24 Questions
What is the approach used by AskMRS to answer queries?
Shallow approach that ignores hard documents and finds easy ones
What is the primary intuition behind query rewriting in AskMSR?
The user's question is often syntactically quite close to sentences that contain the answer
How many categories are used to classify questions in AskMSR?
7
What is the purpose of step 1 in AskMSR?
To rewrite queries to match sentences that contain the answer
What is the reported percentage of questions that top performing systems can answer?
70%
Which of the following is an example of a knowledge-rich approach?
Harabagiu, Moldovan et al.-SMU/UTD/LCC
What is the primary approach used by AskJeeves?
Pattern matching to match user questions to a knowledge base
What is the purpose of category-specific transformation rules in AskMSR?
To rewrite queries to match sentences that contain the answer
What is the primary focus of semantics and reasoning in QA?
Predicate-argument structure
Which of the following is not a key feature of Natural Language Processing?
Search Engine Optimization
What is the primary goal of inference in QA systems?
To justify an answer through lexical chains
What is the Stanford question answering dataset (SQuAD) used for?
Training neural models for reading comprehension
What is the purpose of mapping question and potential answer LFs in QA systems?
To find the best match between question and answer
What is the primary focus of predicate-argument structure in QA systems?
Detecting sentences with specific objects or close synonyms
What is the benefit of using inference in QA systems?
30% improvement in accuracy
What is the name of the process that involves analyzing sentence syntax and semantics to answer questions?
Syntactic analysis plus semantic
What is the purpose of query rewriting in Web question answering?
To rephrase a query to improve its search engine results
What is the primary function of a Web search engine in the Web question answering process?
To provide top N answers to rephrased queries
What is the purpose of mining N-grams in Web question answering?
To identify relevant keywords and phrases
What is the function of data-type filters in Web question answering?
To boost or lower scores of N-grams based on their matching with regular expressions
What is the limitation of the Web question answering technique described in the content?
It has a limited range of question categories and answer data types
What is the purpose of tiling the answers in Web question answering?
To determine the final answer to a question
What is the primary goal of the Web question answering process?
To determine the final answer to a question
What is the purpose of surface matching patterns in Web question answering?
To match a question with its corresponding answer pattern
Study Notes
Question Answering Systems
- Online QA examples include LCC, AnswerBus, EasyAsk, AnswerLogic, and AskJeeves
- AskJeeves is a hyped example of QA that uses pattern matching to match user questions to their own knowledge base of questions
Top Performing Systems
- Top performing systems can answer around 70% of questions
- Approaches include knowledge-rich approaches using many NLP techniques, and shallow approaches like AskMRS
AskMRS: Shallow Approach
- AskMRS ignores hard documents and finds easy ones
- Steps include:
- Query rewriting: classify question into 7 categories and apply transformation rules
- Sending rewrites to a web search engine and retrieving top N answers
- Mining N-grams: enumerating all N-grams in retrieved phrases and weighting by occurrence count
- Filtering N-grams: using data-type filters to boost scores of matching n-grams
- Tiling the answers: combining scores to get the final answer
Surface Matching Patterns
- Surface matching patterns use large collections of surface matching patterns (ISI)
- Important features include precision of recognition, coverage of name classes, and mapping into concept hierarchies
Semantics and Reasoning for QA
- Semantics and reasoning for QA involve analyzing sentence syntax and semantics
- Techniques include predicate-argument structure, syntax to logical forms, and inference
Neural Models for Reading Comprehension
- Neural models for reading comprehension use large-scale supervised datasets like the Stanford question answering dataset (SQuAD)
Find out what vacuum cleaner is recommended by Consumer Guide. Learn about online QA examples and question answering systems.
Make Your Own Quizzes and Flashcards
Convert your notes into interactive study material.
Get started for free