Information-retrieval (IR) based QA
knowledge-based QA

University of Michigan - Ann Arbor

Question Answering is a computer science discipline within the fields of information retrieval and natural language processing (NLP), which is concerned with building systems that automatically answer questions posed by humans in a natural language.

Question Answering

An on-going but a helpful book resource about NLP
https://web.stanford.edu/~jurafsky/slp3/

Speech and Language Processing (3rd ed. draft) 

Two Major Paradigms for QA 

The system relies on the massive information in the collected data resource. Given a user question, information retrieval is used to find relevant
passages. Then neural reading comprehension algorithms read these retrieved passages and draw an answer directly from spans of text.

Information-retrieval (IR) based QA

The system builds a semantic representation of the query and then use it to query databases of facts

Knowledge-based Question Answering (QA)

The system answers questions like “why” questions that require generating long
answers.

Long-form Question Answering(QA)

The system uses datasets of community-created question-answer pairs like Quora or Stack Overflow.

Community Question Answering(QA)

Factoid questions are questions that can be answered with simple facts expressed in short texts, which normally start with words including but not limited to 'when', 'what', and 'where'.

Factoid Questions

The fundamental structure of question answering is a simple correspondence between a given question and its corresponding answer. This direct relationship forms the basis of how QA tasks are conceptualized and implemented.

Question-Answer Format

Open-domain question answering is a task where a system is expected to return an answer in direct response to a user-submitted question, without being restricted to a specific knowledge base.

Open-Domain Question Answering

Prompt templates can be specifically designed for general-purpose question-answering tasks, providing a structured way to formulate queries for a system.

Prompt Templates for General-Purpose Question Answering

Question Answering is a significant discipline within Natural Language Processing because many other complex problems, including reasoning tasks, can be effectively reframed and solved using a question-answering format.

Importance of Question Answering in NLP

Multiple-choice question answering is a task format where a model is presented with a question and a set of predefined options, and it must select the correct answer from the given choices.

Multiple-Choice Question Answering

Based on the scenario provided, analyze this business goal as a Question Answering problem. Describe the primary input, the necessary information source, and the desired output. What is the fundamental challenge this system must solve that distinguishes it from a standard keyword search engine?

Analyzing a Customer Support Automation Task

A software company wants to develop an internal system to help employees quickly find factual information, such as 'What is the company's policy on remote work?' or 'Who is the project manager for Project Phoenix?'. All the necessary information is stored in a highly structured, internal corporate database. Which approach to building this system would be the most direct and efficient for retrieving these specific, factual answers?

Consider two user queries posed to an automated system:
1. 'What year was the Eiffel Tower completed?'
2. 'Why was the Eiffel Tower built?'

Explain the fundamental difference between these two queries from the perspective of designing a system to answer them. What challenges does the second query present that the first one does not?

Learn Before

Related