https://nlp.unibo.it/tag/argument-generation/Argument GenerationHugo Blox Builder (https://hugoblox.com)en-usTue, 13 Oct 2026 00:00:00 +0000https://nlp.unibo.it/media/logo.svghttps://nlp.unibo.it/tag/argument-generation/https://nlp.unibo.it/proposals/legal/argument-generation/Tue, 13 Oct 2026 00:00:00 +0000https://nlp.unibo.it/proposals/legal/argument-generation/<h2 id="context">Context</h2> <p>Legal argument mining identifies, extracts, and reconstructs the arguments of judicial texts automatically. It can reduce manual analysis and support legal research, document retrieval, summarization, and outcome prediction. In a judicial text, an argument is a chain of reasoning steps that leads to a conclusion.</p> <h2 id="objective">Objective</h2> <p>This proposal evaluates the ability of large language models to generate parts of an argumentative chain: the final conclusion or any missing intermediate step. The model may rely on related arguments or similar documents, given as demonstrations or retrieved with retrieval-augmented generation (RAG).</p> <figure class="lt-viz lt-svg" style="--lt-svg-min: 640px"> <svg viewBox="0 0 800 280" role="img" aria-labelledby="arggen-title" xmlns="http://www.w3.org/2000/svg"> <defs><marker id="arggen-arrow" viewBox="0 0 10 10" refX="8" refY="5" markerWidth="7" markerHeight="7" orient="auto-start-reverse"><path class="lt-s-head" d="M0 0 L10 5 L0 10 z"/></marker></defs> <title id="arggen-title">A legal argumentative chain of four reasoning steps: a premise, an intermediate step, a missing step, and the conclusion. A large language model, given related arguments or similar documents as demonstrations or through retrieval-augmented generation, generates the missing part. The missing part may also be the final conclusion.</title> <text class="lt-s-sub" x="20" y="22">AN ARGUMENTATIVE CHAIN WITH A GAP</text> <rect class="lt-s-tok" x="20" y="34" width="160" height="56" rx="10"/> <text class="lt-s-txt lt-s-strong" x="100" y="58">Premise</text> <text class="lt-s-txt lt-s-small" x="100" y="78">a fact of the case</text> <path class="lt-s-arrow" marker-end="url(#arggen-arrow)" d="M184 62 H216"/> <rect class="lt-s-tok" x="220" y="34" width="160" height="56" rx="10"/> <text class="lt-s-txt lt-s-strong" x="300" y="58">Step</text> <text class="lt-s-txt lt-s-small" x="300" y="78">intermediate</text> <path class="lt-s-arrow" marker-end="url(#arggen-arrow)" d="M384 62 H416"/> <rect class="lt-s-off" x="420" y="34" width="160" height="56" rx="10"/> <text class="lt-s-txt lt-s-strong" x="500" y="58">?</text> <text class="lt-s-txt lt-s-small" x="500" y="78">missing step</text> <path class="lt-s-arrow" marker-end="url(#arggen-arrow)" d="M584 62 H616"/> <rect class="lt-s-tok" x="620" y="34" width="160" height="56" rx="10"/> <text class="lt-s-txt lt-s-strong" x="700" y="58">Conclusion</text> <text class="lt-s-txt lt-s-small" x="700" y="78">may be missing too</text> <text class="lt-s-sub" x="20" y="150">RELATED MATERIAL</text> <rect class="lt-s-box lt-s-pred" x="20" y="160" width="260" height="76" rx="12"/> <text class="lt-s-txt lt-s-strong" x="150" y="186">Related arguments</text> <text class="lt-s-txt lt-s-strong" x="150" y="206">or similar documents</text> <text class="lt-s-txt lt-s-small" x="150" y="226">as demonstrations or with RAG</text> <path class="lt-s-arrow" marker-end="url(#arggen-arrow)" d="M284 198 H366"/> <rect class="lt-s-box lt-s-sel" x="370" y="160" width="260" height="76" rx="12"/> <text class="lt-s-txt lt-s-strong" x="500" y="192">Large language model</text> <text class="lt-s-txt lt-s-small" x="500" y="216">generates the missing part</text> <path class="lt-s-arrow" marker-end="url(#arggen-arrow)" d="M500 156 V94"/> <text class="lt-s-txt lt-s-small" x="400" y="268">The generated step is evaluated against the original chain.</text> </svg> <figcaption>Generation of a missing step in an illustrative argumentative chain, built from the description of the proposal.</figcaption> </figure>