public bool Match(SolarixGrammarEngineNET.SyntaxTreeNode proj, int iver, SolarixGrammarEngineNET.GrammarEngine2 gren)
{
if (lexeme != null)
{
return(proj.GetWord().Equals(lexeme, StringComparison.InvariantCultureIgnoreCase));
}
if (id_lemma != null)
{
int ekey = proj.GetVersionEntryID(iver);
if (id_lemma.Contains(ekey))
{
return(true);
}
return(false);
}
if (pos != null)
{
bool pos_matched = false;
int ekey = proj.GetVersionEntryID(iver);
if (ekey != -1)
{
int id_class = gren.GetEntryClass(ekey);
pos_matched = pos.Contains(id_class);
}
if (!pos_matched)
{
return(false);
}
}
if (pairs != null)
{
bool contains_all_required_pairs = true;
foreach (SolarixGrammarEngineNET.CoordPair p in pairs)
{
if (!proj.VersionContains(iver, p))
{
contains_all_required_pairs = false;
break;
}
}
if (!contains_all_required_pairs)
{
return(false);
}
}
return(true);
}
public int MatchTags(WordData token, SolarixGrammarEngineNET.GrammarEngine2 gren)
{
foreach (TagMatcher m in matchers)
{
if (m.Match(token, gren))
{
return(m.GetId());
}
}
int entry_id = token.GetEntryID();
int pos_id = gren.GetEntryClass(entry_id);
string part_of_speech = gren.GetClassName(pos_id);
string tags = string.Join(" ", token.GetTags().Select(z => string.Format("{0}={1}", gren.GetCoordName(z.Item1), gren.GetCoordStateName(z.Item1, z.Item2))).ToArray());
string msg = string.Format("Can not find tag for {0} {{ {1} {2} }}", token.GetWord(), part_of_speech, tags);
throw new ApplicationException(msg);
}
public IEnumerable <SentenceData> Read(SolarixGrammarEngineNET.GrammarEngine2 gren)
{
string[] bin_corpora = System.IO.Directory.GetFiles(System.IO.Path.GetDirectoryName(corpus_path), System.IO.Path.GetFileName(corpus_path));
Console.WriteLine("There are {0} binary corpus segments", bin_corpora.Length);
foreach (string path1 in bin_corpora)
{
Console.WriteLine("Reading corpus {0}...", path1);
int corpus_count = 0;
IntPtr hCorpus = SolarixGrammarEngineNET.GrammarEngine.sol_OpenCorpusStorage8(gren.GetEngineHandle(), path1, false);
if (hCorpus == IntPtr.Zero)
{
throw new ApplicationException(string.Format("Can not open corpus {0}", corpus_path));
}
while (true)
{
corpus_count++;
IntPtr hSample1 = SolarixGrammarEngineNET.GrammarEngine.sol_LoadSyntaxTree(gren.GetEngineHandle(), hCorpus);
if (hSample1 == IntPtr.Zero)
{
break;
}
IntPtr hSample2 = SolarixGrammarEngineNET.GrammarEngine.sol_LoadSyntaxTree(gren.GetEngineHandle(), hCorpus);
IntPtr hSample3 = SolarixGrammarEngineNET.GrammarEngine.sol_LoadSyntaxTree(gren.GetEngineHandle(), hCorpus);
string sample = SolarixGrammarEngineNET.GrammarEngine.sol_GetSentenceW(hSample1);
var morphology = new SolarixGrammarEngineNET.AnalysisResults(gren, SolarixGrammarEngineNET.GrammarEngine.sol_GetTreeHandle(hSample1), false);
var tokenization = new SolarixGrammarEngineNET.AnalysisResults(gren, SolarixGrammarEngineNET.GrammarEngine.sol_GetTreeHandle(hSample2), false);
var syntax_tree = new SolarixGrammarEngineNET.AnalysisResults(gren, SolarixGrammarEngineNET.GrammarEngine.sol_GetTreeHandle(hSample3), false);
if (morphology.Count == tokenization.Count)
{
SentenceData sent = new SentenceData(sample);
for (int i = 0; i < morphology.Count; ++i)
{
WordData word_data = new WordData();
word_data.word = morphology[i].GetWord();
word_data.word_index = morphology[i].GetWordPosition();
word_data.entry_id = morphology[i].GetEntryID();
word_data.part_of_speech = gren.GetEntryClass(word_data.entry_id);
word_data.all_projs = tokenization[i];
foreach (var tag in morphology[i].GetPairs())
{
word_data.tags.Add(Tuple.Create(tag.CoordID, tag.StateID));
}
word_data.lemma = gren.GetEntryName(word_data.entry_id).ToLower();
if (word_data.lemma == "???")
