-
Notifications
You must be signed in to change notification settings - Fork 4
Results Using Word Lists A & B (Lancs) v1.1
Version 1.1 of the List Analyser was updated to include ERRT measurements. For verification we attempted to reproduce the results for the Porter stemmer as documented in a paper produced at Lancaster University by Rob Hooper (2004). A copy of the paper is in the Literature folder of this project.
The Table below shows the results for Word List A reported in Hooper's paper
| Weak Ignored | Weak = Strong | |
|---|---|---|
| Lovins | 95.64 | 90.23 |
| Porter | 83.59 | 67.33 |
| Paice | 71.27 | 54.26 |
The Table below shows the results for Word List B reported in Hooper's paper
| Weak Ignored | Weak = Strong | |
|---|---|---|
| Lovins | 91.79 | 91.47 |
| Porter | 61.66 | 55.08 |
| Paice | 87.97 | 83.13 |
Table 3 shows the results using Word List B (commonwords.txt)for File A and commonwords_Porter.txt for File B using List Analyser 1.1.6916, with the File> Preferences set to Normal.
See Errors in Lancaster University Stemming Evaluation Tool
| Weak Ignored | Weak = Strong | |
|---|---|---|
| Porter | 61.61 | 59.81 |
Table 4 shows the results using Word List B and List Analyser 1.1 for the Porter stemmer with the File> Preferences set to Hooper.
| Weak Ignored | Weak = Strong | |
|---|---|---|
| Porter | 61.61 | 54.99 |