conference on Machine Learning, ICML",
    month = jul,
    year = "2020",
    address = "virtual"
    }

  • M. Hofer, S. Hellmann, M. Dojchinovski, and J. Frey, „The New DBpedia Release Cycle: Increasing Agility and Efficiency in Knowledge Extraction Workflows,“ submitted to Semantics 2020, vol. 16, 2020.
    [BibTeX] [Download PDF]
    @article{hoferdbpedai,
    title = {The New DBpedia Release Cycle: Increasing Agility and Efficiency in Knowledge Extraction Workflows},
    added-at = {2020-06-29T15:55:45.000+0200},
    author = {Hofer, Marvin and Hellmann, Sebastian and Dojchinovski, Milan and Frey, Johannes},
    biburl = {https://www.bibsonomy.org/bibtex/2bbb6a227c9fdcd96b332132877226b6f/aksw},
    booktitle = {16th International Conference on Semantic Systems},
    interhash = {592b96964e8c91d044c175c1708d2d34},
    intrahash = {bbb6a227c9fdcd96b332132877226b6f},
    journal = {submitted to Semantics 2020},
    keywords = {frey group_aksw hellmann hofer kilt},
    publisher = {Springer},
    series = {LNCS},
    timestamp = {2020-06-29T15:55:45.000+0200},
    url = {https://svn.aksw.org/papers/2020/semantics_marvin/public.pdf},
    volume = 16,
    year = 2020
    }

  • M. Hübner, C. Alt, R. Schwarzenberg, and L. Hennig, „Defx at SemEval-2020 Task 6: Joint Extraction of Concepts and Relations for Definition Extraction,“ in Proceedings of the Fourteenth Workshop on Semantic Evaluation, Barcelona (online), 2020, p. 704–709.
    [BibTeX] [Download PDF]
    @inproceedings{defx_at_semeval,
    title = "Defx at {S}em{E}val-2020 Task 6: Joint Extraction of Concepts and Relations for Definition Extraction",
    author = {H{\"u}bner, Marc and
    Alt, Christoph and
    Schwarzenberg, Robert and
    Hennig, Leonhard},
    booktitle = "Proceedings of the Fourteenth Workshop on Semantic Evaluation",
    month = dec,
    year = "2020",
    address = "Barcelona (online)",
    publisher = "International Committee for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2020.semeval-1.92",
    pages = "704--709"
    }

  • K. Lazaridou, A. Löser, M. Mestre, and F. Naumann, „Discovering Biased News Articles Leveraging Multiple Human Annotations,“ in Proceedings of the 12th Language Resources and Evaluation Conference, Marseille, France, 2020, p. 1268–1277.
    [BibTeX] [Abstract] [Download PDF]

    Unbiased and fair reporting is an integral part of ethical journalism. Yet, political propaganda and one-sided views can be found in the news and can cause distrust in media. Both accidental and deliberate political bias affect the readers and shape their views. We contribute to a trustworthy media ecosystem by automatically identifying politically biased news articles. We introduce novel corpora annotated by two communities, i.e., domain experts and crowd workers, and we also consider automatic article labels inferred by the newspapers{‚} ideologies. Our goal is to compare domain experts to crowd workers and also to prove that media bias can be detected automatically. We classify news articles with a neural network and we also improve our performance in a self-supervised manner.

    @inproceedings{lazaridou-etal-2020-discovering,
    title = "Discovering Biased News Articles Leveraging Multiple Human Annotations",
    author = {Lazaridou, Konstantina and
    L{\"o}ser, Alexander and
    Mestre, Maria and
    Naumann, Felix},
    booktitle = "Proceedings of the 12th Language Resources and Evaluation Conference",
    month = may,
    year = "2020",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://www.aclweb.org/anthology/2020.lrec-1.159",
    pages = "1268--1277",
    abstract = "Unbiased and fair reporting is an integral part of ethical journalism. Yet, political propaganda and one-sided views can be found in the news and can cause distrust in media. Both accidental and deliberate political bias affect the readers and shape their views. We contribute to a trustworthy media ecosystem by automatically identifying politically biased news articles. We introduce novel corpora annotated by two communities, i.e., domain experts and crowd workers, and we also consider automatic article labels inferred by the newspapers{'} ideologies. Our goal is to compare domain experts to crowd workers and also to prove that media bias can be detected automatically. We classify news articles with a neural network and we also improve our performance in a self-supervised manner.",
    language = "English",
    ISBN = "979-10-95546-34-4",
    }

