Heuristic and rule-based knowledge acquisition: classification of numeral strings in text
This paper describes the rule-based classification of numerals and strings that include numerals, composed of a number and semantic unit(s) that indicate a SPEED, NUMBER, or other measure, at three levels: morphological, syntactic, and semantic. The approach employs three interpretation processes: word trigram construction with tokeniser, rule-based processing of number strings, and n-gram based classification. We extracted numeral strings from 378 online newspaper articles, finding that, on average, they comprised about 2.2% of the words in the articles. To manually extract n-gram rules to disambiguate the number strings’ meanings, our approach was trained on 886 numeral strings and tested on the remaining 3251 strings. We implemented two heuristic disambiguation methods based on each category’s frequency statistics collected from the sample data, and precision ratios of both methods were 86.8% and 86.3% respectively. This paper focuses on the acquisition and performance of different types of rules applied to numeral strings classification.