search

darkestfloyd / PNet

0 stars 1 forks source link

Paper #1

Closed darkestfloyd closed 6 years ago

darkestfloyd commented 6 years ago

Leave a comment with links to papers you think are relevant, with a small abstract for context.

darkestfloyd commented 6 years ago

A good blog post on NLP problems, may help in future. Mentioning for documentation purposes.

varun-sundar-rabindranath commented 6 years ago

From Natural Language Specifications to Program Input Parsers

The paper discusses techniques to automatically generate input parser code, that given a specification (as commonly found in programming platforms), generates code to parse the input as per the specification

varun-sundar-rabindranath commented 6 years ago

Using Semantic Unification to Generate Regular Expressions from Natural Language

The paper discusses techniques to convert natural language to regular expressions

varun-sundar-rabindranath commented 6 years ago

Base Paper - A Syntactic Neural Model for General-Purpose Code Generation

This paper discusses techniques to convert natural language descriptions (comments found in source code) to Python programs.

GitHub link: pcyin/NL2code

varun-sundar-rabindranath commented 6 years ago

Abstract Syntax Networks for Code Generation and Semantic Parsing

This paper is similar to the base paper, and confirms the results of the base paper. Though the goals are similar between the papers, the latter has tackled problems a little differently.

krishnakaranam3732 commented 6 years ago

A paper using semantic parsers for converting natural language to code on IFTTT dataset

This paper uses program retrieval, machine translation and generation without alignment to map the natural language descriptions to program representations if IFTTT recipes.

darkestfloyd commented 6 years ago

Incorporating Copying Mechanism in Sequence-to-Sequence Learning

The paper describes a CopyNet - a Seq2Seq network with some modification to add the ability of copying text from the input to the output. This could help copy variable names from the input. Although the base paper handles this internally, we could use the technique described here if we see that the network is not able to copy variable names from the input.

darkestfloyd commented 6 years ago

Base papers and data selected. Moving on to #5