-
Notifications
You must be signed in to change notification settings - Fork 1
/
Generator.cs
86 lines (79 loc) · 2.97 KB
/
Generator.cs
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
namespace MagicMau.ProceduralNameGenerator
{
/// <summary>
/// Provides procedural generation of words using high-order Markov chains
/// Uses Katz's back-off model - chooses the next character based on conditional probability given the last n-characters (where model order = n)
/// and backs down to lower order models when higher models fail
/// Uses a Dirichlet prior, which is like additive smoothing and raises the chances of a "random" letter being picked instead of one that's trained in
///
/// <seealso cref="https://github.com/Tw1ddle/MarkovNameGenerator/blob/master/src/markov/namegen/Generator.hx"/>
/// </summary>
public class Generator
{
public int Order { get; set; }
public double Smoothing { get; set; }
private List<Model> models;
/// <summary>
///
/// </summary>
/// <param name="trainingData">training data for the generator, array of words</param>
/// <param name="order">number of models to use, will be of orders up to and including "order"</param>
/// <param name="smoothing">the dirichlet prior/additive smoothing "randomness" factor</param>
public Generator(IEnumerable<string> trainingData, int order, double smoothing)
{
Order = order;
Smoothing = smoothing;
models = new List<Model>();
// Identify and sort the alphabet used in the training data
var letters = new HashSet<char>();
foreach (var word in trainingData)
{
foreach (char c in word)
{
letters.Add(c);
}
}
var domain = letters.OrderBy(c => c).ToList();
domain.Insert(0, '#');
// create models
for (int i = 0; i < order; i++)
{
models.Add(new Model(trainingData, order - i, smoothing, domain));
}
}
/// <summary>
/// Generates a word
/// </summary>
/// <returns></returns>
public string Generate(Random rnd)
{
string name = new string('#', Order);
char letter = GetLetter(name, rnd);
while (letter != '#' && letter != '\0')
{
name += letter;
letter = GetLetter(name, rnd);
}
return name;
}
private char GetLetter(string name, Random rnd)
{
char letter = '\0';
string context = name.Substring(name.Length - Order);
foreach (var model in models)
{
letter = model.Generate(context, rnd);
if (letter == '\0')
context = context.Substring(1);
else
break;
}
return letter;
}
}
}