This is an exercise concerning string similarity and lexical analogy based on experimental data from Albright and Hayes (2003).
It is a version of the 'wug test'.
Test.csv contains the Albright and Hayes data, where:
- Orth - orthographic representation of each wug (in present tense)
- Present - phonetic transcription of present tense of wug
- Rating - participants’ average acceptability rating on scale of 1-7 for each wug
- Reg_Past - phonetic form of regular past tense
- Reg_Rating - participants’ average rating for regular past tense form
- Irreg_Past - phonetic form of irregular past tense form of wug
- Irreg_Rating - participants’ average rating for irregular past tense form
- Class - label indicating morphological transformation required to form irregular past tense
Train.csv is a dictionary of over 4,000 English verbs. It contains the following information for each entry:
- orthographic form of present
- phonetic form of present
- orthographic form of past
- phonetic form of past
- morphological transformation class
The analogical model implemented by Albright and Hayes (2003) is contained in analogy.py.
- defines similarity between two words i and j based on Levenshtein distance