End-to-end Argument Generation System in Debating

Abstract

We introduce an argument generation system in debating, one that is based on sentence retrieval. Users can specify a motion such as This house should ban gambling, and a stance on whether the system agrees or disagrees with the motion. Then the system outputs three argument paragraphs based on "values" automatically decided by the system. The "value" indicates a topic that is considered as a positive or negative for people or communities, such as health and education. Each paragraph is related to one value and composed of about seven sentences. An evaluation over 50 motions from a popular debate website showed that the generated arguments are understandable in 64 paragraphs out of 150.

StruAP: A Tool for Bundling Linguistic Trees through Structure-based Abstract Pattern

“…We have developed 657 patterns in total, and the number of words in the lexicons used for the patterns is 6406. Especially, our development of end-to-end argument generation system in debating (Sato et al, 2015) relied on the proposed tool. We conclude that the proposed model is effective when it is necessary to newly define case-specific relations and non-researchers are involved in development of the relation extraction.…”

Sequence-to-Sequence Learning for Motion-Aware Claim Generation

“…Even though the task of a broad generative model is gaining significant progress, argumentation generation task with a deep generative model is still under-developed. Besides our current work, the only work we found on this task has been done in [8]. They used a large text data from Gigaword corpus [27] and annotated it on the pre-processing step.…”

“…We cannot make an apple-to-apple comparison to any previous work due to the different nature of the dataset. A research conducted before [8] focuses on the information retrieval technique instead of the generation technique. Hence, the dataset is also different.…”

“…Despite much proliferation in the argumentation recognition research, the argument generation is still not well developed. To the best of our knowledge, the only study that focused on topic-dependent argument generation is done in [8]. The paper attempted to generate argument using a sentence retrieval method.…”

Neural Argument Generation Augmented with Externally Retrieved Evidence

“…To date, progress made in argument generation has been limited to retrieval-based methodsarguments are ranked based on relevance to a given topic, then the top ones are selected for inclusion in the output (Rinott et al, 2015;Wachsmuth et al, 2017;Hua and Wang, 2017). Although sentence ordering algorithms are developed for information structuring (Sato et al, 2015;Reisert et al, 2015), existing methods lack the ability of synthesizing information from different resources, leading to redundancy and incoherence in the output. In general, the task of argument generation presents numerous challenges, ranging from aggregating supporting evidence to generating text with coherent logical structure.…”