6#ifndef DEEPLIMA_LEMMATIZATION_IMPL_H
7#define DEEPLIMA_LEMMATIZATION_IMPL_H
19namespace lemmatization
34 size_t buffer_size_per_thread
41 void init(
size_t max_input_word_len,
42 const std::vector<std::string>& class_names,
43 const std::vector<std::vector<std::string>>& class_values);
52 void predict(
const std::u32string& form,
54 std::u32string& target);
virtual ~LemmatizationImpl()=default
bool is_fixed(std::shared_ptr< StdMatrix< uint8_t > > classes, size_t idx)
std::vector< size_t > m_feat2cls
EmbdVectorizer m_feat_vectorizer
size_t m_upos_idx
index in the features matrix of the upos line
std::vector< bool > m_fixed_upos
index i is true if upos whose index is i is considered as fixed
EmbdVectorizer m_vectorizer
size_t m_max_input_word_len
max word length the encoder workbench / vectorizer were sized for (see init)
void init(size_t max_input_word_len, const std::vector< std::string > &class_names, const std::vector< std::vector< std::string > > &class_values)
void predict(const std::u32string &form, std::shared_ptr< StdMatrix< uint8_t > > classes, size_t idx, std::u32string &target)
morph_model::morph_feats_t get_morph_feats(std::shared_ptr< StdMatrix< uint8_t > > classes, size_t idx) const
virtual void load(const std::string &fn, const PathResolver &)
Encoding on one 64 bits integer of the set of morphological features for one token.