Download presentation
Presentation is loading. Please wait.
Published byJanice Chapman Modified over 9 years ago
1
Knowledge based Question Answering System Anurag Gautam Harshit Maheshwari
2
Introduction In order for the computers to interact with the users more naturally the computer must understand and automatically infer what the user wants to say from what he actually says. Our project is an attempt to achieve the same. We call the program : Sarvagya.
3
Our project is a question answering system based on statistical learning approach. We have used Probabilistic Latent Semantic Analysis (PLSA) algorithm to guess the most probable answer from the given set of documents
4
Usefulness As a user, you will probably not want to read the whole document to search for small answers. Time is a big factor in today’s life. You will no doubt want to save it. Document and knowledge management increase with QAS solution.
5
Query Extraction of important concepts from the query. Based on the concepts we search for the documents from wikipedia. Processing …. Our Approach
6
Tokenize the documents into sentences Formation of Term – Document Matrix (TD matrix) Pass the T-D matrix to PLSA Calculating cosine similarity among the document vectors returned by PLSA
7
PLSA – what it does ? PLSA T-D Matrix Document Clusters according to concepts
8
Some Results… We tested Sarvagya on various questions and got really close answers, some of them are shown:
12
Scope of Improvements Technique of extraction of concept from query can be improved. Data can be taken from some other sources also. We can use inbuilt knowledge database which will speed up the answer retreival.
13
Thank you
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.