Rule-Based PoS Tagging

Rule-Based PoS Tagging

• a type of Part-of-Speech (PoS) Tagging utilizing rules
• rules cannot be auto-learned by feeding it training data (requires manual creation)

Process

given:

• dictionary of words with their corresponding PoS Tags (a single word may have multiple tags)
• set of rules that selectively remove PoS Tags
• input sentence to be PoS Tagged

steps:

• according to the dictionary, assign each word of the input sentence all possible PoS Tags
• according to the set of rules, remove tags from word of the sentence, until each word has exactly one PoS Tag

Example

given input sentence:

• She promised to back the bill

according to the dictionary, we assign each word their PoS Tags:

according to the set of rules, we remove tags

say we come across a rule like below:

• Eliminate VBN if VBD is an option when VBN|VBD follows “<start> PRP”

we then remove VBN from under the promised word like so:

we continue until each word has exactly one PoS Tag