/// <summary>
/// Load script file into sentence id and script entry maped dictionary.
/// </summary>
/// <param name="scriptFilePath">Script file to read.</param>
/// <param name="language">Language of the script.</param>
/// <param name="engineType">Engine of the script to support.</param>
/// <param name="outEntries">Output of script items.</param>
/// <returns>Data error set found.</returns>
public static DataErrorSet ReadAllData(string scriptFilePath,
Language language, EngineType engineType,
SortedDictionary<string, ScriptItem> outEntries)
{
if (string.IsNullOrEmpty(scriptFilePath))
{
throw new ArgumentNullException("scriptFilePath");
}
if (outEntries == null)
{
throw new ArgumentNullException("outEntries");
}
Collection<ScriptItem> entriesInCollection = new Collection<ScriptItem>();
DataErrorSet errorSet = ReadAllData(scriptFilePath,
language, engineType, entriesInCollection);
foreach (ScriptItem entry in entriesInCollection)
{
if (outEntries.ContainsKey(entry.Id))
{
DataError error = new DataError();
error.SentenceId = entry.Id;
error.Message = "Sentence id duplicated.";
error.FilePath = scriptFilePath;
errorSet.Errors.Add(error);
continue;
}
outEntries.Add(entry.Id, entry);
}
return errorSet;
}
/// <summary>
/// Given one sentence's pronunciation string, convert
/// One Phone-based segment file to Unit-based segment file.
/// </summary>
/// <param name="pronunciation">Pronunciation string.</param>
/// <param name="filePath">Phone-based segment file.</param>
/// <param name="targetFilePath">Unit-based segment file.</param>
/// <returns>Data error found.</returns>
public virtual DataError CombinePhone(string pronunciation,
string filePath, string targetFilePath)
{
pronunciation = Core.Pronunciation.CleanDecorate(pronunciation);
string[] slices = PronunciationSeparator.SplitSlices(pronunciation);
Collection<WaveSegment> phoneSegs = SegmentFile.ReadAllData(filePath);
DataError dataError = null;
using (StreamWriter sw = new StreamWriter(targetFilePath))
{
int sliceIndex = 0;
StringBuilder slice = new StringBuilder();
for (int i = 0; i < phoneSegs.Count;)
{
if (phoneSegs[i].IsSilenceFeature)
{
sw.WriteLine(phoneSegs[i].ToString());
i++;
continue;
}
if (sliceIndex >= slices.Length)
{
string sid = Path.GetFileNameWithoutExtension(filePath);
dataError = new DataError(filePath,
"Data does not align between phone segmentation and pronunciation in CombinePhone", sid);
break;
}
TtsMetaUnit ttsMetaUnit = new TtsMetaUnit(Language);
ttsMetaUnit.Name = slices[sliceIndex];
sliceIndex++;
// Clear first
slice.Remove(0, slice.Length);
foreach (TtsMetaPhone phone in ttsMetaUnit.Phones)
{
if (string.IsNullOrEmpty(phone.Name))
{
continue;
}
if (slice.Length > 0)
{
slice.Append("+");
}
slice.Append(phone.FullName);
}
if (slice.Length == 0)
{
continue;
}
sw.Write(phoneSegs[i].StartTime.ToString("F5", CultureInfo.InvariantCulture));
sw.WriteLine(" " + slice.ToString());
i += ttsMetaUnit.Phones.Length;
}
}
if (dataError != null)
{
try
{
File.Delete(targetFilePath);
}
catch (IOException ioe)
{
Console.WriteLine(ioe.Message);
}
}
return dataError;
}
/// <summary>
/// Based on script file, build a mono-phone MLF (See HTK document) file .
