This page lists the rules used by POME (Pattern-based Opinion MinEr). The rules are categorized into seven different aspects: community, compatibility, documentation, functional, performance, reliability, usability. In the end of this page, we also include the extra rules we used during training to reduce the false positive.

For each rule, blue blocks are the tokens representing libraries, green blocks are vocabularies extracted during labeling, yellow blocks are pos tags of tokens. In the end, there is a badge representing the default sentiment of the rule (which can be different from the sentiment of the example sentence if negation is detected in the sentence). The sentiment will be reversed if the part of sentence containing the pattern are negated. Black blocks represent that negation is detected.

The number l on the left side of each block is the index. Here we denote the block with index i by Bi, and denote the POS tag of Bi by Pi. The number r on the right side of the block represents that Br should be the first parent node of Bl in the parsed dependency tree which has the POS tag of Pi.

Hovering over the blocks allows users to check the actual meaning of these blocks.