Dataset of "Knowledge-based Sense Disambiguation of Multiword Expressions in Requirements Documents"

This is the dataset used in the paper "Knowledge-based Sense Disambiguation of Multiword Expressions in Requirements Documents" at AIRE'21

In this paper, we explore the use of a multiword expression detection in combination with a knowledge-based word sense disambiguation to disambiguate expressions in requirements documents.

The dataset comprises a gold standard for multiword expression detection and sense disambiguation for Wikipedia and WordNet 3.1.

It covers 18 projects: CM1, EBT and GANTT as well as the 15 projects of the NFR dataset.

File format

We use a tab-separated version of the DiMSUM file format and extended it with sense information.

The nine original DiMSUM tab-separated columns:

1. token offset

2. word

3. lowercase lemma

4. POS

5. MWE tag

6. offset of parent token (i.e. previous token in the same MWE), if applicable

7. strength level encoded in the tag, if applicable. Currently not used

8. supersense label, Currently not used

9. sentence ID

and the two further columns for sense information:

10. Wikipedia article name

11. WordNet 3.1 synset

The last two columns might end with .1 or .0 indicating that the sense is a fully applicable or partial sense of a multiword expression.

Attribution (of datasets used)

The NFR Dataset can be attributed to Jane Cleland-Huang.
Jane Cleland-Huang, Sepideh Mazrouee, Huang Liguo, & Dan Port. (2007). nfr [Data set]. Zenodo. Available: http://doi.org/10.5281/zenodo.268542

The CM1, EBT and GANTT datasets were retrieved from the Center of Excellence for Software & Systems Traceability (CoEST) http://coest.org/