search

validas / NLX

Natural Language Expression
2 stars 0 forks source link

Question by: SAI about POS tag and triplets #1

Open freshNfunky opened 2 years ago

freshNfunky commented 2 years ago

Sir, I have the following doubts regarding the Semantic trainer:

  1. Does it POS tag on its own?
  2. What are the functionality does it have?

As I am thinking of switching the trainer with my own NLP model, before pushing it in the neo4j graph db.

Sir, I have 1 more query: Let's take an example sentance 'The machine replaces module-100, but reverts module 50' For the above sentance, what exactly is pushed into the graph db(are node-edge-node triplets being formed by the project)

Yours Sincerely, Sai Ashish Jana IIT Hyderabad

freshNfunky commented 2 years ago

Hi SAI,

to 1.: can you specify what you mean with POS tag? to 2.: what you mean with functionalities?

the Sentence-Tree currently does not branch in triplets. but just doublets. there is still an open Issue for me if triplets are really required and cannot be stacked by douplets. It depends on the closer look at conjunction words like "And" or "or" wheather the trainer can detect sub groups of the conjunction correctly...

PhantomScythe1738 commented 2 years ago

Hi, Regarding point 1, I wanted to know if the Semantic trainer classifies words into classes like noun/adjective/verb on its own or is it something that has to be done manually. Regarding point 2, I wanted to know what happens after the DOM-tree is created(like at which part does the data get inserted into the neo4j database) And is the insertion into neo4j triggered by simply building the project and creating an .nlx file or do we have to type another command.

freshNfunky commented 2 years ago

1.) no, there is a dictionary in the neo4j database. If the database does not know the word it is underlined with a blue jig-jag-line. so you set it in the first place. second opportunity is in the grammar trainer with the "Plus" button. Some words like "use" can have different type-roles according to the context. so use would be: Noun or Verb. According to the surrounding context it is detected either as Verb or as Noun. 2.) The DOM Tree in the lower left is used per sentence for the grammar trainer. Each sentence represents a statement on where the grammar trainer resolves the grammar tree. If the grammar trainer has found one unique root for the entire sentence, the sentence is resolved and therefore will in later versions be passed on the Entity-Relation parser to resolve an ontology over the entire Document.

PhantomScythe1738 commented 2 years ago

Hello Sir, Could you please clarify the following:

  1. Is there a domain model currently in neo4j?
  2. Are words/objects being pushed into the db(if so how are they being pushed as I presume the dictionary gets updated)?
freshNfunky commented 2 years ago

Hi Sai, to 1.) No. as i mentioned neo4j currently only hosts the dictionary and the grammar to 2.) as we did it on screen together. Unknown words get pushed into the dictionary via the NaturalLangQuicFixProvider they are underlined with a blue jig-jag line if unknown

PhantomScythe1738 commented 2 years ago

Okay sir, also can the neo4j be replaced by another application such as graphDB. Like I would like to find and store the POS tags that the semantic trainer outputs using GraphDB instead, in such a case, what all must be done?

freshNfunky commented 2 years ago

Moved to a new Ticket. Please open a new ticket for each individual question - thanks!