{
word_data.lemma = word_data.word;
}
else if (word_data.lemma == "NUMBER_")
{
word_data.lemma = word_data.word;
}
sent.AddWord(word_data);
}
yield return(sent);
}
SolarixGrammarEngineNET.GrammarEngine.sol_FreeSyntaxTree(hSample1);
SolarixGrammarEngineNET.GrammarEngine.sol_FreeSyntaxTree(hSample2);
SolarixGrammarEngineNET.GrammarEngine.sol_FreeSyntaxTree(hSample3);
}
SolarixGrammarEngineNET.GrammarEngine.sol_CloseCorpusStorage(gren.GetEngineHandle(), hCorpus);
}
}
public IEnumerable <SentenceData> Read(SolarixGrammarEngineNET.GrammarEngine2 gren)
{
int beth_class = gren.FindPartOfSpeech("BETH");
int entry_begin_id = gren.FindEntry("BEGIN", beth_class);
int entry_end_id = gren.FindEntry("END", beth_class);
string[] files = System.IO.Directory.GetFiles(System.IO.Path.GetDirectoryName(corpus_path), System.IO.Path.GetFileName(corpus_path));
foreach (string path1 in files)
{
using (System.IO.StreamReader rdr = new System.IO.StreamReader(path1))
{
List <ConlluItem> items = new List <ConlluItem>();
while (!rdr.EndOfStream)
{
string line = rdr.ReadLine();
if (line == null)
{
break;
}
if (string.IsNullOrEmpty(line))
{
if (items.Count > 0)
{
string sample = string.Join(" ", items.Select(z => z.word));
SentenceData sent = new SentenceData(sample);
bool all_ok = true; // если не сможем выполнить конверию морфологической разметки, то
// выставим этот флаг в false и не будем возвращать объект SentenceData.
int word_index = 0;
SolarixGrammarEngineNET.AnalysisResults tokenization = gren.AnalyzeMorphology(sample, -1, SolarixGrammarEngineNET.GrammarEngine.MorphologyFlags.SOL_GREN_TOKENIZE_ONLY, 0);
if (tokenization.Count - 2 == items.Count)
{
// Добавить <BEGIN>
WordData wd_begin = new WordData();
wd_begin.word_index = 0;
wd_begin.word = "BEGIN";
wd_begin.entry_id = entry_begin_id;
wd_begin.part_of_speech = beth_class;
wd_begin.all_projs = tokenization[word_index];
sent.AddWord(wd_begin);
word_index++;
foreach (var item in items)
{
var projs = SolarixGrammarEngineNET.GrammarEngine.sol_ProjectWord(gren.GetEngineHandle(), item.word, 1);
int selected_proj = 0;
int nproj = SolarixGrammarEngineNET.GrammarEngine.sol_CountProjections(projs);
if (nproj > 1)
{
List <Tuple <int, int> > proj_scores = new List <Tuple <int, int> >();
int required_class = -1;
List <Tuple <int, int> > required_tags = new List <Tuple <int, int> >();
if (item.ud_class == "NOUN" || item.ud_class == "PROPN")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.NOUN_ru;
}
else if (item.ud_class == "PUNCT")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.PUNCTUATION_class;
}
else if (item.ud_class == "ADJ" || item.ud_class == "DET")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.ADJ_ru;
}
else if (item.ud_class == "AUX" || item.ud_class == "VERB")
{
if (item.Contains("VerbForm ", "Inf"))
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.INFINITIVE_ru;
}
else if (item.Contains("VerbForm", "Trans"))
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.GERUND_2_ru;
}
else
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.VERB_ru;
}
}
else if (item.ud_class == "ADP")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.PREPOS_ru;
}
else if (item.ud_class == "ADV")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.ADVERB_ru;
}
else if (item.ud_class == "PRON")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.PRONOUN_ru;
}
else if (item.ud_class == "PART")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.PARTICLE_ru;
}
