Erotetic Reasoning Corpus


Erotetic Reasoning Corpus (ERC) constitutes a data set for research on natural question processing. Intuitively, we are dealing with question processing in a situation when a question is not followed by an answer but with a new question or a strategy of reducing it into auxiliary questions. Usually, such a situation takes place when an agent wants to solve a certain problem (expressed in a form of an initial question) but is not able to reach the solution using his/her own information resources. Thus new data, collected via questioning is necessary.

The corpus consists of the language data collected in the previous studies on the question processing phenomenon. The outcomes of three research projects are employed here. These are: Erotetic Reasoning Test, QuestGen and Mind Maze.



ESS Library and Data

This archive contains:
a) collections of the evaluated erotetic search scenarios: „TeXMain”,”TeXRemovedIC”, „TeXReplacedIC”. Each LaTeX file in each folder contains a diagram of an erotetic search scenario and additional evaluation information,
b) 6 files with the results of quantitative analyses (including the distributions of the criteria values and the outcome of multi-criteria analyses): „quantitative-data-original.txt”, „quantitative-data-original-without-purity.txt”, „quantitative-data-removedIC.txt”, „quantitative-data-removedIC-without-purity.txt”, „quantitative-data-replacedIC.txt”, „quantitative-data-replacedIC-without-purity.txt”,
c) a PDF file („commentary.pdf”) containing a commentary on the results obtained from the experiments.

Please read the [PDF] file to get a clarification of the erotetic search scenarios generation procedure and about the results available in the other files. Also see the readme file to get an explanation of the used naming notation.

FILE (zip archive, 618 KB) [LINK]