/// </summary>
/// <param name="writer">Writer to write result out if not null.</param>
/// <param name="scriptFilePath">Script file path.</param>
/// <returns>Data error set found.</returns>
private DataErrorSet BuildMonoMlf(TextWriter writer, string scriptFilePath)
{
Collection<ScriptItem> entries = new Collection<ScriptItem>();
DataErrorSet errorSet = ReadAllData(scriptFilePath,
Language, EngineType, entries);
if (entries.Count == 0)
{
string message = string.Format(CultureInfo.InvariantCulture,
"There is no script sentence found in file [{0}].", scriptFilePath);
throw new InvalidDataException(message);
}
if (writer != null)
{
// write MLF file header
writer.WriteLine("#!MLF!#");
}
foreach (ScriptItem entry in entries)
{
try
{
BuildMonoMlf(writer, entry);
}
catch (InvalidDataException ide)
{
DataError error = new DataError(scriptFilePath,
Helper.BuildExceptionMessage(ide), entry.Id);
errorSet.Errors.Add(error);
}
}
return errorSet;
}
/// <summary>
/// Append a script into other script file.
/// </summary>
/// <param name="subScriptFilePath">Source script file.</param>
/// <param name="outFilePath">Target script file.</param>
/// <param name="append">Whether appending to target script file.</param>
/// <returns>Invalid format script entry strings.</returns>
private DataErrorSet AppendScript(string subScriptFilePath,
string outFilePath, bool append)
{
DataErrorSet errorSet = new DataErrorSet();
SortedDictionary<string, ScriptItem> existEntries = new SortedDictionary<string, ScriptItem>();
if (append && File.Exists(outFilePath))
{
errorSet = ReadAllData(outFilePath, Language, EngineType, existEntries);
}
else
{
Helper.EnsureFolderExistForFile(outFilePath);
}
SortedDictionary<string, ScriptItem> subEntries = new SortedDictionary<string, ScriptItem>();
DataErrorSet subErrorSet = ReadAllData(subScriptFilePath,
Language, EngineType, subEntries);
errorSet.Merge(subErrorSet);
using (StreamWriter sw = new StreamWriter(outFilePath, append, Encoding.Unicode))
{
foreach (string sid in subEntries.Keys)
{
if (existEntries.ContainsKey(sid))
{
DataError error = new DataError(subScriptFilePath,
"Entry already exists in script file [" + outFilePath + "]", sid);
errorSet.Errors.Add(error);
continue;
}
// hook handling
DataError preAppendError = ProcessPronunciation(subEntries[sid]);
if (preAppendError != null)
{
errorSet.Errors.Add(preAppendError);
continue;
}
sw.WriteLine(subEntries[sid].ToString(true, true, true));
}
}
return errorSet;
}
/// <summary>
/// Build pronunciation script from file list.
/// </summary>
/// <param name="scriptFilePaths">Hiragana script file list.</param>
/// <param name="outFilePath">Romaji pronunciation script file.</param>
/// <returns>Data error set found.</returns>
public DataErrorSet BuildScript(string[] scriptFilePaths,
string outFilePath)
{
if (scriptFilePaths == null)
{
throw new ArgumentNullException("scriptFilePaths");
}
DataErrorSet errorSet = new DataErrorSet();
for (int i = 0; i < scriptFilePaths.Length; i++)
{
string scriptFilePath = scriptFilePaths[i];
if (string.IsNullOrEmpty(scriptFilePath))
{
throw new InvalidDataException("scriptFilePath");
}
if (!scriptFilePath.EndsWith(ScriptFile.Extension, StringComparison.Ordinal))
{
continue;
}
// all script files should be saved in unicode
if (!Helper.IsUnicodeFile(scriptFilePath))
{
DataError error = new DataError(scriptFilePath,
"script file should be saved in Unicode.");
errorSet.Errors.Add(error);
continue;
}
// do appending
DataErrorSet subErrorSet = AppendScript(scriptFilePath,
outFilePath, (i != 0));
// merge error messages
errorSet.Merge(subErrorSet);
}
return errorSet;
}
/// <summary>
/// Validate the phones in slice are valid or not.
