How does a computer know that ‘Smith’ and ‘Smyth’ are the same name? Soundex — patented in 1918 for the US census — keeps the first letter, throws away the vowels, and turns the remaining consonants into digits by how they SOUND. Names that sound alike land on the same four-character code. Slide through the pairs and watch them collide.
Soundex encodes a name to a letter + three digits: keep the first letter; drop vowels and h/w; map the rest by phonetic class (b,f,p,v→1; c,g,j,k,q,s,x,z→2; d,t→3; l→4; m,n→5; r→6); collapse adjacent duplicates; pad or trim to four. Similar-SOUNDING names collapse to the SAME code — Robert and Rupert both give R163, Smith and Smyth both S530. A fail-loud self-check throws unless those homophone pairs encode identically — phonetic matching reduced to a lookup.
Soundex is Anglocentric and crude: it over-merges (many unrelated names share a code) and mishandles non-English names (Metaphone / Double-Metaphone improve on it). The encoding rule and the homophone collisions shown are exact.