Table 3. Pre- and post-test scores for dictionary test sets.
Correct (max. 1pt. deviation from human assessment) | Partially correct (2pt. deviation from human assessment) | Incorrect* | ||
---|---|---|---|---|
DK | Pre-test | 46 | 46 | 8 |
Post-test | 70 | 26 | 4 | |
DE | Pre-test | 58 | 37 | 5 |
Post-test | 73 | 25 | 2 | |
NL | Pre-test | 47 | 42 | 11 |
Post-test | 57 | 35 | 8 | |
SE | Pre-test | 67 | 27 | 6 |
Post-test | 78 | 20 | 2 |
*A score is deemed incorrect, either if it indicates sentiment where there is none, indicates no sentiment where there is sentiment, or if the strength of the sentiment indicated differs with more than one point compared to the human estimation (e.g. if SentiStrength indicates a positive sentiment of 2, but the human estimation is 4). If one score indicates a very strong sentiment where there is none, the tweet is added to the third column (incorrect).