Cateye

A hint-enabled search engine framework for biomedical classification systems

Features

• Hint: Show hints for search terms which can narrow down the results fast.
• Fallback: If no result satisfying the query, the system automatically eliminates less important search terms.
• Spelling correction: Build-in spelling correction for query terms.
• Abbreviation expansion: Pre-defined abbreviation list will be automatically applied during the search
• Sorted results: Sort the results according to the search history.

Installation

$ git clone https://github.com/jeroyang/cateye.git
$ cd cateye
$ pip install -e .

Usage

1. Run the Demo Site:

$ FLASK_APP=application.py FLASK_ENV=development flask run

Then browse the local site http://127.0.0.1:5000/ Try to search "rhinitis"

2. Make your own site:

2-1. Check the constants.py:

Setup the essential variables in the constants.py: SITE_TITLE, SITE_SUBTITLE, TOKEN_FOLDER, SNIPPET_FOLDER, HINT_FOLDER, SPELLING_FILE, ABBREVIATION_FILE, INDEX_URL

The INDEX_URL will be used in the Shove object, which can be a local URL starts with file:// please check the document of Shove.

2-2. Data preparing

Folders overview:

• data: The data source for the search engine, all information in this subfolders using the term id as their filenames
• data/token: The tokens of the documents, after lemmatization
• data/snippet: The HTML snippets of the documents, which will be shown on the search results
• data/hint: The hints for each entity
• data/spelling.txt: The formal spelling of your tokens (before normalization). If possible, sort the tokens with the frequency of usage, the most common word the first.
• data/abbreviation.txt: The abbreviations, one line for one abbreviation pair, using tab to separate the short form and long form

Cateye include some very basic text processing tools: tokenizer (cateye.tokenize) and lemmatizer (cateye.lemmatize)

The tokenize function will be used in two places: the first place is to cut your documents into tokens, and the second place is to cut your query into tokens.

The lemmatizing function will normalize your tokens. If you wish to build a case-insensitive search engine, you may use lowercase lemmatizer on the tokens.

2-3. Build the index:

Run the command in the command line

$ cateye newindex

This command read the files in the token_folder and build an on-disk index in the index_url. It takes time depending on the size of your data.

2-4. Run your application:

$ FLASK_APP=application.py FLASK_ENV=development flask run

License

• Free software: MIT license