COSTRA 1.0: A Dataset of Complex Sentence Transformations

TL;DR

COSTRA 1.0 is a new dataset of complex Czech sentence transformations designed to evaluate sentence embeddings beyond simple paraphrasing, with potential applications across multiple languages.

Contribution

The paper introduces COSTRA 1.0, a novel dataset for complex sentence transformations, enabling advanced semantic analysis of sentence embeddings.

Findings

LASER embeddings do not exhibit expected properties in the dataset

Dataset includes 4,262 sentences with 15 transformation types

Potential for cross-lingual application and semantic property testing

Abstract

We present COSTRA 1.0, a dataset of complex sentence transformations. The dataset is intended for the study of sentence-level embeddings beyond simple word alternations or standard paraphrasing. This first version of the dataset is limited to sentences in Czech but the construction method is universal and we plan to use it also for other languages. The dataset consist of 4,262 unique sentences with average length of 10 words, illustrating 15 types of modifications such as simplification, generalization, or formal and informal language variation. The hope is that with this dataset, we should be able to test semantic properties of sentence embeddings and perhaps even to find some topologically interesting 'skeleton' in the sentence embedding space. A preliminary analysis using LASER, multi-purpose multi-lingual sentence embeddings suggests that the LASER space does not exhibit the desired properties.