TextPHash

Perceptual Hash for text strings.

• Source repository: Github: mlefkon/text-phash
• NPM Package: NPM: text-phash

What it does

• Computes a perceptual hash for a text string.
• Compares perceptual hashes to give a percent similarity between two text strings.

Usage

const TextPHash = require('text-phash')
// OR
import TextPHash from 'text-phash'

let hashA = TextPHash.computePHash("The quick brown fox jumped over the black fence.")
let hashB = TextPHash.computePHash("Over the black fence, the quick brown fox jumped.")
let pctMatch = TextPHash.percentMatch(hashA, hashB)
console.log(hashA) // 00500000000000000000000500000000000F0050005000000000000000500000
console.log(hashB) // 00500005000000000000000500000000000F0000005000000000000000500000
console.log(pctMatch);  // 77.77777777777779

Methodology

1. Supply text (can be one word or a lengthy book)
2. Tokenize text into neighboring word-groups. Number of words in each group is set in options:NGRAM_WORDS.
3. Initialize a [hashHits] array with zeros, one 'counter' for each possible hash value. Number of hash values is set in options:WORD_HASH_BITS.
4. Hash each word-group.
5. For each hash encountered, increment it's 'counter' in the [hashHits] array
6. Normalize all [hashHits] counters between 0, for no hits, and a set maximum (set in options:HIT_VALUE_BITS) hits.
7. Convert [hashHits] array into a hexadecimal string.
8. Compare two hashes by converting hex back into [hashHits] array and comparing the difference in hits.

Functions

For optional options parameter {object}, supply one or more properties from the 'Default Options' object below.

computePHash()

TextPHash.computePHash(text)
TextPHash.computePHash(text, options)

• Returns a hexadecimal number representing a binary string (2 ^ WORD_HASH_BITS x 2 ^ HIT_VALUE_BITS) bits long. Using the default options, this will be a 64 digit hexadecimal string.

percentMatch()

TextPHash.percentMatch(pHashA, pHashB)
TextPHash.percentMatch(pHashA, pHashB, options)

• If options are supplied, they must be the same as those used to create the hashes.
• Returns a number between zero and 100.

Default Options

Available on the static class object TextPHash.DefaultOptions:

• NGRAM_WORDS: default = 2

  Number of 'neighbor' words that will be hashed together.

  For example, a value of 1: ABCDE=>[A,B,C,D,E], 2: ABCDE => [AB, BC, CD, DE], 3: ABCDE => [ABC,BCD,CDE]

• WORD_HASH_FUNCTION: default = TextPHash.WordHashDJB

  A function that does a non-unique hash on each word-group/ngram.

  Select any TextPHash.WordHash... function in TextPHash class (DJB, FNV1a, Murmur3). Or provide your own with signature: (strText, intHashBitSize) => intHash

• WORD_HASH_BITS: default = 6

  The binary size of hash produced by WORD_HASH_FUNCTION.

  Hashes are not meant to be unique, so this can be a low number. The hashes build a histogram of melded word frequencies. This is the 'x value' in the word-group-hash histogram. So if this is '6', there will be 2^6 possible hashes, or 64 'x values'.

• HIT_VALUE_BITS: default = 4

  Binary size of hit counter for a single hash. Actual hits are adjusted down to these discrete values.

  So if this is '4' and hash counters range from 0 to a max of 140 hits, the 140 value will be adjusted to (2^4)-1, or a max value of 15. A hash counter with lower value, say 70 hits, would get an adjusted value of 8. This is the 'y value' in the word-group-hash histogram.