Ein Spezielles Problem Der Frage - Antwort-Beziehungen: Fragen Mit Pluralischen NP

Rudi ConradCOLING 1982

TL;DR

This paper tackles the challenge of accurately matching questions with their answers in artificial dialogue systems, specifically focusing on questions involving plural noun phrases. The authors propose a framework for defining "direct, genuine answers" and explore the complexities that arise when these answers don't perfectly adhere to a strict syntactic structure, suggesting a solution based on semantic networks.

What Problem Does This Paper Address?

Building effective artificial dialogue systems hinges on their ability to correctly pair questions with relevant answers. This paper hones in on a specific, yet crucial, aspect of this problem: how to handle questions that inquire about multiple items or individuals (plural noun phrases). The authors identify that a key challenge lies in defining what constitutes a "direct, genuine answer." They propose that a valid answer must contain a predefined structure (the "answer schema") and replace a "question unknown" with elements from the queried class. However, they observe that real-world answers often deviate from this strict definition, leading to difficulties in accurate question-answer matching. This ambiguity is particularly pronounced with plural noun phrases, where answers might list multiple items or provide partial information. The paper aims to clarify these complexities and propose a robust method for handling them.

Methods

The paper introduces a formal definition of a "direct, genuine answer" based on two components:

  1. Answer Schema: A predefined structure that the answer sentence must conform to.
  2. Question Unknown: A placeholder in the question that is replaced by one or more elements from the queried class in the answer.

The authors then explore how this definition applies to questions with plural noun phrases. They analyze cases where answers do strictly adhere to the schema and, more importantly, cases where they don't. The proposed solution for handling these less-than-perfect answers involves representing semantic structures using semantic networks. This approach allows for a more flexible and nuanced understanding of the relationship between questions and answers, moving beyond purely syntactic matching.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The "Direct, Genuine Answer" is More Complex Than It Seems: The paper highlights that a strict, purely syntactic definition of a "direct, genuine answer" is insufficient for real-world dialogue systems, especially when dealing with plural noun phrases.
  • Plural Noun Phrases Introduce Ambiguity: Questions like "Who among you can speak multiple languages?" or "What is your parents' profession?" can have answers that are technically correct but don't perfectly fit a rigid structural mold.
  • Beyond Syntax: The Need for Semantics: The authors implicitly argue that to accurately match questions and answers, especially in complex cases, a deeper semantic understanding is required. This points towards the limitations of purely syntactic parsing for this task.
  • Semantic Networks as a Solution: The paper proposes semantic networks as a promising way to represent the underlying meaning and relationships, enabling more robust question-answer matching even when answers are not perfectly structured.

Key Results

The paper doesn't present quantitative results in the form of accuracy scores or performance metrics. Instead, its key results are conceptual:

  • It clearly articulates the problem of handling non-strictly conforming answers to questions with plural noun phrases.
  • It provides examples of such questions and their varied answers, illustrating the challenge.
  • It proposes a direction for a solution by suggesting the use of semantic networks for representing semantic structures.

Limitations and Future Work

The paper acknowledges that the proposed framework for defining "direct, genuine answers" leads to specific problems when dealing with plural noun phrases. The authors suggest that these problems are discussed and a solution is proposed. Future work would likely involve implementing and evaluating the proposed semantic network-based solution to demonstrate its effectiveness in practice. The paper also raises questions about whether such questions should be considered ambiguous or if this is a specific phenomenon of vagueness, implying further theoretical exploration is needed.

Summary updated

BibTeX
@inproceedings{conrad1982ein,
  author = {Rudi Conrad},
  title = {{Ein Spezielles Problem Der Frage - Antwort-Beziehungen: Fragen Mit Pluralischen NP}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1982},
}

Loading related content...

Figure