Understanding about Text Summarization and Implementing various Models used for Text Summarization
Read Notebook Summarise Notebook.ipynb
The seq2seq models encodes the content of an article (encoder input) and one character (decoder input) from the summarized text to predict the next character in the summarized text
The implementation can be found in keras_text_summarization/library/seq2seq.py
There are three variants of seq2seq model implemented for the text summarization
- Seq2SeqSummarizer (one hot encoding)
- training: run demo/seq2seq_train.py
- prediction: demo code is available in demo/seq2seq_predict.py
- Seq2SeqGloVeSummarizer (GloVe encoding for encoder input)
- training: run demo/seq2seq_glove_train.py
- prediction: demo code is available in demo/seq2seq_glove_predict.py
- Seq2SeqGloVeSummarizerV2 (GloVe encoding for both encoder input and decoder input)
- training: run demo/seq2seq_glove_v2_train.py
- prediction: demo code is available in demo/seq2seq_glove_v2_predict.py
There are currently 3 other encoder-decoder recurrent models based on some recommendation here
The implementation can be found in keras_text_summarization/library/rnn.py
- One-Shot RNN (OneShotRNN in rnn.py):
The one-shot RNN is a very simple encoder-decoder recurrent network model which encodes the content of an article and decodes the entire content of the summarized text
- training: run demo/one_hot_rnn_train.py
- prediction: run demo/one_hot_rnn_predict.py
- Recursive RNN 1 (RecursiveRNN1 in rnn.py):
The recursive RNN 1 takes the artcile content and the current built-up summarized text to predict the next character of the summarized text.
- training: run demo/recursive_rnn_v1_train.py
- prediction: run demo/recursive_rnn_v1_predict.py
- Recursive RNN 2 (RecursiveRNN2 in rnn.py):
The recursive RNN 2 takes the article content and the current built-up summarized text to predict the next character of the summarized text + one layer of LSTM decoder.
- training: run demo/recursive_rnn_v2_train.py
- prediction: run demo/recursive_rnn_v2_predict.py
The trained models are available in the demo/models folder
After a week of nonstop criticism from Democrats and Republicans alike for comments many condemned as racially charged, Donald Trump claims to be altering his campaign to be a little more inclusive. While the presumptive G.O.P. has long promised to “make America great again,” Trump now says he’s adding two words to slogan to illustrate just how non-racist he really is.
“You know, I have the theme ‘make America great again,’ and I've added a couple of things,” Trump announced to supporters at a campaign rally in Richmond, Virginia, on Friday night. “Right now I’m adding make America great again—I’m adding ‘for everyone,’ because it’s really going to be for everyone. It’s not going to be for a group of people, it’s going to be for everyone. It’s true.”
The allegedly amended slogan, which has yet to appear on any official signage or Trump merchandise, comes after the presidential candidate spent the first half of June repeatedly denouncing Gonzalo Curiel, the federal judge of Mexican heritage presiding over the Trump University class action lawsuit, as inherently biased against him. (Curiel was born in Indiana.) His comments were widely condemned by the Washington political establishment, including Senate Minority Leader Mitch McConnell, who suggested he may be an idiot, and House Speaker Paul Ryan, who called Trump’s statement the “textbook definition of a racist comment.”
Trump, who hasn’t apologized or taken back any of his comments, indicated on Friday that he realized his words have had a negative effect on his campaign and declared he is not a racist.
“I am the least racist person. The least racist person that you’ve ever seen. I mean give me a break,” he said at the rally. “I am the least racist person that you’ve ever looked at, believe me.”
Generated Headline: donald trump is changing his campaign refusal to blame it get i
Original Headline: Donald Trump Is Changing His Campaign Slogan to Prove He’s Not Racist
Lavrov and Kerry discuss Syrian settlement October 28, 2016 TASS Russian Foreign Ministry Sergey Lavrov and U.S. Secretary of State John Kerry discussed the Syrian settlement as well as the situation in Yemen and Libya by telephone on Oct. 28. The Russian Foreign Ministry said that the conversation had taken place at the U.S. side’s request. "The foreign policy chiefs continued discussing ways of settling the Syrian conflict, including the normalization of the situation around Aleppo, with account taken of fundamental approaches contained in the previously reached Russian-U.S. agreements. For that, the United States should ultimately separate moderate opposition (in Syria) from terror groups," the Russian Foreign Ministry said. "Lavrov and Kerry also discussed assistance to the solution of crises in Yemen and Libya as well as separate issues of bilateral agenda," the Russian Foreign Ministry stressed.
Generated Headline: lavrov and kerry discuss syrian war
Original Headline: Lavrov and Kerry discuss Syrian settlement
pip install requirements.txt
cd demo
For training make sure you have tensorflow-gpu and CUDA installed
There are already pre-trained models available at (demo/models) it can be used directly or you can train on your own
To train a deep learning model, say Seq2SeqSummarizer, run the following commands:
python seq2seq_train.pyTo summarise a sequence of sample texts run
python main.pyTo summarise one sentence of your own run
python main1sent.pyAlso for train and validation there is Accuracy and Loss graphs available at [demo/reports]
Data source : https://github.com/GeorgeMcIntire/fake_real_news_dataset
- https://machinelearningmastery.com/encoder-decoder-models-text-summarization-keras/
- https://github.com/chen0040/keras-text-summarization
* Other Models wil soon be updated
* Android App and Chrome Extension will soon be launched for Summarization of Articles/Blogs