Snowball: extracting relations from large plain-text collections
DL '00 Proceedings of the fifth ACM conference on Digital libraries
Scaling question answering to the Web
Proceedings of the 10th international conference on World Wide Web
Learning search engine specific query transformations for question answering
Proceedings of the 10th international conference on World Wide Web
Mining the web for answers to natural language questions
Proceedings of the tenth international conference on Information and knowledge management
Context and Page Analysis for Improved Web Search
IEEE Internet Computing
Using Syntactic Dependency-Pairs Conflation to Improve Retrieval Performance in Spanish
CICLing '02 Proceedings of the Third International Conference on Computational Linguistics and Intelligent Text Processing
Learning surface text patterns for a Question Answering system
ACL '02 Proceedings of the 40th Annual Meeting on Association for Computational Linguistics
Answer formulation for question-answering
AI'03 Proceedings of the 16th Canadian society for computational studies of intelligence conference on Advances in artificial intelligence
Semantic similarity between sentences through approximate tree matching
IbPRIA'05 Proceedings of the Second Iberian conference on Pattern Recognition and Image Analysis - Volume Part II
Towards a multilingual QA system based on the web data redundancy
AWIC'05 Proceedings of the Third international conference on Advances in Web Intelligence
Machine learning for query formulation in question answering
Natural Language Engineering
| Hi-index | 0.00 |
In this paper, we present a method for the automatic acquisition of semantic-based reformulations from natural language questions. Our goal is to find useful and generic reformulation patterns, which can be used in our question answering system to find better candidate answers. We used 1343 examples of different types of questions and their corresponding answers from the TREC-8, TREC-9 and TREC-10 collection as training set. The system automatically extracts patterns from sentences retrieved from the Web based on syntactic tags and the semantic relations holding between the main arguments of the question and answer as defined in WordNet. Each extracted pattern is then assigned a weight according to its length, the distance between keywords, the answer sub-phrase score, and the level of semantic similarity between the extracted sentence and the question. The system differs from most other reformulation learning systems in its emphasis on semantic features. To evaluate the generated patterns, we used our own Web QA system and compared its results with manually created patterns and automatically generated ones. The evaluation on about 500 questions from TREC-11 shows comparable results in precision and MRR scores. Hence, no loss of quality was experienced, but no manual work is now necessary.