public void TestSequenceProbability()
{
UniGram unigram = new UniGram();
Assert.AreEqual(
0,
unigram.Compile().SequenceProbability(new string[] { "a", "b" }));
unigram.AddData(null, "a");
Assert.AreEqual(
1,
unigram.Compile().SequenceProbability(new string[] { "a", "a" }));
unigram.AddData(null, "b");
Assert.AreEqual(
0.5,
unigram.Compile().SequenceProbability(new string[] { "a" }));
Assert.AreEqual(
0.25,
unigram.Compile().SequenceProbability(new string[] { "b", "b" }));
unigram.AddData(null, "b");
unigram.AddData(null, "b");
Assert.AreEqual(
0.75,
unigram.Compile().SequenceProbability(new string[] { "b" }));
Assert.AreEqual(
0.75 * 0.25 * 0.75,
unigram.Compile().SequenceProbability(new string[] { "b", "a", "b" }));
}
public void TestPerplexity()
{
UniGram unigram = new UniGram();
Assert.AreEqual(
double.PositiveInfinity,
unigram.Compile().Perplexity(new string[] { "a", "b" }));
unigram.AddData(null, "a");
Assert.AreEqual(
1,
unigram.Compile().Perplexity(new string[] { "a", "a" }));
unigram.AddData(null, "b");
Assert.AreEqual(
1 / 0.5,
unigram.Compile().Perplexity(new string[] { "a" }));
Assert.AreEqual(
1 / 0.25,
unigram.Compile().Perplexity(new string[] { "b", "b" }));
unigram.AddData(null, "b");
unigram.AddData(null, "b");
Assert.AreEqual(
1 / 0.75,
unigram.Compile().Perplexity(new string[] { "b" }));
Assert.AreEqual(
1 / (0.75 * 0.25 * 0.75),
unigram.Compile().Perplexity(new string[] { "b", "a", "b" }));
}
public void TestGetValues()
{
UniGram unigram = new UniGram();
Assert.AreEqual(
new Dictionary <string, float>(),
unigram.Compile().GetValues(null));
unigram.AddData(null, "a");
Assert.AreEqual(
new Dictionary <string, float>()
{
{ "a", 1f }
},
unigram.Compile().GetValues(null));
unigram.AddData(null, "a");
unigram.AddData(null, "a");
unigram.AddData(null, "b");
Assert.AreEqual(
new Dictionary <string, float>()
{
{ "a", 0.75f }, { "b", 0.25f }
},
unigram.Compile().GetValues(null));
}
public void TestUpdateMemory()
{
UniGram a = new UniGram();
a.AddData(null, "a");
a.UpdateMemory(0.9f);
Assert.AreEqual(0.9f, a.Grammar["a"]);
a.AddData(null, "b");
a.UpdateMemory(0.9f);
Assert.IsTrue(Mathf.Approximately(0.81f, a.Grammar["a"]));
Assert.AreEqual(0.9f, a.Grammar["b"]);
}
public void TestClone()
{
UniGram unigram = new UniGram();
unigram.AddData(null, "a");
unigram.AddData(null, "a");
unigram.AddData(null, "b");
CompiledUniGram compiledUnigram = new CompiledUniGram(unigram.Grammar);
ICompiledGram clone = compiledUnigram.Clone();
Assert.AreEqual(compiledUnigram.GetN(), clone.GetN());
Assert.AreEqual(compiledUnigram.GetValues(null), clone.GetValues(null));
}
public void TestHasNextStep()
{
UniGram unigram = new UniGram();
Assert.IsFalse(unigram.Compile().HasNextStep(null));
unigram.AddData(null, "a");
Assert.IsTrue(unigram.Compile().HasNextStep(null));
unigram.AddData(null, "a");
Assert.IsTrue(unigram.Compile().HasNextStep(null));
unigram.AddData(null, "b");
Assert.IsTrue(unigram.Compile().HasNextStep(null));
}
public void TestAddDataString()
{
UniGram a = new UniGram();
a.AddData(null, "a");
Assert.AreEqual(1, a.Grammar.Keys.Count);
Assert.AreEqual(1f, a.Grammar["a"]);
a.AddData(new string[] { "asdf" }, "a");
Assert.AreEqual(1, a.Grammar.Keys.Count);
Assert.AreEqual(2f, a.Grammar["a"]);
a.AddData(new string[] { "31234" }, "b");
Assert.AreEqual(2, a.Grammar.Keys.Count);
Assert.AreEqual(1f, a.Grammar["b"]);
Assert.AreEqual(2f, a.Grammar["a"]);
}
public void TestSequenceProbability()
{
HierarchicalNGram gram = new HierarchicalNGram(3, 0.9f);
Assert.AreEqual(
0,
gram.Compile().SequenceProbability(new string[] { "a", "b", "c" }));
double denominator = 0.9 + 0.81 + 0.729;
double triWeight = 0.9 / denominator;
double biWeight = 0.81 / denominator;
double uniweight = 0.729 / denominator;
UniGram gram1 = new UniGram();
gram1.AddData(null, "a");
