Questions for self-monitoring

• What's the benefit of using dense document representations for IR?
• How can dense vectors be generated for IR?

• What are the major paradigms for QA?

• Which tasks have to be solved in IR-based (or open domain) QA?
• How do data sets for the answer extraction look like?
• How can the task of answer extraction be solved?

• What's the purpose of entity linking?
• Which tasks have to be solved for entity linking?
• Which approaches have been pursued for entity linking?
• What are the advantages of knowledge-based QA?

• What are the drawbacks of using a language model as QA system?