jowoni / team-project-team18

Korean-tag-recommeder

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Korean-tag-recommeder

A prototype program that recommends appropriate KOREAN HASHTAGS depending on Instagram body
Recommend word hashtags that are popular but not appearing in the post, using Word2Vec model.

인스타그램에 업로드할 본문을 입력하면, 해당 본문에 적절한 한글 해쉬태그를 추천하는 프로그램입니다.
Word2Vec을 이용한 모델로 많이 사용되면서도 본문에 등장하지 않는 단어를 태그로 추천합니다.

※ 'Instagram body' means Instagram post without hashtags and must include two or more nouns..

구현 범위 및 개발 방법

이 프로그램은 prototype으로, '필카'에 관련된 2개 이상의 명사를 가진 본문에 대한 태그만을 추천합니다.
※ '필카': '필름 카메라'의 준말 Short for film camera

  • 개발 언어: Python 3.6
  • 개발 툴: Anaconda, Jupytal Notebook, Pycharm
  • 협업 툴: Github, Slack
  • 사용 라이브러리: KoNLPy, Gensim

example:

input:

계단 그림자 기다리는 중. 필카의 묘미.

output:

#사람 #필카 #필카감성 #카메라 #친구 #코닥 #필름 #구닥 #그림자 #셀피 #일회용카메라  

Manual

Environment configuration

To use Word2Vec, Gensim must be installed.

Run in your terminal (recommended):

$ pip install --upgrade gensim

or, alternatively for conda environments:

$ conda install -c conda-forge gensim

For preprocessing of data, you need to install KoNLPy:

$ pip install --upgrade pip
$ conda -c conda-forge install jpype1

$ pip install konlpy

Testing

  1. Run Recommend_tags.py

    $ python Recommend_tags.py

  2. GUI appears.

start

  1. Write down your post on. Press Process.

write

  1. Check HASHTAGS RECOMMENDED!

result

  • If you don't want to remove results, uncheck "auto remove"

check


For more instructions of PREPROCESSING/MODELING, please read WIKI

What This Repository Contains

Workflow

  • Dataset1/ data for Model 1

    • `kowiki_data_tokenizing.py' Tokenize Korean Wiki text file
  • Dataset2/ data for Model 2

    • keyword_recomend_by_counted.py Choose keyword from insta_noun.bin by `insta_counted.bin'
    • create_keyword_tags.py create keyword_tags.bin for Model 2
  • Model1_word2vec.model Model 1 for recommending WORDS similar to the keywords

  • Model2_word2vec.model Model 2 for recommending TAGS related to keywords

  • make_model1.py Make Model 1

  • make_model2.py Make Model 2

  • model_test.py Check the results of the model created


Contribution

If you want to contribute on our project, please read Contribution Guide before Contributing.


About

Korean-tag-recommeder

License:MIT License


Languages

Language:Python 100.0%