Automatic text summarizer

Simple library and command line utility for extracting summary from HTML pages or plain texts. The package also contains simple evaluation framework for text summaries. Implemented summarization methods:

• Luhn - heurestic method, reference
• Edmundson heurestic method with previous statistic research, reference
• Latent Semantic Analysis, LSA - one of the algorithm from http://scholar.google.com/citations?user=0fTuW_YAAAAJ&hl=en I think the author is using more advanced algorithms now. Steinberger, J. a Ježek, K. Using latent semantic an and summary evaluation. In In Proceedings ISIM '04. 2004. S. 93-100.
• LexRank - Unsupervised approach inspired by algorithms PageRank and HITS, reference
• TextRank - some sort of combination of a few resources that I found on the internet. I really don't remember the sources. Probably Wikipedia and some papers in 1st page of Google :)
• SumBasic - Method that is often used as a baseline in the literature. Source: Read about SumBasic
• KL-Sum - Method that greedily adds sentences to a summary so long as it decreases the KL Divergence. Source: Read about KL-Sum

Here are some other summarizers:

• https://github.com/thavelick/summarize/ - Python, TF (very simple)
• Reduction - Python, TextRank (simple)
• Open Text Summarizer - C, TF without normalization
• Simple program that summarize text - Python, TF without normalization
• Intro to Computational Linguistics - Java, LexRank
• Sumtract: Second project for UW LING 572 - Python
• TextTeaser - Scala
• PyTeaser - TextTeaser port in Python
• Automatic Document Summarizer - Java, Bipartite HITS (no sources)
• Pythia - Python, LexRank & Centroid
• SWING - Ruby
• Topic Networks - R, topic models & bipartite graphs
• Almus: Automatic Text Summarizer - Java, LSA (without source code)
• Musutelsa - Java, LSA (always freezes)
• http://mff.bajecni.cz/index.php - C++
• MEAD - Perl, various methods + evaluation framework

Installation

Make sure you have Python 2.6+/3.2+ and pip (Windows, Linux) installed. Run simply (preferred way):

$ [sudo] pip install sumy

Or for the fresh version:

$ [sudo] pip install git+git://github.com/miso-belica/sumy.git

Or if you have to:

$ wget https://github.com/miso-belica/sumy/archive/master.zip # download the sources
$ unzip master.zip # extract the downloaded file
$ cd sumy-master/
$ [sudo] python setup.py install # install the package

Usage

Sumy contains command line utility for quick summarization of documents.

$ sumy lex-rank --length=10 --url=http://en.wikipedia.org/wiki/Automatic_summarization # what's summarization?
$ sumy luhn --language=czech --url=http://www.zdrojak.cz/clanky/automaticke-zabezpeceni/
$ sumy edmundson --language=czech --length=3% --url=http://cs.wikipedia.org/wiki/Bitva_u_Lipan
$ sumy --help # for more info

Various evaluation methods for some summarization method can be executed by commands below:

$ sumy_eval lex-rank reference_summary.txt --url=http://en.wikipedia.org/wiki/Automatic_summarization
$ sumy_eval lsa reference_summary.txt --language=czech --url=http://www.zdrojak.cz/clanky/automaticke-zabezpeceni/
$ sumy_eval edmundson reference_summary.txt --language=czech --url=http://cs.wikipedia.org/wiki/Bitva_u_Lipan
$ sumy_eval --help # for more info

Python API

Or you can use sumy like a library in your project.

# -*- coding: utf8 -*-

from __future__ import absolute_import
from __future__ import division, print_function, unicode_literals

from sumy.parsers.html import HtmlParser
from sumy.parsers.plaintext import PlaintextParser
from sumy.nlp.tokenizers import Tokenizer
from sumy.summarizers.lsa import LsaSummarizer as Summarizer
from sumy.nlp.stemmers import Stemmer
from sumy.utils import get_stop_words


LANGUAGE = "czech"
SENTENCES_COUNT = 10


if __name__ == "__main__":
    url = "http://www.zsstritezuct.estranky.cz/clanky/predmety/cteni/jak-naucit-dite-spravne-cist.html"
    parser = HtmlParser.from_url(url, Tokenizer(LANGUAGE))
    # or for plain text files
    # parser = PlaintextParser.from_file("document.txt", Tokenizer(LANGUAGE))
    stemmer = Stemmer(LANGUAGE)

    summarizer = Summarizer(stemmer)
    summarizer.stop_words = get_stop_words(LANGUAGE)

    for sentence in summarizer(parser.document, SENTENCES_COUNT):
        print(sentence)

Tests

Run tests via

$ py.test-2.6 && py.test-3.2 && py.test-2.7 && py.test-3.3 && py.test-3.4