Table 2. Transferability r t example.
Article | Japanese | Thai |
Fever | ||
Chills | ||
Headache | ||
Influenza |
This table shows simplified models for influenza in two locations: Japan, where Japanese is spoken, and Thailand, where Thai is spoken. The Japanese model yielded correlations for Japanese versions of the articles “Fever”, “Chills”, “Headache”, and “Influenza” of 0.23, 0.59, −0.10, and 0.85, respectively. The Thai model yielded correlations of 0.21, 0.15, and 0.77 for “Fever”, “Headache”, and “Influenza”, respectively. Note that the article “Chills” is not currently present in the Thai Wikipedia. Therefore, the correlation vectors are and for the two languages. The meta-correlation, r t, for these two vectors, which provides a gross estimate of how similar the models are, is 0.97. Extending this computation to the full models yields r t = 0.81, as noted below in Table 4.