This project used 100% python (v 3.7)

Features:

Corpus search tool:

Our tool can search in a corpus by:

• Ambiguous: you can search everything such as character, number, morpheme,…
• Noun (POS tagging)
• Verb (POS tagging)
• Adjective (POS tagging)
• Name of Person (NER model)
• Name of Location (NER model)
• Name of Organization (NER model)
• Show the top 10 similar words of your input (gensim word2Vec)

Corpus dataset:

• I did web scrapping and got 12k description lines on vnexpress.net

Libraries used:

• Dash + Dash bootstrap components
• Plotly
• Gensim
• Underthesea (now Underthesea requires pytorch 1.4.0)
• nltk
• numpy
• pandas
• statsmodels

How to run:

• Open terminal in the following directory: “Vietnamese-corpus-search-and-analysis-Web-app/“

  Using Corpus search app

• Run terminal “python src/app.py”

  python src/app.py

• Wait about 1 minute for the server, if you see the local host link in terminal, then ctrl click open it or copy and paste it into browser

Using corpus statistical analysis app

• Run terminal

  python src_statistics/app.py

• Wait about 1 minute for the server, if you see the local host link in terminal, then ctrl click open it or copy and paste it into browser

Using another corpus

• Rename your corpus file to “vn_express.txt” and replace it in resources/
• You have to run “python src/create_NER_pickle.py”, then type in your corpus’ directory: “resources/vn_express.txt” to build the NER model and Word2vec model, output as 2 files ner.pik and w2v.pik
• You only need to run once when using a new corpus

Folders structure:

• docs/: documentation folder
  • NLP.pptx: slides
• src/: source code of corpus search app
• src_statistics/: source code of corpus statistical analysis app
• resources/:
  • ner.pik: pickle file of NER model
  • w2v.pik: pickle file of Word2vec model
  • vn_express.txt: main corpus data
  • corpus_mini.txt: small 2k corpus for fast debugging
  • stop_words.txt: File contains Vietnamese stopwords

Demo

Corpus search tool

Statistical analysis tool