gram1.AddData(null, "a");
gram1.AddData(null, "b");
NGram gram2 = new NGram(2);
gram2.AddData(new string[] { "a" }, "a");
gram2.AddData(new string[] { "a" }, "b");
NGram gram3 = new NGram(3);
gram3.AddData(new string[] { "a", "a" }, "b");
gram.AddData(new string[] { "a", "a" }, "b");
ICompiledGram c1 = gram1.Compile();
ICompiledGram c2 = gram2.Compile();
ICompiledGram c3 = gram3.Compile();
string[] input = new string[] { "a", "a", "b" };
double expected =
uniweight * c1.SequenceProbability(input) +
biWeight * c2.SequenceProbability(input) +
triWeight * c3.SequenceProbability(input);
double actual = gram.Compile().SequenceProbability(new string[] { "a", "a", "b" });
Assert.IsTrue(
Mathf.Approximately((float)expected, (float)actual),
$"Expected {expected} but received {actual}.");
}
public void TestAddDataWeight()
{
UniGram a = new UniGram();
a.AddData("a", 2.2f);
Assert.AreEqual(1, a.Grammar.Keys.Count);
Assert.AreEqual(2.2f, a.Grammar["a"]);
a.AddData("a", 0.8f);
Assert.AreEqual(1, a.Grammar.Keys.Count);
Assert.AreEqual(3f, a.Grammar["a"]);
a.AddData("b", 2.823f);
Assert.AreEqual(3f, a.Grammar["a"]);
Assert.AreEqual(2.823f, a.Grammar["b"]);
Assert.AreEqual(2, a.Grammar.Keys.Count);
}
public void TestGet()
{
UniGram unigram = new UniGram();
unigram.AddData(null, "a");
unigram.AddData(null, "a");
unigram.AddData(null, "b");
unigram.AddData(null, "c");
CompiledUniGram compiledUnigram = new CompiledUniGram(unigram.Grammar);
bool seenA = false;
bool seenB = false;
bool seenC = false;
for (int i = 0; i < 1000; ++i)
{
string val = compiledUnigram.Get(null);
switch (val)
{
case "a":
seenA = true;
break;
case "b":
seenB = true;
break;
case "c":
seenC = true;
break;
default:
Assert.Fail($"{val} should not be possible.");
break;
}
}
// in theory, we could potentially not see one of these but it is very unlikely.
Assert.IsTrue(seenA);
Assert.IsTrue(seenB);
Assert.IsTrue(seenC);
}
public void TestAddGrammar()
{
UniGram a = new UniGram();
UniGram b = new UniGram();
a.AddGrammar(b);
Assert.AreEqual(0, a.Grammar.Keys.Count);
b.AddData(null, "a");
a.AddGrammar(b);
Assert.AreEqual(1, a.Grammar.Keys.Count);
Assert.AreEqual(1f, a.Grammar["a"]);
b.AddData(null, "b");
b.AddData(null, "a");
a.AddGrammar(b);
Assert.AreEqual(2, a.Grammar.Keys.Count);
Assert.AreEqual(3f, a.Grammar["a"]);
Assert.AreEqual(1f, a.Grammar["b"]);
}
public void TestGetGuesses()
{
UniGram unigram = new UniGram();
Assert.AreEqual(0, unigram.Compile().GetGuesses(null).Length);
unigram.AddData(null, "a");
Assert.AreEqual(1, unigram.Compile().GetGuesses(null).Length);
unigram.AddData(null, "a");
Assert.AreEqual(1, unigram.Compile().GetGuesses(null).Length);
unigram.AddData(null, "b");
Assert.AreEqual(2, unigram.Compile().GetGuesses(null).Length);
unigram.AddData(null, "c");
string[] guesses = unigram.Compile().GetGuesses(null);
Assert.AreEqual(3, guesses.Length);
Assert.IsTrue(guesses.Contains("a"));
Assert.IsTrue(guesses.Contains("b"));
Assert.IsTrue(guesses.Contains("c"));
}
public void TestAddUnigram()
{
HierarchicalNGram a = new HierarchicalNGram(2, 0.6f);
UniGram u1 = a.Grammars[0] as UniGram;
NGram n2 = a.Grammars[1] as NGram;
UniGram unigram = new UniGram();
unigram.AddData(null, "a");
a.AddGrammar(unigram);
Assert.AreEqual(1, u1.Grammar.Keys.Count);
Assert.AreEqual(0, n2.Grammar.Keys.Count);
Assert.AreEqual(1, u1.Grammar["a"]);
}
public void TestConstructor()
{
UniGram unigram = new UniGram();
Assert.DoesNotThrow(() => { unigram.Compile(); });
unigram.AddData(null, "a");
Assert.DoesNotThrow(() => { unigram.Compile(); });
unigram.AddData(null, "b");
Assert.DoesNotThrow(() => { unigram.Compile(); });
unigram.AddData(null, "b");
unigram.AddData(null, "b");
unigram.AddData(null, "b");
unigram.AddData(null, "a");
Assert.DoesNotThrow(() => { unigram.Compile(); });
}