else if (item.ud_class == "NUM")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.NUMBER_CLASS_ru;
}
else if (item.ud_class == "CONJ" || item.ud_class == "SCONJ")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.CONJ_ru;
}
else if (item.ud_class == "INTJ")
{
required_class = SolarixGrammarEngineNET.GrammarEngineAPI.PARTICIPLE_ru;
}
else
{
all_ok = false;
break;
}
foreach (string tv in item.ud_tags)
{
int coord_id = -1;
int state_id = -1;
string[] tv2 = tv.Split('=');
string tag = tv2[0].Trim();
string val = tv2[1].Trim();
if (tag == "Animacy")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.FORM_ru;
if (val == "Inan")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.INANIMATIVE_FORM_ru;
}
else
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.ANIMATIVE_FORM_ru;
}
}
if (tag == "Case")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.CASE_ru;
if (val == "Nom")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.NOMINATIVE_CASE_ru;
}
else if (val == "Acc")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.NOMINATIVE_CASE_ru;
}
else if (val == "Gen")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.GENITIVE_CASE_ru;
}
else if (val == "Ins")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.INSTRUMENTAL_CASE_ru;
}
else if (val == "Loc")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PREPOSITIVE_CASE_ru;
}
else if (val == "Dat")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.DATIVE_CASE_ru;
}
else if (val == "Par")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PARTITIVE_CASE_ru;
}
else if (val == "Voc")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.VOCATIVE_CASE_ru;
}
else
{
all_ok = false;
break;
}
}
if (tag == "Gender")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.GENDER_ru;
if (val == "Fem")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.FEMININE_GENDER_ru;
}
else if (val == "Masc")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.MASCULINE_GENDER_ru;
}
else if (val == "Neut")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.NEUTRAL_GENDER_ru;
}
else
{
all_ok = false;
break;
}
}
// Number = Sing
if (tag == "Number")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.NUMBER_ru;
if (val == "Sing")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.SINGULAR_NUMBER_ru;
}
else
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PLURAL_NUMBER_ru;
}
}
// Degree=Pos
if (tag == "Degree")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.COMPAR_FORM_ru;
if (val == "Pos")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.ATTRIBUTIVE_FORM_ru;
}
else if (val == "Sup")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.SUPERLATIVE_FORM_ru;
}
else if (val == "Cmp")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.COMPARATIVE_FORM_ru;
}
}
if (tag == "Variant" && val == "Brev")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.SHORTNESS_ru;
state_id = 1;
}
if (tag == "Person")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.PERSON_ru;
if (val == "1")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PERSON_1_ru;
}
else if (val == "2")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PERSON_2_ru;
}
else if (val == "3")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PERSON_3_ru;
}
}
if (tag == "Tense")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.TENSE_ru;
if (val == "Pres")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PRESENT_ru;
}
else if (val == "Past")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.PAST_ru;
}
else if (val == "Fut")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.FUTURE_ru;
