Universität Wien FIND

Due to the COVID-19 pandemic, changes to courses and exams may be necessary at short notice. Inform yourself about the current status on u:find and check your e-mails regularly. Registration is mandatory for courses and exams. Wearing a FFP2 face mask and a valid evidence of being tested, vaccinated or have recovered from an infection are mandatory on site.

Please read the information on studieren.univie.ac.at/en/info.

136041 SE Weakly Supervised Machine Learning (2021S)

Continuous assessment of course work

Registration/Deregistration

Note: The time of your registration within the registration period has no effect on the allocation of places (no first come, first served).

Details

max. 25 participants
Language: German, English

Lecturers

Classes (iCal) - next class is marked with N

BigBlueButton link for the first session:
https://bbb.cs.univie.ac.at/b/ben-chx-4jk-opw

Thursday 04.03. 09:45 - 11:15 Digital
Thursday 11.03. 09:45 - 11:15 Digital
Thursday 18.03. 09:45 - 11:15 Digital
Thursday 25.03. 09:45 - 11:15 Digital
Thursday 15.04. 09:45 - 11:15 Digital
Thursday 22.04. 09:45 - 11:15 Digital
Thursday 29.04. 09:45 - 11:15 Digital
Thursday 06.05. 09:45 - 11:15 Digital
Thursday 20.05. 09:45 - 11:15 Digital
Thursday 27.05. 09:45 - 11:15 Digital
Thursday 10.06. 09:45 - 11:15 Digital
Thursday 17.06. 09:45 - 11:15 Digital
Thursday 24.06. 09:45 - 11:15 Digital

Information

Aims, contents and method of the course

In this seminar, participants will read, present and discuss recent papers on weakly supervised machine learning for natural language processing.
Weak supervision allows the use of prior knowledge so that machine learning models can be trained even if there is no annotated training data available.

Assessment and permitted materials

Participants will have to present one topic from the list in the seminar, the presentation should be roughly 25 minutes (hard limits: min. 20 minutes, max. 30 minutes). The presentation is followed by a QA-session and discussion. Participants will also have to submit a written report (15-20 pages, exact requirements TBD), describing the main contents of the presented paper and putting it in a wider context.

Please send an email to nlp.datamining@univie.ac.at including a selection of *5 topics* from the list below, and indicate your *study program* (Computer Science, Digital Humanities, ...). You will be assigned one topic from your selection (for your presentation and report). For additional two topics (also from your selection, but presented by somebody else) you will have to prepare some questions that can get a discussion started.

Minimum requirements and assessment criteria

Your presentation will account for 45% of the grade, participation in discussions for 10%, and the written report for 45%.

Examination topics

.

Reading list

*Classics, non-neural approaches*
Craven et al. (1999). Constructing biological knowledge bases by extracting information from text sources.

Mintz et al. (2009). Distant supervision for relation extraction without labeled data.

Riedel et al. (2010). Modeling relations and their mentions without labeled text.

Surdeanu et al. (2012). Multi-instance multi-label learning for relation extraction.

Ritter et al. (2013). Modeling missing data in distant supervision for information extraction.

*Universal Schemas*
Riedel et al. (2013). Relation extraction with matrix factorization and universal schemas.

Verga et al. (2015). Multilingual relation extraction using compositional universal schema.

*Label Denoising*
"Topic Models"
Griffiths et al. (2004). Finding scientific topics.
Alfonseca, E., Filippova, K., Delort, J. Y., & Garrido, G. (2012). Pattern Learning for Relation Extraction with Hierarchical Topic Models.

Lin et al. (2016). Neural relation extraction with selective attention over instances.

Luo et al. (2017). Learning with noise: Enhance distantly supervised relation extraction with dynamic transition matrix.

"Cleanlab"
Northcutt et al. (2019). Confident learning: Estimating uncertainty in dataset labels.
https://pypi.org/project/cleanlab/

"Snorkel"
Stephen et al. (2017). Learning the Structure of Generative Models without Labeled Data.
https://www.snorkel.org/

Feng et al. (2018). Reinforcement learning for relation classification from noisy data.

Dehghani et al. (2017). Fidelity-weighted learning.

Zheng et al. (2019). Meta label correction for learning with weak supervision.

Awasthi et al. (2020). Learning from rules generalizing labeled exemplars.

*Adversarial methods*
Miyato et al. (2016). Adversarial training methods for semi-supervised text classification.
Wu et al. (2017). Adversarial training for relation extraction.

Zeng et al. (2018). Adversarial learning for distant supervised relation extraction.

Qin et al. (2018). Dsgan: Generative adversarial training for distant supervision relation extraction.

* Applications *

"Named Entity Recognition"
Rehbein et al. (2017). Detecting annotation noise in automatically labelled data.
Hovy et al. (2013). Learning whom to trust with MACE.

Lison et al. (2020). Named entity recognition without labelled data: A weak supervision approach.

Liang et al. (2020). Bond: Bert-assisted open-domain named entity recognition with distant supervision.

"Low-resource Named Entity Recognition"
Hedderich et al. (2020). Transfer Learning and Distant Supervision for Multilingual Transformer Models: A Study on African Languages.
Hedderich and Klakow (2018). Training a neural network in a low-resource settingon automatically annotated noisy data.
Lv et al. (2020). Matrix smoothing: A regularization for DNN with transition matrix under noisy labels.

Cao et al. (2020). Unsupervised Parsing via Constituency Tests.

"Sentiment"
Turney (2002). Thumbs up or thumbs down? Semantic orientation applied to unsupervised classification of reviews.
Go et al. (2009). Twitter sentiment classification using distant supervision.
Hogenboom et al. (2013). Exploiting emoticons in sentiment analysis.

Guo et al. (2011). A weakly-supervised approach to argumentative zoning of scientific documents.

Rutherford et al. (2015). Improving the inference of implicit discourse relations via classifying explicit discourse connectives.

Wallace et al. (2016). Extracting PICO sentences from clinical trial reports using supervised distant supervision.

Keith et al. (2017). Identifying civilians killed by police with distantly supervised entity-event extraction.

Li et al. (2014). Weakly supervised user profile extraction from twitter.

Lee et al. (2019). Latent retrieval for weakly supervised open domain question answering.

Hashimoto (2019). Weakly supervised multilingual causality extraction from wikipedia.

Association in the course directory

S-DH (Cluster I: Language and Literature)

Last modified: We 03.03.2021 13:08