Skip to content

YaleDHLab/intertext

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Intertext

Detect and visualize text reuse within collections of plain text or XML documents.

Intertext uses machine learning and interactive visualizations to identify and display intertextual patterns in text collections. The text processing is based on minhashing vectorized strings and the web viewer is based on interactive React components. [Demo]

App preview

Installation

To install Intertext, run the steps below:

# optional: install Anaconda and set up conda virtual environment
conda create --name intertext python=3.7
conda activate intertext

# install the package
pip uninstall intertext -y
pip install https://github.com/yaledhlab/intertext/archive/master.zip

Usage

# search for intertextuality in some documents
intertext --infiles "sample_data/texts/*.txt"

# serve output
python -m http.server 8000

Then open a web browser to http://localhost:8000/output and you'll see any intertextualities the engine discovered!

CUDA Acceleration

To enable Cuda acceleration, we recommend using the following steps when installing the module:

# set up conda virtual environment
conda create --name intertext python=3.7
conda activate intertext

# set up cuda and cupy
conda install cudatoolkit
conda install -c conda-forge cupy

# install the package
pip uninstall intertext -y
pip install https://github.com/yaledhlab/intertext/archive/master.zip

Providing Metadata

To indicate the author and title of matching texts, one should pass the flag to a metadata file to the intertext command, e.g.

intertext --infiles "sample_data/texts/*.txt" --metadata "sample_data/metadata.json"

Metadata files should be JSON files with the following format:

{
  "a.xml": {
    "author": "Author A",
    "title": "Title A",
    "year": 1751,
    "url": "https://google.com?text=a.xml"
  },
  "b.xml": {
    "author": "Author B",
    "title": "Title B",
    "year": 1753,
    "url": "https://google.com?text=b.xml"
  }
}

Deeplinking

If your text documents can be read on another website, you can add a url attribute to each of your files within your metadata JSON file (see example above).

If your documents are XML files and you would like to deeplink to specific pages within a reading environment, you can use the --xml_page_tag flag to designate the tag within which page breaks are identified. Additionally, you should include $PAGE_ID in the url attribute for the given file within your metadata file, e.g.

{
  "a.xml": {
    "author": "Author A",
    "title": "Title A",
    "year": 1751,
    "url": "https://google.com?text=a.xml&page=$PAGE_ID"
  },
  "b.xml": {
    "author": "Author B",
    "title": "Title B",
    "year": 1753,
    "url": "https://google.com?text=b.xml&page=$PAGE_ID"
  }
}

If your page ids are specified within an attribute in the --xml_page_tag tag, you can specify the relevant attribute using the --xml_page_attr flag.