INSTALL.md

Installation Guide

This document deals with how to install libraries required for Pororo installation.

Common modules

• The libraries that should be installed in common for the use of Pororo are:
  • These libraries are installed when Pororo is installed through the pip install command, so you do not have to take additional action

requirements = [
    "torch==1.6.0",
    "torchvision==0.7.0",
    "pillow>=4.1.1",
    "fairseq>=0.10.2",
    "transformers>=4.0.0",
    "sentence_transformers>=0.4.1.2",
    "nltk>=3.5",
    "word2word",
    "wget",
    "joblib",
    "lxml",
    "g2p_en",
    "whoosh",
    "marisa-trie",
    "kss",
    'dataclasses; python_version<"3.7"',
]

Korean

• You may need to install additional libraries to perform specific tasks in Korean.

• python-mecab-ko is a library needed for the Korean Tokenization, PoS Tagging, Dependency Parsing, etc.

pip install python-mecab-ko==1.0.9

• kollocate is a library needed for the Korean Collocation task.

pip install kollocate

• koparadigm is a library needed for the Korean Morphological Inflection task.

pip install koparadigm

• g2pk is a library needed for the Korean Grapheme-to-Phoneme task.

pip install g2pk

Japanese

• You may need to install additional libraries to perform specific tasks in Japanese.

• fugashi and ipadic are the libraries needed for the Japanese RoBERTa model and the Japanese PoS Tagging.

pip install fugashi ipadic

• romkan is a library needed for the Japanese Grapheme-to-Phoneme task.

pip install romkan

Chinese

• You may need to install additional libraries to perform specific tasks in Chinese.

• g2pM is a library needed for the Chinese Grapheme-to-Phoneme task.

pip install g2pM

• jieba is a library needed for the Chinese PoS Tagging task.

pip install jieba

Etc.

Linux Supported Tasks

• Automatic Speech Recognition
• Speech Translation
• Optical Character Recognition
• Image Captioning

Automatic Speech Recognition

• To utilize the Automatic Speech Recognition module, wav2letter is required. asr-install.sh can be used for installation of th wav2letter
• Note! CUDA must be pre-installed for installation of the wav2letter.

bash asr-install.sh

Speech Synthesis

• To utilize the Speech Synthesis module, you need to install additional modules below.

requirements = [
    "editdistance==0.5.3",
    "epitran==1.2",
    "fastdtw==0.3.4",
    "future",
    "jieba==0.42.1",
    "librosa==0.7.0",
    "phonemizer==2.1",
    "Pillow==7.1.0",
    "pinyin==0.4.0",
    "scipy",
    "SoundFile==0.10.2",
    "numba==0.48",
    "ko_pron",
]

Optical Character Recognition

• To utilize the OCR module, you need to install the following libraries

apt-get install -y libgl1-mesa-glx

pip install opencv-python scikit-image