Table 2.
Comparison of shared terms among dictionaries
| Harvard-IV | LM | L1 | L2 | VADER | |
|---|---|---|---|---|---|
| Harvard-IV | 3642 | 597 | 312 | 235 | 1291 |
| (100%) | (97.3%) | (75.6%) | (93.2%) | (97.7%) | |
| LM | 2709 | 191 | 217 | 870 | |
| (100%) | (96.9%) | (99.5%) | (97.6%) | ||
| L1 | 8000 | 805 | 418 | ||
| (100%) | (99.9%) | (81.6%) | |||
| L2 | 1311 | 373 | |||
| (100%) | (97.3%) | ||||
| VADER | 7517 | ||||
| (100%) |
Note: This table summarizes commonalities and differences among five publicly available sentiment dictionaries, i.e., Harvard-IV, LM (Loughran and McDonald 2011), the two dictionaries introduced by Renault (2017) (L1 and L2), and VADER (Hutto and Gilbert 2014). More precisely, the table shows the number of common terms occurring in two dictionaries. The diagonal elements report the total number of words in each dictionary and the off-diagonal elements report the number of shared terms between two dictionaries. The share of words having the same sentiment connotation (positive or negative) in both dictionaries is reported in parentheses below the respective number of common terms