numeralclassification

Numeralclassification is the study and practice of organizing numerals—both digits and words—that encode quantity, rank, or numerical relation. In linguistics, numerals are classified by their functional type: cardinal (one, two, three), ordinal (first, second), multiplicative (double, triple), distributive (each, either), and fractional (half, quarter). Some languages employ numeral classifiers, a system of measure or noun-class markers that accompany numerals when counting or quantifying nouns, as seen in Chinese, Japanese, Korean, Thai, and many Austronesian languages. The position and agreement of numerals also vary cross-linguistically; in some languages numerals precede the noun, in others they follow, and they may interact with adjectives and determiners in complex ways.

In computational linguistics, numeralclassification refers to automatically labeling numerals in text with their semantic category or normalizing them for downstream tasks. Tasks include type tagging (cardinal, ordinal, date, time, money, percentage, unit), unit extraction, and numeric value normalization. Numerals often require disambiguation and contextual cues to resolve scale (million vs thousand) and currency or date semantics.

Common approaches combine rule-based heuristics with statistical or neural methods, using features from surrounding words, punctuation, currency signs, units, and formatting. Challenges include multilingual text, mixed-language numerals, ambiguous forms, and noisy data. Applications span information extraction, question answering, automated report generation, and natural language interfaces.