Figure 1
The flow of PMML.
Table 1
Matching patterns of phases.
| Pattern | Collocations | Example |
|---|---|---|
| E1 | Adjective + noun | Intimate service |
| E2 | Adjective + adverb | High enough |
| E3 | Adjective + verb | Efficient operation |
| E4 | Noun + adjective | Allocation perfect |
| E5 | Adverb + adjective | Especially like |
| E6 | Negative word + verb | Not satisfactory |
| E7 | Negative word + adjective | Not beautiful |
| E8 | Negative word + adverb + verb | Not very like |
| E9 | Verb + adjective | Consume fast |
| E10 | Adverb + negative word + verb | Not happy |
Table 2
The classification situation.
| Actual number of texts which belong to the class | Actual number of texts which do not belong to the class | |
|---|---|---|
| Number of texts which are identified as the class by the classifier | w | x |
| Number of texts which are identified not as the class by the classifier | y | z |
Table 3
The comparison of classification performance of N-Gram.
| Feature | UniGrams | BiGrams | TriGrams |
|---|---|---|---|
| PP | 62.72% | 67.62% | 65.85% |
| PR | 54.27% | 47.55% | 43.72% |
| NP | 60.38% | 63.82% | 59.32% |
| NR | 72.54% | 76.41% | 64.78% |