ufal / udpipe

UDPipe: Trainable pipeline for tokenizing, tagging, lemmatizing and parsing Universal Treebanks and other CoNLL-U files
Mozilla Public License 2.0
358 stars 75 forks source link

new requirement in parsers #178

Closed Shasetty closed 10 months ago

Shasetty commented 10 months ago
          Hi sir,

I totally respect the decision process of this form.

My main requirement, is to understand any given text, by using 1) common sense 2) grammar

As UDpipe is good in grammar, incorporating common sense in UDPIPE will make it complete.

lets us consider common sense step by step

1)To find the real meaning of any text, we have to:

a1) read them a2) analyze them a3) understand them

2)What are the real challenges in understanding any text?

Taking entire economy (USA or INDIA) as an input.

b1) We are not experts, in all the fields. b2) Information keeps flowing from all the fields.

C1) informations will be unique C2) hidden information C3) indirect information C4) providing unwanted information C5) and many more


3) Style of available text

d1) Big text with many paragraph d2) small text with few paragraph d3) 1 paragraph d4) 1 liners d5) single words

4) Availability of time

5) How to analyze the text to understand the text?

E1) problems are unique E2) solutions for the problem will be unique

E3) requirements are unique E4) solution for the requirement will be unique

E5) Decision process will be universal (common sense)

6) What is the universal decision process?

Decision process, is looking for; F1) problem F2) requirement

F3) solution

F4) path that leads for problem F5) path that leads for requirement F6) path that leads for solution

considering the above inputs , i would like to see the text ,analyze them and understand them as per point 6, and find the main word using parsers(grammar)

Originally posted by @Shasetty in https://github.com/ufal/udpipe/issues/177#issuecomment-1793277375

arademaker commented 10 months ago

Dear @Shasetty, please don't take me wrong, but I assume you don't have the necessary background to understand why #177 was closed.

Natural Language Processing is a massive area, including many different tasks. Given a piece of text, we usually understand that many steps must be done to transform text (a blob of symbols in a computer) to a representation that some application can manipulate or use directly. Here is a list of excellent textbooks that can help you be introduced to the area:

  1. https://web.stanford.edu/~jurafsky/slp3/
  2. https://www.amazon.com/Natural-Language-Understanding-James-Allen/dp/0805303340
  3. https://www.amazon.com/Linguistic-Fundamentals-Natural-Language-Processing/dp/1681736713
  4. https://www.amazon.com/Linguistic-Fundamentals-Natural-Language-Processing/dp/1681736713

To make a long story short. We don't do everything in one step. The repository here is about a statistical parser, a tool designed to take a string and produce a data structure representing one particular syntactic analysis in the theory of http://universaldependencies.org. UDPipe can already do too much: tokenization, segmentation into sentences, morphosyntactic analysis of the words, and parsing (syntactic analysis). By construction, these are the functionalities of UDPipe.

Other tools can take the output of UDPipe to produce further representations of semantics — for example, https://github.com/sivareddyg/UDepLambda. Indeed, we know that semantics can contribute to syntactic... see more in Allen's book above. However, UDPipe is not implemented following this principle of language processing.

So please, this is not an issue for this repository. You are not reporting a bug. You are not making a reasonable request for features. Please educate yourself in the area. Use other internet forums to discuss ideas for projects.

foxik commented 10 months ago

@arademaker Thank you very much for a very nice response! :heart: