Uses frequency analysis to summarize text. Scores are assigned to sentences and sentences that are rich with relevant words will have higher scores. The highest scoring sentences are then printed in chronological order to produce a summary.
The following are all summaries of articles. If any of the links are dead, the content of each article can be found in the folder "sample_articles". These articles are all randomly selected and do not represent my opinions, views or what news websites I follow.
Generated Summary: Interviews with people familiar with Apple's inner workings reveal that the Mac is getting far less attention than it once did. People now have more options. There is now just one team, and most of the engineers are iOS first, giving the people working on the iPhone and iPad more power. In recent years, Apple managers have also become more likely to float two or more competing ideas, meaning designers and engineers must work on more than one concept at a time.
Generated Summary: The box jellyfish’s eyes are part of an almost endless variation of eyes in the animal kingdom. There are eyes with bifocal lenses, eyes with mirrors, and eyes that look up, down, and sideways all at the same time. Cup-shaped eyes with more receptors or compound eyes with additional cups can produce crude images of objects. There are thousands of different opsins, but they are all related.
Generated Summary: In 2016, according to a trends analysis based on this report by the Royal Society of London, the energy return on energy investment (EROEI) for oil appears to have fallen below a ratio of 15 to 1 globally. The higher energy returns for solar come as module efficiency, supply chain efficiency, and production and installation efficiency are all on the rise. Higher energy return on investment ratios for solar is one of the primary drivers enabling such low overall power prices.
Generated Summary: Search Google for something simple, like "buy term paper" or "buy academic paper online," and you end up with dozens of paper writing factories to choose from. The paper writer assigned to me would be an anonymous wordsmith. "Our clients are mostly undergraduate students. Graduate students (such as the online MBA students discussed above) are the second most common. PhD students are uncommon," he says.
Generated Summary: The NEA says that China will seek to add more than 110 gigawatts within the next three years, which could help the nation up the proportion of its renewable energy use to 20 percent by 2030. China’s geography certainly lends itself to large solar energy farms. Weaning itself off of fossil fuels will require quite a hefty investment; one that China appears ready to make.
Generated Summary: Google has developed and open-sourced a new JPEG algorithm that reduces file size by about 35 percent—or alternatively, image quality can be significantly improved while keeping file size constant.
Generated Summary: All the major technology companies are investing aggressively in AI software, including companies beyond IBM like Salesforce, SAP and Oracle that focus on business customers. Amazon, Google and Microsoft are the front-runners. "AI machines are only as smart as the data you give them," Mr. Kelly noted.
Generated Summary: The self-driving technology being sold by John Deere and some of its competitors are less technically complex than the fully driverless cars that big tech companies and car manufacturers are working on. There are no federal rules specifically addressing self-driving tech for tractors, largely because farm equipment is designed for use in fields where it doesn't pose the same level of risk to other vehicles or people as a self-driving vehicle on a public road. The systems are pricey: Outfitting a new tractor with top-of-the-line auto-steering, navigation and guidance tech could cost upwards of $20,000, Reed said.
Generated Summary: The great accomplishment of Jobs’s life is how effectively he put his idiosyncrasies—his petulance, his narcissism, and his rudeness—in the service of perfection. In his mind, what he did was special.
Generated Summary: Doctors in London say they have cured two babies of leukemia in the world’s first attempt to treat cancer with genetically engineered immune cells from a donor. "The patient could be treated immediately, as opposed to taking cells from a patient and manufacturing them," says Julianne Smith, vice president of CAR-T development for Cellectis, which specializes in supplying universal cells. "What they can do in the future is what we can do today," Nelsen said in an interview last year.
Usage: main.py <article.txt> <summary length> Explanation: Parameter 1: the location and name of the text to summarize Parameter 2: the number of sentences for the summary to contain
- Extract all the sentences from the text.
- Remove all the newlines (“\n”) from the text.
- Remove periods from all multi-period abbreviations.
- Example: change “Y.M.C.A.” to “YMCA” but leave single-period abbreviations like “Dr.” and “Mr.” alone.
- This reduces ambiguities and code complexity when it must be known if a period marks the end of a sentence.
- Split the text into sentences by dividing it in the locations where there are periods.
- Make sure no sentence has been incorrectly split in two due to abbreviations such as “Dr.”, “Mr.”, etc….
- Perform some more clean up such as whitespace removal and grouping words within quotes together.
- Remove transition sentences by scanning for keywords.
- Extract all the words from the text.
- Split the text into words by dividing it in the locations where there is whitespace (“ “).
- Trim words that have punctuation around it (“cat!” should be changed to “cat”).
- Group words that are technically the same (such as “berries” and “berry”).
- For each word, remove ownership (“schools’” and “school’s” should be changed to “school”).
- Convert plural words to singular.
- Assign a score to each word.
- Word scores are based on their frequency in the text.
- Ignore common words such as “the”, “to”, “and”, etc….
- Assign a score to each sentence.
- The sentence score is the sum of all the scores of its words divided by the number of words.
- Sentences with high information density will have high scores.
- Put the sentences with the highest scores together in chronological order to produce the summary.
While each module serves a function and contributes to the overall algorithm, they are also designed to work independently to serve their own particular function by themselves. For example, “extractor.py” is used by the Summarizer to extract sentences and words from text, but it can also be used independently to see exactly what it extracts. This is useful for debugging as it allows one to test each individual component separately.
This is used to extract sentences and words from text. It can be used to check if the sentences are being divided up correctly. It can also provide info about the text such as the sentence and word count.
Usage: extractor.py <article.txt> [parameter] Parameters: -i --info display basic info about <article.txt> -s --sentences extract sentences from <article.txt> -w --words extract words from <article.txt>
$ python3.5 extractor.py "sample_articles/apple.txt" -i Sentence count: 78 Word count: 1592
This module is used to manipulate words and sentences.
Usage: parser.py <word> [--parameter] parser.py <sentence> [--parameter] Parameters for <word>: -a --abbreviation remove all periods from an abbreviation -s --singular convert most words to singular and remove ownership -p --punctuation remove the surrounding punctuation -w --whitespace remove the surrounding whitespace Parameters for <sentence>: -a --abbreviation remove all periods from an abbreviation
$ python3.5 parser.py berries -s berry
This module is used for scoring the words and sentences.
Usage: scoring.py <article.txt> <parameter> <quantity> Parameters: -s print the top scoring sentences -w print the top scoring words
Example: $ python3.5 scoring.py "sample_articles/apple.txt" -w 5 Rank: Score: Content: #1. 35.0 apple #2. 24.0 mac #3. 17.0 with #4. 15.0 company #5. 15.0 more
This module was used to create the abbreviation lists that are referenced in the algorithm so it is not part of the algorithm itself.
Usage: tools.py [--options] Options: --create_abbr create abbreviations.txt from words.txt --create_abbr_multi create abbreviations_multi.txt from words.txt
This software was written for educational purposes and as a means for me to learn Python. Not all the summaries that are produced are as perfect as the samples shown above. Sometimes the summaries come out awkward and certain special characters can throw the algorithm off. Furthermore, titles and ads should be removed from an article before attempting to summarize it since the software does not automatically do that for you. For more accurate summary generation consider using NLTK.