https://github.com/indianwhocodes/patternrec

Recurrent NN to process text as an intelligent dialogue agent applicable to deductive reasoning. Multiple feature sets were generated to progressively equip the system with textural reasoning. Testing on comprehension, robustness to accuracy for queries with transitivity, induction.
https://github.com/indianwhocodes/patternrec

Last synced: 19 days ago
JSON representation

Recurrent NN to process text as an intelligent dialogue agent applicable to deductive reasoning. Multiple feature sets were generated to progressively equip the system with textural reasoning. Testing on comprehension, robustness to accuracy for queries with transitivity, induction.

• Host: GitHub
• URL: https://github.com/indianwhocodes/patternrec
• Owner: indianwhocodes
• Created: 2018-07-12T05:33:04.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2021-03-25T22:36:07.000Z (almost 4 years ago)
• Last Synced: 2024-03-21T17:34:52.035Z (10 months ago)
• Language: Python
• Size: 404 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Overview

Recurrent NN to process text as an intelligent dialogue agent applicable to deductive reasoning. Multiple feature sets were generated to progressively equip the system with textural reasoning. Testing on comprehension, robustness to accuracy for queries with transitivity, induction.

### Goals

My project’s objective is to fabricate a system which involves building an entire AI system modularly to comprehend textual information, processing it as an intelligent dialogue agent which is applicable to deductive reasoning and natural language processing. 

### Specifics

- A plethora of pre-meditated tasks will be applied to evaluate comprehension of the designed system and its robustness to accuracy in terms of answering queries posed via transitivity, induction etc. 

- This project on its completion should provide a substantial baseline over which development of an automated system is possible which has the potential to converse intelligently with humans. The problem will be segregated into multiple feature sets that would progressively equip my system with textual reasoning.

- We will be parsing the words composited in a sentence to apposite vector representations to be utilized as an input to our learning model. I will also provide a comparative analysis of different learning models that would be implemented to train my system. 

### Technologies Used

Torch library for Lua will be used to develop my source code. A later extension would be to implement my source code in python by using Theano and Keras libraries.

### Resources

- [Presentation on Design](https://www.youtube.com/watch?v=HK7_A05vqwQ)