Is there any method to summarize and gain the big picture from an article?

Question

Background

Recently I just develop a method to summarize and find the big picture of an article, especially the ones that are so abstract and introduce too many new terms. The method is as follow:

Step 1: Identify the central concepts in the text

One can do this by using a word frequency counter and list all most frequent words.

Step 2: Group all sentences that having the same subject into their basket

Let's say from step 1 we have three words that are most frequent: X, Y and Z. Then the sentences in the text should indicate the relationships between them. Then for basket X, search for every sentence whose subject is X. These sentences may have these forms:

• X is Y
• X does Y on Z
• X is Z-ed by Y
• X of Y is Z
• X is Y of Z
• etc.

Step 3: Simplify sentences

Sentences in the form "X is Z-ed by Y" can be change to "Y does Z on X"

Sentences in the form "X of Y is Z" usually have X is a verb in the noun form. In this case we can change the sentence to "Z X Y". For example:

Original sentence: The limitation of the universality is the particularity.
Reformed sentence: The particularity limits the universality.

Similarly, sentences in the form "X is Y of Z" usually have Y is a verb in the noun form. In this case we can change the sentence to "X Y Z". For example:

Original sentence: The particularity is the limitation of the individuality.
Reformed sentence: The particularity limits the individuality.

We can then rearrange the reformed sentences to their suitable basket. Read it again and then what the article says is much clearer.

Question

I suspect linguistics in general, and discourse analysis in specific may give me more insight on this. Can you give me some pointer to learn more?

Answer 1

First, automated summarization is indeed a task in computational linguistics, although I suspect (I'm not really close to that particular field) that they employ some neural networks now and no longer a rule-based system as sketched in the question.

The sketch in the question also touches another important task in computational linguistics and natural language processing, namely knowledge extraction.

Searching for those keyword will provide you a wealth of published literature, maybe more than you expected.