Usage

You can find executables for Windows, Linux and macOS (Darwin) in the ./bin directory. The executable will be referred to as ./spt throughout this document.

Run ./spt --help for usage:

Usage of ./spt:
  -all_pairs
    	compare each wordlist in file
  -consonants string
    	path to file with consonant encodings
  -cost_groups_plot string
    	path to file with cost groups plot
  -count_groups_plot string
    	path to file with count groups plot
  -lang_1 string
    	first language to compare (optional)
  -lang_2 string
    	second language to compare (optional)
  -num_trials int
    	number of trials (default 1000000)
  -output string
    	path to output file (stdout if not specified)
  -set_a string
    	path to file containing wordlists for A (triggers AB mode)
  -set_b string
    	path to file containing wordlists for B (triggers AB mode)
  -sounds string
    	path to file containing sound classes (default "./data/sounds.xlsx")
  -verbose
    	verbose output
  -weights string
    	path to file containing class weights
  -wordlists string
    	path to file containing wordlists (default "./data/wordlists.xlsx")

Please note that --output and --consonant flag values are modified in the following way. For wordlists A and B:

• If path ends with .txt, then /some/path.txt --> /some/path_A_B.txt
• If path doesn't end with .txt, then /some/path --> /some/path_A_B.txt

Running test on two wordlists

To run test on a pair of wordlists contained in a single file, run:

$ ./spt --num_trials=1000000 --sounds=./data/sounds.xlsx --wordlists=./data/wordlists.xlsx --weights=./data/weights.xlsx
Positive pair 0: 19 drink: eːgʰʷ - ɨɣi
Positive pair 1: 39 hear: kʸlew - kuwli
Positive pair 2: 42 I : me - mi
Positive pair 3: 57 name: nom - nimi
Positive pair 4: 87 thou : ti - ti
Positive pair 5: 94 water: wed - weti
Positive pair 6: 98 who: kʷi - ku
N = 7 (number of positive pairs in the original list)
S = 436.000000 (cost of positive pairs in the original list)

k = 0:	89679 trial(s)
k = 1:	231915 trial(s)
k = 2:	281519 trial(s)
k = 3:	214596 trial(s)
k = 4:	115495 trial(s)
k = 5:	46891 trial(s)
k = 6:	15108 trial(s)
k = 7:	3863 trial(s)
k = 8:	768 trial(s)
k = 9:	148 trial(s)
k = 10:	17 trial(s)
k = 11:	1 trial(s)
P (counts) = 4797 / 1000000 = 0.004797

s = 0.000: 89679 trial(s)
s = 37.000: 24056 trial(s)
s = 39.000: 3864 trial(s)
s = 41.000: 20597 trial(s)
<...>
s = 574.000: 1 trial(s)
s = 580.000: 1 trial(s)
s = 600.000: 1 trial(s)
s = 601.000: 1 trial(s)
P (costs) = 681 / 1000000 = 0.000681

• --num_trials specifies how many times we shuffle the wordlists and count scores; default value is 1000000.
• --sounds is the path to file with sound tables; sample file can be found at ./data/sounds.xlsx (also the default value).
• --wordlists is the path to file with wordlists; sample file can be found at ./data/wordlists.xlsx (also the default value).
• --weights is the path containing mapping from Swadesh ID to its weight (missing IDs get weight value of 1.0); sample file can be found at ./data/weights.xlsx.

Running test on two sets of wordlists (AB mode)

$ ./spt --num_trials=1000000 --sounds=./data/sounds.xlsx --set_a=./data/wordlists.xlsx --set_b=./data/wordlists.xlsx

• --num_trials specifies how many times we shuffle the wordlists and count scores; default value is 1000000.
• --sounds is the path to file with sound tables; sample file can be found at ./data/sounds.xlsx (also the default value).
• --set_a is the path to file with wordlists set A (no default value).
• --set_b is the path to file with wordlists set B (no default value).
• --weights is the path containing mapping from Swadesh ID to its weight (missing IDs get weight value of 1.0); sample file can be found at ./data/weights.xlsx (also the default value).

If you run the command as specified above (using the same file for sets A and B), the program will execute correctly (but probability will always be zero, which is expected).

Building plots

Pass the --count_groups_plot option to build a plot representing how many trials gave a certain amount of matches:

$ ./spt --count_groups_plot=./count_groups.svg

Sample count groups plot:

Pass the --cost_groups_plot option to build a plot representing how many trials gave a certain cost of matches:

$ ./spt --cost_groups_plot=./cost_groups.svg

Sample cost groups plot:

Building from source

You need to have Go installed. Project must be put to $GOPATH/src/github.com/. Run:

make deps
make build