Ahmed is investigating Neo4j for NLP

[asashour]

Hi all,

I was hoping to find (or build) something that automatically processes arbitrary text, and creates corresponding relations.

Since long journeys start with a single step, I am gradually investigating different technologies.

[oleg_neo4j]

Hi Asashour
What kind of text and what kind of relations are you hoping to find?

[asashour]

Hi Oleg,

Any English text, and relations possibly similar to ConceptNet ones

So far the NLP tools I checked, they return whether a word is a Noun/Verb/Adjective/etc, some return if it is a subject/object. But I think it would be helpful if we have the sentence transformed into relations.

E.g. John Smith (person, male, type of Human) is a worker (job, he can make furniture, has a average salary of X in England, he can afford to buy a car), and he goes to the workshop (place of work, has tools) everyday (persistent, mostly has experience).

Those relations are detected by parsing text, and the relations are collected form various sentences.

[oleg_neo4j]

Ok, have you looked into the Graphaware NLP plugin (https://github.com/graphaware/neo4j-nlp) for Neo4j? It will do some things with ConceptNet. There is also Stanford CoreNLP (https://corenlp.run), which can make trees out of sentences, and you can write rules for which parts of grammar exactly you want.