There are various challenges that are interesting to the Semantic Computing group. Typically, a challenge involves a dataset and a task and a description of an evaluation procedure.

Within a thesis (every topic can be framed as a B.Sc. thesis or a M.Sc. thesis), one can approach the problem, but since the timing of a thesis may not always fit to the deadline of a challenge, we do not expect students to actually participate in (i.e., submit results to) a challenge.

Generating Text from RDF Data - The WebNLG Challenge

The task consists in verbalizing RDF triples. For instance, given the 3 DBPedia triples shown in (a), the aim is to generate a text such as (b).

a. (John_E_Blaha birthDate 1942_08_26) (John_E_Blaha birthPlace San_Antonio) (John_E_Blaha occupation Fighter_pilot)
b. John E Blaha, born in San Antonio on 1942-08-26, worked as a fighter pilot.

Further information available at webnlg.loria.fr/pages/challenge.html

This thesis would be supervized by Basil Ell

TextExt - DBpedia Open Extraction Challenge

Participants are asked to submit their engines that extract facts and knowledge from Wikipedia article texts to dramatically broaden and deepen the amount of structured DBpedia/Wikipedia data

Further information available at https://wiki.dbpedia.org/textext

This thesis would be supervized by Basil Ell

Fact Extraction and Verification (fever.ai)

Given a factual claim involving one or more entities (resolvable to Wikipedia pages), the system must extract textual evidence (sets of sentences from Wikipedia pages) that support or refute the claim.

Further information available at http://fever.ai

This thesis would be supervized by Basil Ell

Natural Language Decathlon

While Natural Language Processing (NLP) architectures are mostly created with one specific task in mind, the need for NLP models that perform well on multiple tasks is mostly neglected. This challenge spans ten tasks: question answering, machine translation, summarization, natural language inference, sentiment analysis, semantic role labeling, relation extraction, goal-oriented dialogue, database query generation, and pronoun resolution. The goal of the Decathlon is to explore models that generalize to all ten tasks and investigate how such models differ from those trained for single tasks.

Further information available at https://einstein.ai/research/the-natural-language-decathlon

This thesis would be supervized by Frank Grimm and Basil Ell