2018 ProvidingASimpleQuestionAnsweri
- (Larson et al., 2018) ⇒ Tait Larson, Johnson Heng Gong, and Josh Daniel. (2018). “Providing A Simple Question Answering System By Mapping Questions to Questions.”
Subject Headings: Question Answering (QA) Task; Question Answering (QA) System
Notes
Cited By
Quotes
Abstract
Most current QA systems attempt to find answers to input questions by cleverly selecting portions of related documents. In this paper, we discuss an alternative approach for taking advantage of the large amount of pre-answered questions available on the web and finding a similar question that has already been answered. Our approach involves query transformation as breadth-first search with expansion by Wordnet and pruning by language model, in order to transform the question into the language of the question corpus. We use a combination of NLP and IR techniques to return the most relevant preanswered question. We discuss experimental results comparing our system to other QA systems.
Introduction
Algorithms/Method
Overall Description
Given a set of questions and answers as a knowledge base, we take as input a user's question phrased as one to several English sentences. We then attempt to map the user's question to the questions and answers in our repository. Our system returns a set of ranked relevant question and answer pairs to the user.
Note that this is distinctly different from traditional Information Retrieval in that we are attempting to map input questions to stored questions (that we already have answers to), as opposed to mapping input questions directly to documents.
The core of our approach is an attempt to transform input questions into the language of the question corpus, and then find the closest matching repository question. The transformation is done by a breadth-first search via rewriting of words in the input question according to Wordnet synonyms, and pruning the search tree by removing low probability rewritten sentences according to a language model trained on the question corpus. Once the input question has been rewritten into language close to that of the question corpus, we run tf-idf information retrieval algorithms with only bigrams to find the closest questions. If the final scores are too low, we determine that there is no matching question/answer pair in the repository. A diagram of the process is below.
Figure 1: Overall data flow for our Question ⇒ Question system.
References
;
| Author | volume | Date Value | title | type | journal | titleUrl | doi | note | year | |
|---|---|---|---|---|---|---|---|---|---|---|
| 2018 ProvidingASimpleQuestionAnsweri | Tait Larson Johnson Heng Gong Josh Daniel | Providing A Simple Question Answering System By Mapping Questions to Questions |