we can classify the text based on the required parameters using various classification algorithms which data mining uses and then based on the class of the text we may use the ontologies which we might have developed beforehand or train the system according to the test cases

It's an interesting but difficult task, great if you are going to tackle it.

I would recommend reading the following:

Summarizing Online Forum Discussions – Can Dialog Acts of Individual Messages Help? (EMNLP 2014)

An example of using ML from colleagues from Kyushu Uni:

Conversation summarization using machine learning and scoring method (PACLING 2013)

Because the conversation structure is usually lost, I'd like to see more "John explained his difficult family situation and Mary expressed sorry for him" -like summarizations of long conversations. I think you'll find hints in currently popular argument-related papers as

Using Summarization to Discover Argument Facets in Online Idealogical Dialog (NAACL 2015).

Good luck!

A nice overview with R script and data can be found here:

https://www.hackerearth.com/practice/machine-learning/advanced-techniques/text-mining-feature-engineering-r/tutorial/

Also have a look at the challenges from that page for more data related to text mining.

first of all read the books suggested then read the papers

you have to read about Discourse Analysis in NLP

Refer below articles

https://arxiv.org/pdf/1707.02268

https://www.technologyreview.com/s/607828/an-algorithm-summarizes-lengthy-text-surprisingly-well/

https://dl.acm.org/citation.cfm?id=1613813

https://www.researchgate.net/post/How_can_we_summarize_conversation_using_machine_learning_and_data_mining_techniques

Try to summarize it in the same way documents are summarized. Simply, Google documents summarization and you will be guided to your desire.