README

Know everything, like Asimov did.

asimov is a python module that uses machine learning to help you keep abrest of new scientific journal articles.

General principle

asimov uses a corpus of text, that you identify as being interesting or not, to train a model whose job it is to predict if a new piece of text is something that you're interested in.

Usage

0. Configuration

Copy default.txt to config.txt and fill out the missing values. This information is needed for asimov to work. Then, import asimov:

import asimov

1. Generate training data

Training data consists of a list of abstracts (or any text that you want to train on), and a corresponding list of labels identifying whether each piece of text is to be trained for, or trained against.

asimov can generate positive training data from a BibTex file containing papers that you find interesting. Once you have exported all the papers you are interested in a .bib file, use asimov to pull some useful text from the .bib file:

asimov.get_cool_papers_from_bib()

You may not have abstracts for every paper in this .bib file. If this is the case, you can use PubMed to get these abtracts:

asimov.get_missing_abstracts()

Hopefully, this dataset is mostly complete, and somewhat useable.

Generating negative training data

asimov can also scan RSS feeds to determine a corpus of text to feed into the negative training data. To use this, you need to edit the file called boring-feeds.txt and add links to RSS feeds from sources that would never interest you. You might want to edit this file and remove entries that you might be interested in. Once this done,

asimov.get_boring_papers_from_rss()

Installation

Download this repository.

License

GPL v3