RuOpinionNE-2024: Extraction of Opinion Tuples from Russian News Texts
Natalia Loukachevitch, Natalia Tkachenko, Anna Lapanitsyna, Mikhail Tikhomirov, Nicolay Rusnachenko
TL;DR
This work presents RuOpinionNE-2024, a shared task and dataset for extracting structured opinion tuples (holder, target, polarity, expression) from Russian news sentences. It builds on the RuSentNE corpus and defines a strict evaluation protocol (F1 with token-overlap criteria) over JSON Lines data, reporting results from over 100 submissions that utilized zero-shot, few-shot, and fine-tuned large language models. The top performance (F1 ≈ 0.41) comes from parameter-efficient fine-tuning of a very large model, with extensive exploration of prompts (30 prompts) and 11 models in 1-shot and 10-shot settings. The work provides a publicly available dataset, a rigorous evaluation framework, and practical insights into prompt design and model scaling for structured opinion extraction in Russian, paving the way for more nuanced media analysis.
Abstract
In this paper, we introduce the Dialogue Evaluation shared task on extraction of structured opinions from Russian news texts. The task of the contest is to extract opinion tuples for a given sentence; the tuples are composed of a sentiment holder, its target, an expression and sentiment from the holder to the target. In total, the task received more than 100 submissions. The participants experimented mainly with large language models in zero-shot, few-shot and fine-tuning formats. The best result on the test set was obtained with fine-tuning of a large language model. We also compared 30 prompts and 11 open source language models with 3-32 billion parameters in the 1-shot and 10-shot settings and found the best models and prompts.