BOWV-summarization is a automatic summarization system, which is base on the bag of word vectors (BOWV) representation.
BOWV is a semantic representation, which can represent words, phrases, sentences, paragraphs and documents in a same vector space. In the BOWV space, we can approximately measure the semantic similarity with the distance between BOWV representations. We utilize it for extractive automatic summarization.
##Usage
summarization.py [-k sentence-num] [-v word-vector-version] file
file: the text file to be obtained a summary
sentence-num: the number of extracted sentence, default is 5
word-vector-version: 0-5, default is 0 ...
0 glove.6B.50d.txt
1 glove.6B.100d.txt
2 glove.6B.200d.txt
3 glove.6B.300d.txt
4 glove.42B.300d.txt
5 glove.840B.300d.txt
##Examples
$python summarization.py example/dlbook-c1.md
The automatic extractive summary of the 1st chapter in the deep learning book from Yoshua Bengio.
Deep learning is a particular kind of machine learning that achieves great power and flexibility by learning to represent the world as a nested hierarchy of concepts, with each concept defined in relation to simpler concepts. Fig 1 shows how a deep learning system can represent the concept of an image of a person by combining simpler concepts, such as corners and contours, which are in turn defined in terms of edges. Deep learning solves this central problem in representation learning by introducing representations that are expressed in terms of other, simpler representations. A major source of difficulty in many real-world artificial intelligence applications is that many of the factors of variation influence every single piece of data we are able to observe. The introduction of machine learning allowed computers to tackle problems involving knowledge of the real world and make decisions that appear subjective.
$python summarization.py -v 4 -k 4 example/news.md
The automatic extractive summary of a news from CNN.
Bank of Tokyo-Mitsubishi UFJ is trying out "Nao," a customer service robot that answers basic questions and is designed to speak 19 languages. A growing number of Japanese businesses are testing out robots as a possible solution to the country's shrinking workforce. But he says the technology is evolving quickly and someday, robots like Chihira could replace humans for certain jobs. The regular greeter, Ayako Seiryu, says she's not worried about a robot replacing her -- even one made to resemble a real 32-year-old woman.