machine learning - Concept Based Text Summarization (Abstraction) -

i looking engine ai text summarization based on concept or meaning of sentence, looked @ open-source projects (ginger, paraphrase, ace) don't job. way work try find synonyms each word , replace current words, way generate alot of alternatives sentence meaning wrong of times.

i have worked stanford's engine highlights article , based on extract important sentences, still not abstraction, extraction.

it make sense engine i'm looking learns on time , results improved after each summary.

please out here, appreciated!

i don’t know open source project fits requirements abstraction , meaning assume.

but have ideas how build such engine , how train it.

in few words think keep in mind bayesian-network structure in our minds, helps not classify data, form abstract meaning text or message.

since impossible extract abstract categories structure our mind think it’s better build mechanism allow reconstruct step-by-step.

abstract

the key idea of proposed solution in extraction of meaning of conversation using approaches easier in operation automated computer system. allow creating level of illusion of real conversation person.

proposed model supports 2 levels of abstraction:

first of them, less complex level consists in recognition of groups of words or single word group related category, instance or instance attribute.

instance means instantiation general category of real or abstract subject, object, action, attribute or other kind of instances. example – concrete relation between 2 or more subjects: concrete relations between employer , employee, concrete city , country it’s situated , on. basic meaning recognition approach allows create bot ability sustain conversation. ability based on recognition of basic elements of meaning: categories, instances , instances attributes.

second, complicated method based on scenario recognition , storing them conversation context instances/categories using them completion of recognized scenarios.

related scenarios used complete next message of conversation of scenarios can used generate next message or recognizing meaning element using of conditions , using meaning elements context.

something that:

basic classification should entered manually , future correction/addition of teachers.

words sentence in conversation , scenarios sentence can filled context

conversation scenarios/categories can fulfilled recognized instances or instances described in future conversation (self-learning)

pic 1 – word detection/categorization flow vision

pic 2 – general system vision big picture view

pic 3 - meaning element classification

pic 4 – categories structure that