/// </summary>
/// <param name="entry">Script item.</param>
/// <param name="slice">Pronunciation of slice.</param>
/// <returns>Data error found.</returns>
public static DataError ValidatePhones(ScriptItem entry, string slice)
{
if (entry == null)
{
throw new ArgumentNullException("entry");
}
if (entry.PronunciationSeparator == null)
{
string message = Helper.NeutralFormat("entry.PronunciationSeparator should not be null.");
throw new ArgumentException(message);
}
if (string.IsNullOrEmpty(entry.PronunciationSeparator.Phone))
{
string message = Helper.NeutralFormat("entry.PronunciationSeparator.Phone should not be null.");
throw new ArgumentException(message);
}
if (string.IsNullOrEmpty(slice))
{
throw new ArgumentNullException("slice");
}
DataError error = null;
Phoneme phoneme = Localor.GetPhoneme(entry.Language);
string[] items = slice.Split(new string[] { entry.PronunciationSeparator.Phone },
StringSplitOptions.RemoveEmptyEntries);
for (int i = 0; i < items.Length; i++)
{
// TODO: PS#13181 Offline tools:Syllable veridation and pronunciation design
if (items[i] == "1" || items[i] == "2" || items[i] == "3")
{
continue;
}
if (items[i].StartsWith("_", StringComparison.Ordinal) && items[i].EndsWith("_", StringComparison.Ordinal))
{
// special phone
continue;
}
if (phoneme.ToneManager.NameMap.ContainsKey(items[i]))
{
// tone
continue;
}
if (phoneme.TtsPhones.IndexOf(items[i]) < 0)
{
// invalid tts phone found
string message = string.Format(CultureInfo.InvariantCulture,
"The phone[{0}] in slice[{1}] is invalid",
items[i], slice);
error = new DataError("null", message, entry.Id);
break;
}
}
return error;
}
/// <summary>
/// Validate the slices in syllable are valid or not.
/// </summary>
/// <param name="entry">Script item.</param>
/// <param name="syllable">Pronunciation of syllable.</param>
/// <returns>Data error found.</returns>
public static DataError ValidateSlices(ScriptItem entry, string syllable)
{
if (entry == null)
{
throw new ArgumentNullException("entry");
}
if (entry.PronunciationSeparator == null)
{
string message = Helper.NeutralFormat("entry.PronunciationSeparator should not be null.");
throw new ArgumentException(message);
}
if (string.IsNullOrEmpty(entry.PronunciationSeparator.Slice))
{
string message = Helper.NeutralFormat("entry.PronunciationSeparator.Slice should not be null.");
throw new ArgumentException(message);
}
if (string.IsNullOrEmpty(syllable))
{
throw new ArgumentNullException("syllable");
}
DataError dataError = null;
string[] slices = syllable.Split(new string[] { entry.PronunciationSeparator.Slice },
StringSplitOptions.None);
for (int k = 0; k < slices.Length; k++)
{
if (string.IsNullOrEmpty(slices[k]))
{
string message = string.Format(CultureInfo.InvariantCulture,
"The slice[{0}] of syllable[{1}] is empty by separator [{2}]",
k, syllable, entry.PronunciationSeparator.Slice);
dataError = new DataError("null", message, entry.Id);
break;
}
// check phones
dataError = ValidatePhones(entry, slices[k]);
if (dataError != null)
{
break;
}
}
return dataError;
}
/// <summary>
/// Validate the syllable in word pronunciation are valid or not.