  • T. Oberhauser, T. Bischoff, K. Brendel, M. Menke, T. Klatt, A. Siu, F. A. Gers, and A. Löser, „TrainX – Named Entity Linking with Active Sampling and Bi-Encoders,“ in Proceedings of the 28th International Conference on Computational Linguistics: System Demonstrations, Barcelona, Spain (Online), 2020, p. 64–69. doi:10.18653/v1/2020.coling-demos.12
    [BibTeX] [Abstract] [Download PDF]

    We demonstrate TrainX, a system for Named Entity Linking for medical experts. It combines state-of-the-art entity recognition and linking architectures, such as Flair and fine-tuned Bi-Encoders based on BERT, with an easy-to-use interface for healthcare professionals. We support medical experts in annotating training data by using active sampling strategies to forward informative samples to the annotator. We demonstrate that our model is capable of linking against large knowledge bases, such as UMLS (3.6 million entities), and supporting zero-shot cases, where the linker has never seen the entity before. Those zero-shot capabilities help to mitigate the problem of rare and expensive training data that is a common issue in the medical domain.

    @inproceedings{oberhauser-etal-2020-trainx,
    title = "{T}rain{X} {--} Named Entity Linking with Active Sampling and Bi-Encoders",
    author = {Oberhauser, Tom and
    Bischoff, Tim and
    Brendel, Karl and
    Menke, Maluna and
    Klatt, Tobias and
    Siu, Amy and
    Gers, Felix Alexander and
    L{\"o}ser, Alexander},
    booktitle = "Proceedings of the 28th International Conference on Computational Linguistics: System Demonstrations",
    month = dec,
    year = "2020",
    address = "Barcelona, Spain (Online)",
    publisher = "International Committee on Computational Linguistics (ICCL)",
    url = "https://www.aclweb.org/anthology/2020.coling-demos.12",
    doi = "10.18653/v1/2020.coling-demos.12",
    pages = "64--69",
    abstract = "We demonstrate TrainX, a system for Named Entity Linking for medical experts. It combines state-of-the-art entity recognition and linking architectures, such as Flair and fine-tuned Bi-Encoders based on BERT, with an easy-to-use interface for healthcare professionals. We support medical experts in annotating training data by using active sampling strategies to forward informative samples to the annotator. We demonstrate that our model is capable of linking against large knowledge bases, such as UMLS (3.6 million entities), and supporting zero-shot cases, where the linker has never seen the entity before. Those zero-shot capabilities help to mitigate the problem of rare and expensive training data that is a common issue in the medical domain.",
    }

  • R. Schneider, T. Oberhauser, P. Grundmann, F. A. Gers, A. Löser, and S. Staab, „Is Language Modeling Enough? Evaluating Effective Embedding Combinations,“ in Proceedings of the 12th Language Resources and Evaluation Conference., Marseille, France, 2020.
    [BibTeX]
    @inproceedings{schneider2020language,
    title = {Is Language Modeling Enough? Evaluating Effective Embedding Combinations},
    booktitle = {Proceedings of the 12th Language Resources and Evaluation Conference.},
    author = {Schneider, Rudolf and Oberhauser, Tom and Grundmann, Paul and Gers, Felix Alexander and Löser, Alexander and Staab, Steffen},
    year = {2020},
    month = May,
    volume = {12},
    publisher = {ELRA},
    address = {Marseille, France}
    }

  • H. Zhang, L. Hennig, C. Alt, C. Hu, Y. Meng, and C. Wang, „Bootstrapping Named Entity Recognition in E-Commerce with Positive Unlabeled Learning,“ in Proceedings of the Third Workshop on e-Commerce and NLP, Seattle, Washington, 2020.
    [BibTeX] [Abstract]

    Named Entity Recognition (NER) in domains like e-commerce is an understudied problem due to the lack of annotated datasets. Recognizing novel entity types in this domain, such as products, components, and attributes, is challenging because of their linguistic complexity and the low coverage of existing knowledge resources. To address this problem, we present a bootstrapped positive-unlabeled learning algorithm that integrates domain-specific linguistic features to quickly and efficiently expand the seed dictionary. The model achieves an average F1 score of 72.02\% on a novel dataset of product descriptions, an improvement of 3.6{\textbackslash}\% over a baseline BiLSTM classifier, and in particular exhibits better recall (4.96\% on average)

