Address-Finder

Address Finder is a package for matching textual addresses. Several similarity scoring methods are implemented in this package. The methods implemented in this package are targeting non-uniform textual addresses which may contain misspelling, missing words, etc. The package can be used with Spark.

Textual Address similarity

The following similarity measures are defined in this package.

• LetterPrefixDistance measures the length of common prefix in letters
• WordPrefixDistance measures the length of common prefix in words
• NumberOverlapDistance measures the differences of numbers appearing the addresses
• NumberSeqDistance measures the numbers difference by Levenshtein distance
• PriorityWordDistance measures word differences and assign more weights in the difference if the word appears at the beginning of the string
• StrictNumberOverlapDistance a combination of number overlap and number seq differences with binary results
• WordBagDistance measures the difference in word bag (word counts)
• WordSetDistance measures the difference in words without considering of the numbers of appearance
• SymmetricWordSetDistance similar to WordSetDistance and is symmetric
• SymmetricWordSetWithIDF similar to SymmetricWordSetDistance and words are weighted by IDF in the collection

Searcher Composition

Searchers can be composed with different components. Please refer to uk.ac.cdrc.data.utility.text.AddressSearcher for an example of search composition.

Quick Example

import uk.ac.cdrc.data.utility.text.AddressSearcher
val addressSetA = IndexedSeq(
    "1 some street some city",
    "2 some street some city")
val addressSetB = IndexedSeq(
    "2 some street some city",
    "2a some street some city")
val as = AddressSearcher(addressSetA)
val matching = for {a <- addressSetB
     r <- as search a
     if !r.multiTops
    } yield (r.top, a)