Machine Reading Research Task
A Machine Reading Research Task is a computational research task that exposes computation knowledge that facilitates automated readings.
- …
- Counter-Example(s):
- See: Machine Reading Practical Task.
References
2014
- http://www.pgafamilyfoundation.org/MediaLibraries/PGA-Foundations/Documents/PDFs/AllenFoundation-AI-ADI-RFP-March-2014.pdf
- Machine reading is one of the grand challenges of AI. While much current work has focused on entity linking and fact extraction, the next step in machine reading involves the computer constructing richer semantic representations from language that better support automated reasoning.
We are looking for proposals that explore innovative approaches to this challenge. We are particularly interested in proposals that include consideration of background knowledge and an operational use of machine-read knowledge to evaluate progress. A particular application context of interest is machine reading of material required for taking science exams. Examples include on-line text in support of the New York Regents exams (http://www.nysedregents.org/ ), or reading narrative text and answering comprehension questions (e.g.,http://nces.ed.gov/nationsreportcard/itmrlsx/search.aspx?subject=reading).
Lines of inquiry might include:
- How can one extract inference - supporting representations from language?
- How does one "fill in the gaps" (implied facts) in textually presented information?
- What kinds of prior knowledge structures are useful for interpreting language? How might such knowledge structures be acquired?
- How might larger knowledge structures (scripts, frames) that encode expectations be acquired and used to guide reading?
- Machine reading is one of the grand challenges of AI. While much current work has focused on entity linking and fact extraction, the next step in machine reading involves the computer constructing richer semantic representations from language that better support automated reasoning.