private void Align(List <string> database, List <string> query, params int[] result)
{
var aligner = new LongTextAligner(database, 1);
int[] alignment = aligner.Align(query);
Assert.IsTrue(Helper.Contains(Utilities.AsList(alignment), result));
}
public void TextAligner_SmallAlign()
{
var wordList = new Dictionary <string[], int[]>
{
{ new[] { "foo", "foo" }, new[] { -1, -1 } },
{ new[] { "foo", "baz" }, new[] { 2, 3 } },
{ new[] { "foo", "bar", "foo", "bar", "baz", "42" }, new[] { 0, 1, 2, 4, 5, 6 } },
{ new[] { "foo", "bar", "foo", "baz", "bar" }, new[] { 0, 1, 2, 3, 4 } }
};
foreach (var item in wordList.Keys)
{
Assert.IsTrue(Helper.Contains(Utilities.AsList(_aligner.Align(item.ToList())), wordList[item]));
}
}
public List <WordResult> Align(FileInfo audioUrl, List <string> sentenceTranscript)
{
var transcript = SentenceToWords(sentenceTranscript);
var aligner = new LongTextAligner(transcript, TupleSize);
var alignedWords = new Dictionary <int, WordResult>();
var ranges = new LinkedList <Range>();
//var texts = new ArrayDeque();
//var timeFrames = new ArrayDeque();
var texts = new LinkedList <List <string> >();
var timeFrames = new LinkedList <TimeFrame>();
ranges.AddLast(new Range(0, transcript.Count));
texts.Offer(transcript);
TimeFrame totalTimeFrame = TimeFrame.Infinite;
timeFrames.Offer(totalTimeFrame);
long lastFrame = TimeFrame.Infinite.End;
for (int i = 0; i < 4; i++)
{
if (i == 3)
{
_context.SetLocalProperty("decoder->searchManager", "alignerSearchManager");
}
while (texts.Count != 0)
{
Debug.Assert(texts.Count == ranges.Count);
Debug.Assert(texts.Count == timeFrames.Count);
var text = texts.Poll();
var frame = timeFrames.Poll();
var range = ranges.Poll();
if (i < 3 && texts.Count < MinLmAlignSize)
{
continue;
}
this.LogInfo("Aligning frame " + frame + " to text " + text + " range " + range);
if (i < 3)
{
_languageModel.SetText(text);
}
_recognizer.Allocate();
if (i == 3)
{
_grammar.SetWords(text);
}
_context.SetSpeechSource(audioUrl.OpenRead(), frame);
var hypothesis = new List <WordResult>();
Result speechResult;
while (null != (speechResult = _recognizer.Recognize()))
{
hypothesis.AddRange(speechResult.GetTimedBestResult(false));
}
if (i == 0)
{
if (hypothesis.Count > 0)
{
lastFrame = hypothesis[hypothesis.Count - 1].TimeFrame.End;
}
}
var words = new List <string>();
foreach (WordResult wr in hypothesis)
{
words.Add(wr.Word.Spelling);
}
int[] alignment = aligner.Align(words, range);
List <WordResult> results = hypothesis;
this.LogInfo("Decoding result is " + results);
// dumpAlignment(transcript, alignment, results);
DumpAlignmentStats(transcript, alignment, results);
for (int j = 0; j < alignment.Length; j++)
{
if (alignment[j] != -1)
{
alignedWords.Add(alignment[j], hypothesis[j]);
}
}
_recognizer.Deallocate();
}
ScheduleNextAlignment(transcript, alignedWords, ranges, texts, timeFrames, lastFrame);
}
return(new List <WordResult>(alignedWords.Values));
}
