Automatic evaluation systems in the field of automatic summarization have been relying on the availability of gold standard summaries for over ten years. Gold standard summaries are expensive to obtain and often require the availability of domain experts to achieve high quality. In this paper, we propose an alternative evaluation approach based on pairwise preferences of sentences. In comparison to gold standard summaries, they are simpler and cheaper to obtain. In our experiments, we show that humans are able to provide useful feedback in the form of pairwise preferences. The new framework performs better than the three most popular versions of ROUGE with less expensive human input. We also show that our framework can reuse already available evaluation data and achieve even better results.

Most automatic text summarization systems proposed to date rely on centrality and structural features as indicators for information importance. In this paper, we argue that these features cannot reliably detect important information in heterogeneous document collections. Instead, we propose CPSum, a summarizer that learns the importance of information objects from a background source. Our hypothesis is tested on a multi-document corpus where we remove centrality and structural features. CPSum proves to be able to perform well in this challenging scenario whereas reference systems fail.

Answer selection aims at identifying the correct answer for a given question from a set of potentially correct answers. Contrary to previous works, which typically focus on the semantic similarity between a question and its answer, our hypothesis is that question-answer pairs are often in analogical relation to each other. Using analogical inference as our use case, we propose a framework and a neural network architecture for learning dedicated sentence embeddings that preserve analogical properties in the semantic space. We evaluate the proposed method on benchmark datasets for answer selection and demonstrate that our sentence embeddings indeed capture analogical properties better than conventional embeddings, and that analogy-based question answering outperforms a comparable similaritybased technique.

The task of automatic text summarization is to generate a short text that summarizes the most important information in a given set of documents. Sentence regression is an emerging branch in automatic text summarizations. Its key idea is to estimate the importance of information via learned utility scores for individual sentences. These scores are then used for selecting sentences from the source documents, typically according to a greedy selection strategy. Recently proposed state-ofthe-art models learn to predict ROUGE recall scores of individual sentences, which seems reasonable since the final summaries are evaluated according to ROUGE recall. In this paper, we show in extensive experiments that following this intuition leads to suboptimal results and that learning to predict ROUGE precision scores leads to better results. The crucial difference is to aim not at covering as much information as possible but at wasting as little space as possible in every greedy step.

It has been shown that a message passing neural networks (MPNNs), a popular family of neural networks for graph-structured data, are at most as expressive as the firstorder Weisfeiler-Leman (1-WL) graph isomorphism test, which has motivated the development of more expressive architectures.In this work, we analyze if the limited expressiveness is actually a limiting factor for MPNNs and other WL-based models in standard graph datasets. Interestingly, we find that the expressiveness of WL is sufficient to identify almost all graphs in most datasets. Moreover, we find that the classification accuracy upper bounds are often close to 100%. Furthermore, we find that simple WL-based neural networks and several MPNNs can be fitted to several datasets. In sum, we conclude that the performance of WL/MPNNs is not limited by their expressiveness in practice.