}
}
if (tag == "Mood")
{
coord_id = SolarixGrammarEngineNET.GrammarEngineAPI.VERB_FORM_ru;
if (val == "Ind")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.VB_INF_ru;
}
else if (val == "Imp")
{
state_id = SolarixGrammarEngineNET.GrammarEngineAPI.VB_ORDER_ru;
}
}
if (coord_id != -1 && state_id != -1)
{
required_tags.Add(Tuple.Create(coord_id, state_id));
}
}
for (int iproj = 0; iproj < nproj; ++iproj)
{
int proj_score = 0;
int entry_id = SolarixGrammarEngineNET.GrammarEngine.sol_GetIEntry(projs, iproj);
int class_id = gren.GetEntryClass(entry_id);
if (class_id == required_class)
{
proj_score++;
}
foreach (var required_tag in required_tags)
{
int state_id = SolarixGrammarEngineNET.GrammarEngine.sol_GetProjCoordState(gren.GetEngineHandle(), projs, iproj, required_tag.Item1);
if (state_id == required_tag.Item2)
{
proj_score++;
}
}
proj_scores.Add(Tuple.Create(iproj, proj_score));
}
selected_proj = proj_scores.OrderByDescending(z => z.Item2).Select(z => z.Item1).First();
}
// теперь выбранную проекцию слова selected_proj векторизуем
WordData word_data = new WordData();
word_data.word = item.word;
word_data.word_index = word_index;
word_data.all_projs = tokenization[word_index];
word_data.entry_id = SolarixGrammarEngineNET.GrammarEngine.sol_GetIEntry(projs, selected_proj);
word_data.part_of_speech = gren.GetEntryClass(word_data.entry_id);
// Лемму определяем в несколько шагов.
// Прежде всего пытаемся взять наименование словарной статьи в лексиконе Solarix.
word_data.lemma = gren.GetEntryName(word_data.entry_id).ToLower();
// Если получили наменование статьи "???", то значит это несловарный элемент и
// надо брать лемму из корпуса.
if (word_data.lemma == "???")
{
word_data.lemma = item.ud_lemma;
}
else if (word_data.lemma == "NUMBER_")
{
// Для чисел берем исходный токен в качестве леммы.
word_data.lemma = item.word;
}
for (int j = 0; j < SolarixGrammarEngineNET.GrammarEngine.sol_GetProjCoordCount(gren.GetEngineHandle(), projs, selected_proj); ++j)
{
int coord_id = SolarixGrammarEngineNET.GrammarEngine.sol_GetProjCoordId(gren.GetEngineHandle(), projs, selected_proj, j);
int state_id = SolarixGrammarEngineNET.GrammarEngine.sol_GetProjStateId(gren.GetEngineHandle(), projs, selected_proj, j);
word_data.tags.Add(Tuple.Create(coord_id, state_id));
}
sent.AddWord(word_data);
SolarixGrammarEngineNET.GrammarEngine.sol_DeleteProjections(projs);
// DEBUG
//Console.WriteLine("\nDEBUG word={0}\ncorpus:{1}({2})\nselected:{3}", item.word, item.ud_class, string.Join(" ", item.ud_tags), GetTokenMorph(word_data, gren));
word_index++;
}
// Добавить <END>
WordData wd_end = new WordData();
wd_end.word_index = word_index;
wd_end.word = "END";
wd_end.entry_id = entry_end_id;
wd_end.part_of_speech = beth_class;
wd_end.all_projs = tokenization[word_index];
sent.AddWord(wd_end);
// Для отладки - выведем текстовое представление тегов в выбранной проекции.
if (all_ok)
{
yield return(sent);
}
}
}
items.Clear();
}
else
{
string[] tx = line.Trim().Split('\t');
ConlluItem item = new ConlluItem();
item.word = tx[word_column];
item.ud_lemma = tx[lemma_column];
item.ud_class = tx[class_column];
item.ud_tags = tx[tags_column].Split('|')
.Select(z => z.Split('='))
.Where(z => z.Length == 2)
.Select(z => z[0].Trim() + '=' + z[1].Trim())
.ToArray();
items.Add(item);
}
}
}
}
}
public IEnumerable <SentenceData> Read(SolarixGrammarEngineNET.GrammarEngine2 gren)
{
// Сначала надо получить из словарной список слов, которые дают однозначную лемматизацию.