    @inproceedings{zhang_bootstrapping_2020,
    address = {Seattle, Washington},
    title = {Bootstrapping {Named} {Entity} {Recognition} in {E}-{Commerce} with {Positive} {Unlabeled} {Learning}},
    abstract = {Named Entity Recognition (NER) in domains like e-commerce is an understudied problem due to the lack of annotated datasets. Recognizing novel entity types in this domain, such as products, components, and attributes, is challenging because of their linguistic complexity and the low coverage of existing knowledge resources. To address this problem, we present a bootstrapped positive-unlabeled learning algorithm that integrates domain-specific linguistic features to quickly and efficiently expand the seed dictionary. The model achieves an average F1 score of 72.02\% on a novel dataset of product descriptions, an improvement of 3.6{\textbackslash}\% over a baseline BiLSTM classifier, and in particular exhibits better recall (4.96\% on average)},
    booktitle = {Proceedings of the {Third} {Workshop} on e-{Commerce} and {NLP}},
    author = {Zhang, Hanchu and Hennig, Leonhard and Alt, Christoph and Hu, Changjian and Meng, Yao and Wang, Chao},
    year = {2020}
    }

2019

  • B. van Aken, B. Winter, A. Löser, and F. A. Gers, „How Does BERT Answer Questions?,“ Proceedings of the 28th ACM International Conference on Information and Knowledge Management – CIKM ’19, 2019.
    [BibTeX]
    @article{van_Aken_2019,
    title={How Does BERT Answer Questions?},
    journal={Proceedings of the 28th ACM International Conference on Information and Knowledge Management - CIKM ’19},
    publisher={ACM Press},
    author={van Aken, Betty and Winter, Benjamin and Löser, Alexander and Gers, Felix A.},
    year={2019}
    }

  • C. Alt, M. Hübner, and L. Hennig, „Fine-tuning Pre-Trained Transformer Language Models to Distantly Supervised Relation Extraction,“ in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, 2019, p. 1388–1398.
    [BibTeX] [Abstract] [Download PDF]

    Distantly supervised relation extraction is widely used to extract relational facts from text, but suffers from noisy labels. Current relation extraction methods try to alleviate the noise by multi-instance learning and by providing supporting linguistic and contextual information to more efficiently guide the relation classification. While achieving state-of-the-art results, we observed these models to be biased towards recognizing a limited set of relations with high precision, while ignoring those in the long tail. To address this gap, we utilize a pre-trained language model, the OpenAI Generative Pre-trained Transformer (GPT) (Radford et al., 2018). The GPT and similar models have been shown to capture semantic and syntactic features, and also a notable amount of “common-sense” knowledge, which we hypothesize are important features for recognizing a more diverse set of relations. By extending the GPT to the distantly supervised setting, and fine-tuning it on the NYT10 dataset, we show that it predicts a larger set of distinct relation types with high confidence. Manual and automated evaluation of our model shows that it achieves a state-of-the-art AUC score of 0.422 on the NYT10 dataset, and performs especially well at higher recall levels.

    @inproceedings{alt_fine-tuning_2019,
    address = {Florence, Italy},
    title = {Fine-tuning {Pre}-{Trained} {Transformer} {Language} {Models} to {Distantly} {Supervised} {Relation} {Extraction}},
    url = {https://www.aclweb.org/anthology/P19-1134},
    abstract = {Distantly supervised relation extraction is widely used to extract relational facts from text, but suffers from noisy labels. Current relation extraction methods try to alleviate the noise by multi-instance learning and by providing supporting linguistic and contextual information to more efficiently guide the relation classification. While achieving state-of-the-art results, we observed these models to be biased towards recognizing a limited set of relations with high precision, while ignoring those in the long tail. To address this gap, we utilize a pre-trained language model, the OpenAI Generative Pre-trained Transformer (GPT) (Radford et al., 2018). The GPT and similar models have been shown to capture semantic and syntactic features, and also a notable amount of “common-sense” knowledge, which we hypothesize are important features for recognizing a more diverse set of relations. By extending the GPT to the distantly supervised setting, and fine-tuning it on the NYT10 dataset, we show that it predicts a larger set of distinct relation types with high confidence. Manual and automated evaluation of our model shows that it achieves a state-of-the-art AUC score of 0.422 on the NYT10 dataset, and performs especially well at higher recall levels.},
    booktitle = {Proceedings of the 57th {Annual} {Meeting} of the {Association} for {Computational} {Linguistics}},
    publisher = {Association for Computational Linguistics},
    author = {Alt, Christoph and Hübner, Marc and Hennig, Leonhard},
    month = jul,
    year = {2019},
    pages = {1388--1398}
    }

  • C. Alt, M. Hübner, and L. Hennig, „Improving Relation Extraction by Pre-trained Language Representations,“ in Proceedings of AKBC 2019, Amherst, Massachusetts, 2019, p. 1–18.
    [BibTeX] [Abstract] [Download PDF]

    Current state-of-the-art relation extraction methods typically rely on a set of lexical, syntactic, and semantic features, explicitly computed in a pre-processing step. Training feature extraction models requires additional annotated language resources, which severely restricts the applicability and portability of relation extraction to novel languages. Similarly, pre-processing introduces an additional source of error. To address these limitations, we introduce TRE, a Transformer for Relation Extraction, extending the OpenAI Generative Pre-trained Transformer [Radford et al., 2018]. Unlike previous relation extraction models, TRE uses pre-trained deep language representations instead of explicit linguistic features to inform the relation classification and combines it with the self-attentive Transformer architecture to effectively model long-range dependencies between entity mentions. TRE allows us to learn implicit linguistic features solely from plain text corpora by unsupervised pre-training, before fine-tuning the learned language representations on the relation extraction task. TRE obtains a new state-of-the-art result on the TACRED and SemEval 2010 Task 8 datasets, achieving a test F1 of 67.4 and 87.1, respectively. Furthermore, we observe a significant increase in sample efficiency. With only 20\% of the training examples, TRE matches the performance of our baselines and our model trained from scratch on 100\% of the TACRED dataset. We open-source our trained models, experiments, and source code.

    @inproceedings{alt_improving_2019,
    address = {Amherst, Massachusetts},
    title = {Improving {Relation} {Extraction} by {Pre}-trained {Language} {Representations}},
    url = {https://openreview.net/forum?id=BJgrxbqp67},
    abstract = {Current state-of-the-art relation extraction methods typically rely on a set of lexical, syntactic, and semantic features, explicitly computed in a pre-processing step. Training feature extraction models requires additional annotated language resources, which severely restricts the applicability and portability of relation extraction to novel languages. Similarly, pre-processing introduces an additional source of error. To address these limitations, we introduce TRE, a Transformer for Relation Extraction, extending the OpenAI Generative Pre-trained Transformer [Radford et al., 2018]. Unlike previous relation extraction models, TRE uses pre-trained deep language representations instead of explicit linguistic features to inform the relation classification and combines it with the self-attentive Transformer architecture to effectively model long-range dependencies between entity mentions. TRE allows us to learn implicit linguistic features solely from plain text corpora by unsupervised pre-training, before fine-tuning the learned language representations on the relation extraction task. TRE obtains a new state-of-the-art result on the TACRED and SemEval 2010 Task 8 datasets, achieving a test F1 of 67.4 and 87.1, respectively. Furthermore, we observe a significant increase in sample efficiency. With only 20\% of the training examples, TRE matches the performance of our baselines and our model trained from scratch on 100\% of the TACRED dataset. We open-source our trained models, experiments, and source code.},
    booktitle = {Proceedings of {AKBC} 2019},
    author = {Alt, Christoph and Hübner, Marc and Hennig, Leonhard},
    year = {2019},
    pages = {1--18}
    }

