International AI success for DIKU’s Information Retrieval Lab
In the last month, the Information Retrieval Lab has had a string of huge international successes in artificial intelligence.
The IR Lab has seven scientific paper accepted at top international AI conferences (ACM SIGIR, ICLR, WWW, EDM) with typical acceptance rates lower than 20%.
- I’m very proud that one of our papers received glowing reviews and was ranked at position 2 out of all submissions. This means that our research, when competing with the biggest international players from both academia and industry, such as Google, Microsoft, Baidu and Deep Mind, is ranked in the top 1%, says associate professor Christina Lioma, who leads the IR Lab.
Winner of international AI competition on fake news detection
Furthermore, researchers from the IR Lab have won a highly competitive international AI competition on fake news detection. They took part in the competition last year too, in two different tasks, and came in first in one task and second in the other. This year they took part in one task and came in first. See this year’s results here.
- In the last 12 months, we have participated in four international AI competitions, and have come 1st in two of them and 2nd in two of them. We are competing with top academics and industry globally, and I’m happy to say, that we have now established Copenhagen as a clear leader within this field, says Christina Lioma.
The seven accepted papers
- Unsupervised Neural Generative Semantic Hashing. Casper Hansen, Christian Hansen, Jakob Grue Simonsen, Stephen Alstrup, Christina Lioma. SIGIR 2019 (ranked <1% of all papers submitted).
- Contextually Propagated Term Weights for Document Representation. Casper Hansen, Christian Hansen, Stephen Alstrup, Jakob Grue Simonsen, Christina Lioma. SIGIR 2019.
- Neural Speed Reading with Structural-Jump-LSTM. Casper Hansen, Christian Hansen, Stephen Alstrup, Jakob Grue Simonsen, Christina Lioma. ICLR 2019.
- Neural Check-Worthiness Ranking with Weak Supervision: Finding Sentences for Fact-Checking. Casper Hansen, Christian Hansen, Stephen Alstrup, Jakob Grue Simonsen, Christina Lioma. WWW 2019.
- Contextual Compositionality Detection with External Knowledge Bases and Word Embeddings. Dongsheng Wang, Qiuchi Li, Lucas Chaves Lima, Jakob Grue Simonsen, Christina Lioma. WWW 2019.
- Modelling End-of-Session Actions in Educational Systems. Christian Hansen, Casper Hansen, Stephen Alstrup, Christina Lioma. EDM 2019.
- Neural Weakly Supervised Check-Worthiness Detection with Contrastive Sampling-Based Ranking Loss. Christian Hansen, Casper Hansen, Jakob Grue Simonsen, Christina Lioma. CLEF 2019 (1st prize in international AI competition).
Related News
Check-worthiness
Click on image for larger version.
Table from the paper Contextual Compositionality Detection with External Knowledge Bases and Word Embeddings. Dongsheng Wang, Qiuchi Li, Lucas Chaves Lima, Jakob Grue Simonsen, Christina Lioma. WWW 2019.
The table highlights both check-worthy and normal sentences by US politicians from the 2016 presidential election. Dark red areas are deemed the most important for the prediction of check-worthiness, and correspond to sections that a journalist or automatic fact-checker should investigate.