HashSet <int> classes = new HashSet <int>();
foreach (var class_name in "СУЩЕСТВИТЕЛЬНОЕ ПРИЛАГАТЕЛЬНОЕ ГЛАГОЛ ИНФИНИТИВ ДЕЕПРИЧАСТИЕ НАРЕЧИЕ".Split(' '))
{
classes.Add(gren.FindPartOfSpeech(class_name));
}
HashSet <string> forms = new HashSet <string>();
MultiValueDictionary <string, string> form2lemma = new MultiValueDictionary <string, string>();
MultiValueDictionary <string, int> form2entry = new MultiValueDictionary <string, int>();
Console.WriteLine("Generating the list of words and lemmas...");
IntPtr hList = SolarixGrammarEngineNET.GrammarEngine.sol_ListEntries(hEngine: gren.GetEngineHandle(),
Flags: 0, EntryType: 0, Mask: ".+", Language: -1, PartOfSpeech: -1);
int nb_entries = SolarixGrammarEngineNET.GrammarEngine.sol_CountInts(hList);
for (int i = 0; i < nb_entries; ++i)
{
int id_entry = SolarixGrammarEngineNET.GrammarEngine.sol_GetInt(hList, i);
if (classes.Contains(gren.GetEntryClass(id_entry)))
{
string lemma = gren.GetEntryName(id_entry);
if (char.IsLetter(lemma[0]))
{
IntPtr hForms = SolarixGrammarEngineNET.GrammarEngine.sol_ListEntryForms(gren.GetEngineHandle(), id_entry);
int nb_forms = SolarixGrammarEngineNET.GrammarEngine.sol_CountStrings(hForms);
for (int iform = 0; iform < nb_forms; ++iform)
{
string form = SolarixGrammarEngineNET.GrammarEngine.sol_GetStringFX(hForms, iform);
form = gren.RestoreCasing(id_entry, form);
if (!form2lemma.Contains(form, lemma))
{
form2lemma.Add(form, lemma);
form2entry.Add(form, id_entry);
forms.Add(form);
}
}
SolarixGrammarEngineNET.GrammarEngine.sol_DeleteStrings(hForms);
}
if ((i % 10000) == 0)
{
Console.Write("{0}/{1}\r", i, nb_entries);
}
}
}
SolarixGrammarEngineNET.GrammarEngine.sol_DeleteInts(hList);
List <string> unambiguous_forms = forms.Where(z => form2lemma[z].Count == 1).ToList();
Console.WriteLine("{0} forms are good for lemmatization", unambiguous_forms.Count);
foreach (string word in unambiguous_forms)
{
//SolarixGrammarEngineNET.AnalysisResults morphology = gren.AnalyzeMorphology(word, -1);
SolarixGrammarEngineNET.AnalysisResults tokenization = gren.AnalyzeMorphology(word, -1, SolarixGrammarEngineNET.GrammarEngine.MorphologyFlags.SOL_GREN_TOKENIZE_ONLY, 0);
bool ok = true;
SentenceData sent = new SentenceData(word);
for (int i = 0; i < tokenization.Count; ++i)
{
if (i == 1 && !form2entry[word].Contains(tokenization[i].GetEntryID()))
{
ok = false;
}
WordData word_data = new WordData();
word_data.word = tokenization[i].GetWord();
word_data.word_index = tokenization[i].GetWordPosition();
word_data.entry_id = tokenization[i].GetEntryID();
word_data.part_of_speech = gren.GetEntryClass(word_data.entry_id);
word_data.all_projs = tokenization[i];
foreach (var tag in tokenization[i].GetPairs())
{
word_data.tags.Add(Tuple.Create(tag.CoordID, tag.StateID));
}
word_data.lemma = gren.GetEntryName(word_data.entry_id).ToLower();
if (word_data.lemma == "???")
{
word_data.lemma = word_data.word;
}
else if (word_data.lemma == "NUMBER_")
{
word_data.lemma = word_data.word;
}
sent.AddWord(word_data);
}
if (ok)
{
yield return(sent);
}
}
}