  • J. Frey, M. Hofer, D. Obraczka, J. Lehmann, and S. Hellmann, „DBpedia FlexiFusion the Best of Wikipedia \textgreater Wikidata \textgreater Your Data,“ in The Semantic Web – ISWC 2019 – 18th International Semantic Web Conference, Auckland, New Zealand, October 26-30, 2019, Proceedings, Part II, 2019, p. 96–112. doi:10.1007/978-3-030-30796-7_7
    [BibTeX] [Download PDF]
    @inproceedings{DBLP:conf/semweb/FreyHO0H19,
    author = {Johannes Frey and
    Marvin Hofer and
    Daniel Obraczka and
    Jens Lehmann and
    Sebastian Hellmann},
    editor = {Chiara Ghidini and
    Olaf Hartig and
    Maria Maleshkova and
    Vojtech Sv{\'{a}}tek and
    Isabel F. Cruz and
    Aidan Hogan and
    Jie Song and
    Maxime Lefran{\c{c}}ois and
    Fabien Gandon},
    title = {DBpedia FlexiFusion the Best of Wikipedia {\textgreater} Wikidata
    {\textgreater} Your Data},
    booktitle = {The Semantic Web - {ISWC} 2019 - 18th International Semantic Web Conference,
    Auckland, New Zealand, October 26-30, 2019, Proceedings, Part {II}},
    series = {Lecture Notes in Computer Science},
    volume = {11779},
    pages = {96--112},
    publisher = {Springer},
    year = {2019},
    url = {https://doi.org/10.1007/978-3-030-30796-7\_7},
    doi = {10.1007/978-3-030-30796-7\_7},
    timestamp = {Fri, 09 Apr 2021 18:46:51 +0200},
    biburl = {https://dblp.org/rec/conf/semweb/FreyHO0H19.bib},
    bibsource = {dblp computer science bibliography, https://dblp.org}
    }

  • P. Gupta, Y. Chaudhary, and H. Schütze, „BioNLP-OST 2019 RDoC Tasks: Multi-grain Neural Relevance Ranking Using Topics and Attention Based Query-Document-Sentence Interactions,“ in Proceedings of The 5th Workshop on BioNLP Open Shared Tasks, Hong Kong, China, 2019, p. 227–236. doi:10.18653/v1/D19-5730
    [BibTeX] [Download PDF]
    @inproceedings{gupta-etal-2019-bionlp,
    title = "{B}io{NLP}-{OST} 2019 {RD}o{C} Tasks: Multi-grain Neural Relevance Ranking Using Topics and Attention Based Query-Document-Sentence Interactions",
    author = {Gupta, Pankaj and
    Chaudhary, Yatin and
    Sch{\"u}tze, Hinrich},
    booktitle = "Proceedings of The 5th Workshop on BioNLP Open Shared Tasks",
    month = nov,
    year = "2019",
    address = "Hong Kong, China",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D19-5730",
    doi = "10.18653/v1/D19-5730",
    pages = "227--236"
    }

  • P. Gupta, K. Saxena, U. Yaseen, T. Runkler, and H. Schütze, „Neural Architectures for Fine-Grained Propaganda Detection in News,“ in Proceedings of the Second Workshop on Natural Language Processing for Internet Freedom: Censorship, Disinformation, and Propaganda, Hong Kong, China, 2019, p. 92–97.
    [BibTeX] [Download PDF]
    @inproceedings{gupta-etal-2019-neural,
    title = "Neural Architectures for Fine-Grained Propaganda Detection in News",
    author = {Gupta, Pankaj and
    Saxena, Khushbu and
    Yaseen, Usama and
    Runkler, Thomas and
    Sch{\"u}tze, Hinrich},
    booktitle = "Proceedings of the Second Workshop on Natural Language Processing for Internet Freedom: Censorship, Disinformation, and Propaganda",
    month = nov,
    year = "2019",
    address = "Hong Kong, China",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D19-5012",
    pages = "92--97"
    }

  • P. Gupta, U. Yaseen, and H. Schütze, „Linguistically Informed Relation Extraction and Neural Architectures for Nested Named Entity Recognition in BioNLP-OST 2019,“ in Proceedings of The 5th Workshop on BioNLP Open Shared Tasks, Hong Kong, China, 2019, p. 132–142.
    [BibTeX] [Download PDF]
    @inproceedings{gupta-etal-2019-linguistically,
    title = "Linguistically Informed Relation Extraction and Neural Architectures for Nested Named Entity Recognition in {B}io{NLP}-{OST} 2019",
    author = {Gupta, Pankaj and
    Yaseen, Usama and
    Sch{\"u}tze, Hinrich},
    booktitle = "Proceedings of The 5th Workshop on BioNLP Open Shared Tasks",
    month = nov,
    year = "2019",
    address = "Hong Kong, China",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D19-5720",
    pages = "132--142"
    }