SemEval 2017 – Task 4B

Goal: Topic-Based Sentiment Classification

Technology: Distant-trained Convolutional Neural Network (CNN)

Result: Recall 84.6 (best score: 88.2)

Rank: 4th place out of 23

Paper:

TopicThunder at SemEval-2017 Task 4: Sentiment Classification Using a Convolutional Neural Network with Distant Supervision

Simon Müller, Tobias Huonder, Jan Deriu, and Mark Cieliebak

SemEval 2016 – Task 4

Goal: Message-level sentiment classification of English micro-blog messages from Twitter

Technology: Distant-trained, 2-layer Convolutional Neural Network (CNN) 

Result: F1-Score 63.3 (was best score)

Rank: 1st place out of 34

Paper: SwissCheese at SemEval-2016 Task 4: Sentiment Classification Using an Ensemble of Convolutional Neural Networks with Distant Supervision

Jan Deriu, Maurice Gonzenbach, Fatih Uzdilli, Aurelien Lucchi, Valeria De Luca, and Martin Jaggi

EVALITA 2016 – SENTIPOLC – Task 2

Goal: Message Level Sentiment Classification for Italian Tweets

Technology: Multi-task trained Convolutional Neural Network (CNN) with weakly labelled distant learning phase

Result: F1-Score 68.28 (was best score)

Rank: 1st place out of 26

Paper:

Sentiment Detection using Convolutional Neural Networks with Multi-Task Training and Distant Supervision on Italian Tweets

Jan Deriu and Mark Cieliebak

SemEval 2015 – Task 10

Goal: Message Level Sentiment Classification for Tweets

Technology: Meta-Classifier on several flipout-regularized Support Vector Machines

Result: F1-Score 62.61 (best score: 64.84)

Rank: 8th place out of 40

Paper:

Swiss-Chocolate: Combining Flipout Regularization and Random Forests with Artificially Built Subsystems to Boost Text-Classification for Sentiment

Fatih Uzdilli, Martin Jaggi, Dominic Egger, Pascal Julmy, Leon Derczynski and Mark Cieliebak

SemEval 2014 – Task 9B

Goal: Message Level Sentiment Classification for Tweets

Technology: Regularized Support Vector Machine (SVM) with hand-crafted features.

Result: F1-Score 67.54 (best score: 70.96)

Rank: 8th place out of 50

Paper:

Swiss-Chocolate: Sentiment Detection using Sparse SVMs and Part-Of-Speech n-Grams

Martin Jaggi, Fatih Uzdilli and Mark Cieliebak

SemEval 2014 – Task 9B

Goal: Message Level Sentiment classification for Tweets

Technology: Combining 12 sentiment classification systems with a meta-classifier

Result: F1-Score 66.79 (best score: 70.96)

Rank: 12th place out of 50
Paper:

JOINT_FORCES: Unite Competing Sentiment Classifiers with Random Forest

Oliver Dürr, Fatih Uzdilli, and Mark Cieliebak

Coling 2018 – VarDial

Goal: German Dialect Identification

Technology: Meta SVM with multiple Feature Extraction Methods

Results: F-1 macro 64.6 % (best score 68.6%)

Rank: 2nd out of 8

Paper:

Twist Bytes – German Dialect Identification with Data Mining Optimization
Fernando Benites, Ralf Grubenmann, Pius von Däniken, Dirk von Grünigen, Jan Deriu and Mark Cieliebak

CLEF 2017 – PAN

Goal: Gender and Language Variety Detection from Tweets

Technology: Bi-directional Recurrent Neural Network (RNN) with attention mechanism

Results:

Gender Classification: Accuracy 75.31% (best score: 82.53)

Language Variety: Accuracy 85.22% (best score: 91.84)

Rank: 12th place out of 22

Paper:

Author Profiling with Bidirectional RNNs using Attention with GRUs – Notebook for PAN at CLEF 2017

Don Kodiyan, Florin Hardegger, Stephan Neuhaus, and Mark Cieliebak

CAp 2017

Goal: Named Entity Recognition on French Tweets

Technology: Deep learning with partially annotated data

Result: F-score 50.05 (best score: 58.89)

Rank: 5th place out of 8

Paper:

Swiss Chocolate at CAp 2017 NER Challenge: Partially Annotated Data and Transfer Learning

Nicole Falkner, Stefano Dolce, Pius von Däniken, and Mark Cieliebak

WNUT 2017

Goal: Named Entity Recognition on Tweets

Technology:

Result: F1-Score 40.78 (best score: 41.86)

Rank: 2nd place out of 7

Paper:

Transfer Learning and Sentence Level Features for Named Entity Recognition on Tweets

Pius von Däniken and Mark Cieliebak.

PSB 2016 – Task 1

Goal: Detect Mentions of Adverse Drug Reactions (ADR) in Tweets

Technology: Adaption of a feature-based sentiment classifier to ADR

Result: F1-Score 31.74 (best score: 41.95)

Rank: 5th place out of 8

Paper:

Adverse Drug Reaction Detection using an Adapted Sentiment Classifier

Dominic Egger, Fatih Uzdilli, and Mark Cieliebak

SemEval 2017 – Task 3A

Goal: Finding Relevant Responses to never-before seen Questions

Technology: Siamese Convolutional Neural Network (CNN) with attention mechanism

Result: MAP-Score 86.24 (best score: 88.43)

Rank: 7th place out of 13

Paper:

SwissAlps at SemEval-2017 Task 3: Attention-based Convolutional Neural

Network for Community Question Answering

Jan Deriu and Mark Cieliebak

E2E NLG Challenge 2017

Goal: Generate Restaurant Reviews from Structured Data

Technology: Character-based Semantically Controlled Long Short-term Memory Network (SC-LSTM) with first-word control

Rank: 2nd rank out of 4 clusters

Paper:

End-to-End Trainable System for Enhancing Diversity in Natural Language Generation

Jan Deriu and Mark Cieliebak