/// </summary>
/// <param name="entry">Script item.</param>
/// <param name="word">Pronunciation of word.</param>
/// <returns>Data error found.</returns>
public static DataError ValidateSyllables(ScriptItem entry, string word)
{
if (entry == null)
{
throw new ArgumentNullException("entry");
}
if (entry.PronunciationSeparator == null)
{
string message = Helper.NeutralFormat("entry.PronunciationSeparator should not be null.");
throw new ArgumentException(message);
}
if (string.IsNullOrEmpty(entry.PronunciationSeparator.Syllable))
{
string message = Helper.NeutralFormat("entry.PronunciationSeparator.Syllable should not be null.");
throw new ArgumentException(message);
}
if (string.IsNullOrEmpty(word))
{
throw new ArgumentNullException("word");
}
Phoneme phoneme = Localor.GetPhoneme(entry.Language);
DataError dataError = null;
string[] syllables = word.Split(new string[] { entry.PronunciationSeparator.Syllable },
StringSplitOptions.None);
for (int j = 0; j < syllables.Length; j++)
{
string syllable = syllables[j].Trim();
if (string.IsNullOrEmpty(syllable))
{
string message = string.Format(CultureInfo.InvariantCulture,
"The syllable[{0}] of word[{1}] pronunciation is empty by separator [{2}]",
j, word, entry.PronunciationSeparator.Syllable);
dataError = new DataError("null", message, entry.Id);
break;
}
if (Regex.Match(syllable, "^_(.*)_$").Success)
{
// Special unit
continue;
}
string[] itmes = entry.PronunciationSeparator.SplitPhones(syllable);
TtsMetaUnit ttsMetaUnit = new TtsMetaUnit(entry.Language);
ttsMetaUnit.Name = string.Join(" ", itmes);
string[] phones = ttsMetaUnit.GetPhonesName();
// Tell whether is a valid nucleus,
// which could be syllable with no vowel in some languages, like fr-CA
SliceData sliceData = Localor.GetSliceData(entry.Language);
if (sliceData.NucleusSlices.IndexOf(ttsMetaUnit.Name) < 0)
{
bool goodSyllable;
if (entry.Language == Language.EnUS)
{
// syllable that must have vowels
goodSyllable = IsGoodSyllableWithVowel(entry, phoneme, phones);
}
else if (entry.Language == Language.RuRU)
{
// A Russian syllable can have no sonorant
goodSyllable = IsSyllableWithLessVowel(entry, phoneme, phones);
}
else
{
// syllable that must have vowels or sonorants
goodSyllable = IsGoodSyllableWithSonorant(entry, phoneme, phones);
}
if (!goodSyllable)
{
int[] vowelIndexes = phoneme.GetVowelIndexes(phones);
string str1 = "There must be minimum {0} vowels or maximum {1} included in syllable ";
string str2 = "or the syllable should have one sonorant and more than one consonants, ";
string str3 = "but {2} vowels are found in syllable [{3}] of word [{4}].";
string message = string.Format(CultureInfo.InvariantCulture, str1 + str2 + str3,
entry.MinVowelCountInSyllable, entry.MaxVowelCountInSyllable,
vowelIndexes.Length, syllables[j], word);
dataError = new DataError("null", message, entry.Id);
break;
}
}
// check slice's pronunciation
dataError = ValidateSlices(entry, syllable);
if (dataError != null)
{
break;
}
}
return dataError;
}
/// <summary>
/// Validate the word pronunciations in sentence are valid or not.
/// </summary>
/// <param name="entry">Script item.</param>
/// <param name="isBadPhoneValid">True means that bad phone is valid.</param>
/// <returns>Data error found.</returns>
public static DataError ValidatePronunciation(ScriptItem entry, bool isBadPhoneValid)
{
DataError dataError = null;
string[] words = entry.Pronunciation.Split(
new string[] { entry.PronunciationSeparator.Word },
StringSplitOptions.None);
for (int i = 0; i < words.Length; i++)
{
if ((i == 0 || i == words.Length - 1) &&
string.IsNullOrEmpty(words[i]))
{
// It makes sense for first or last one is with empty string
continue;
}
if (string.IsNullOrEmpty(words[i]))
{
string message = string.Format(CultureInfo.InvariantCulture,
"The word[{0}] pronunciation is empty by separator [{1}]",
i, entry.PronunciationSeparator.Word);
dataError = new DataError("null", message, entry.Id);
break;
}
// Check syllable's pronunciation
string newWord = words[i];
if (isBadPhoneValid)
{
// remove the bad phone label from bad phone.
string pattern = @"([a-zA-Z]+)" + BadPhoneLabel;
string replacement = @"$1";
newWord = Regex.Replace(newWord, pattern, replacement, RegexOptions.CultureInvariant);
}
dataError = ValidateSyllables(entry, newWord);
if (dataError != null)
{
break;
}
}
return dataError;
}
