t2hk/lexrank

LexRank

LexRank implementation using MarkLogic10 CNTK

This is the LexRank implementation for text summarization using CNTK of MarkLogic10.

Overview

This is the LexRank implementation for text summarization using CNTK of MarkLogic10. The word2vec is used for text vectorization. This program is implemented in XQuery.

Requirements

• MarkLogic10 with GPU

If you run as a web application, you need:

• vue.js(https://jp.vuejs.org/index.html)
• axios(https://github.com/axios/axios)
• vue-table-2(https://github.com/matfish2/vue-tables-2)
• bootstrap(https://getbootstrap.com/)

How to use

1. Load files you want to summarize into the MarkLogic.

   • The contents files are into ./news directory
   • See news_load.sh

2. Load word2vec onnx model and the vocabulary file into the MarkLogic.

   • See https://github.com/t2hk/marklogic_cntk_word2vec

3. Load this project files(xqy, html, js) into the module database of the MarkLogic.

4. Get the required libs and load into the module database of the MarkLogic.

   • Javascript files are into ./js directory.
   • CSS files are into ./css directory.

5. Access to the lexrank.html via the MarkLogic HTTP server.