cccccroge/NLP-chatbot

An easy NLP application running on Windows, using state-of-art deep learning models.

NLP Bot

An easy NLP application running on Windows, using state-of-art deep learning models.

功能

• 故事生成：選擇生成故事風-->輸入開頭文字-->生成該風格故事段落
• 情感分析：輸入文字-->回覆可能之情緒（回覆一至兩種，共13種）
• 經驗猜測：輸入文字-->產生數個可能猜測-->使用者給予回饋（是/否）

架構

注意

實作上目前僅提供local端的CPU運算，第一個及第三個功能將會花費5分鐘以上的時間

Project Development

Installation

Windows

official guide from Kivy.org

1. Install Python 3.7
2. Setup pip, wheel, virtualenv

python -m pip install --upgrade pip wheel setuptools virtualenv	// install
python -m virtualenv [your virtual environment name]		// create virtual env. (you don't need to execute this one, the project already has a virtual env.)
source [your virtual environment name]\Scripts\activate	// activate virtual env.

3. Install the dependencies for Kivy

python -m pip install docutils pygments pypiwin32 kivy_deps.sdl2==0.1.22 kivy_deps.glew==0.1.12

4. Install Kivy

python -m pip install kivy==1.11.1

5. Install application associated packages

python -m pip install tensorflow==1.14 regex

Linux

official guide from Kivy.org

1. Install Python 3.7 and pip

sudo apt-get update
sudo apt-get install python3-pip

2. Setup pip, wheel, virtualenv (note that wheel in linux might run into issues)

python3 -m pip install --upgrade --user pip setuptools virtualenv
python3 -m virtualenv ./[your_worn_venv]   // create virtual env at project root
source [your_worn_venv]/bin/activate	// activate virtual env.

3. Install Kivy

pip3 install kivy

4. Install application associated packages

pip3 install tensorflow==1.14 regex

Language model setup

1. for story telling

pip install gpt-2-simple

2. for emotion classification

pip install bert-tensorflow
pip install tensorflow_hub
pip install pandas

3. for experience guessing

pip install torch===1.3.1 torchvision===0.4.2 -f https://download.pytorch.org/whl/torch_stable.html
pip install pytorch_pretrained_bert
pip install nltk