Existing approaches to active learning maximize the system performance by sampling unlabeled instances for annotation that yield the most efficient training.However, when active learning is integrated with an end-user application, this can lead to frustration for participating users, as they spend time labeling instances that they would not otherwise be interested in reading.In this paper, we propose a new active learning approach that jointly optimizes the seemingly counteracting objectives of the active learning system (training efficiently) and the user (receiving useful instances).We study our approach in an educational application, which particularly benefits from this technique as the system needs to rapidly learn to predict the appropriateness of an exercise to a particular user, while the users should receive only exercises that match their skills.We evaluate multiple learning strategies and user types with data from real users and find that our joint approach better satisfies both objectives when alternative methods lead to many unsuitable exercises for end users.1

Large state-of-the-art corpora for training neural networks to create abstractive summaries are mostly limited to the news genre, as it is expensive to acquire human-written summaries for other types of text at a large scale. In this paper, we present a novel automatic corpus construction approach to tackle this issue as well as three new large open-licensed summarization corpora based on our approach that can be used for training abstractive summarization models. Our constructed corpora contain fictional narratives, descriptive texts, and summaries about movies, television, and book series from different domains. All sources use a creative commons (CC) license, hence we can provide the corpora for download. In addition, we also provide a ready-to-use framework that implements our automatic construction approach to create custom corpora with desired parameters like the length of the target summary and the number of source documents from which to create the summary. The main idea behind our automatic construction approach is to use existing large text collections (e.g., thematic wikis) and automatically classify whether the texts can be used as (query-focused) multi-document summaries and align them with potential source texts. As a final contribution, we show the usefulness of our automatic construction approach by running state-of-the-art summarizers on the corpora and through a manual evaluation with human annotators.

Abstract Interactive NLP is a promising paradigm to close the gap between automatic NLP systems and the human upper bound. Preference-based interactive learning has been successfully applied, but the existing methods require several thousand interaction rounds even in simulations with perfect user feedback. In this paper, we study preference-based interactive summarisation. To reduce the number of interaction rounds, we propose the Active Preference-based ReInforcement Learning (APRIL) framework. APRIL uses active learning to query the user, preference learning to learn a summary ranking function from the preferences, and neural Reinforcement learning to efficiently search for the (near-)optimal summary. Our results show that users can easily provide reliable preferences over summaries and that APRIL outperforms the state-of-the-art preference-based interactive method in both simulation and real-user experiments.

Diagnostic reasoning is a key component of many professions. To improve students’ diagnostic reasoning skills, educational psychologists analyse and give feedback on epistemic activities used by these students while diagnosing, in particular, hypothesis generation, evidence generation, evidence evaluation, and drawing conclusions. However, this manual analysis is highly time-consuming. We aim to enable the large-scale adoption of diagnostic reasoning analysis and feedback by automating the epistemic activity identification. We create the first corpus for this task, comprising diagnostic reasoning selfexplanations of students from two domains annotated with epistemic activities. Based on insights from the corpus creation and the task’s characteristics, we discuss three challenges for the automatic identification of epistemic activities using AI methods: the correct identification of epistemic activity spans, the reliable distinction of similar epistemic activities, and the detection of overlapping epistemic activities. We propose a separate performance metric for each challenge and thus provide an evaluation framework for future research. Indeed, our evaluation of various state-of-the-art recurrent neural network architectures reveals that current techniques fail to address some of these challenges.

We propose two novel manipulation strategies for increasing and decreasing the difficulty of C-tests automatically. This is a crucial step towards generating learner-adaptive exercises for self-directed language learning and preparing language assessment tests. To reach the desired difficulty level, we manipulate the size and the distribution of gaps based on absolute and relative gap difficulty predictions. We evaluate our approach in corpus-based experiments and in a user study with 60 participants. We find that both strategies are able to generate C-tests with the desired difficulty level.

Claudia Schulz, Christian M. Meyer, Jan Kiesewetter, Michael Sailer, Elisabeth Bauer, Martin R. Fischer, Frank Fischer, Iryna Gurevych. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019.

Florian Böhm, Yang Gao, Christian M. Meyer, Ori Shapira, Ido Dagan, Iryna Gurevych. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Wei Zhao, Maxime Peyrard, Fei Liu, Yang Gao, Christian M. Meyer, Steffen Eger. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Document summarisation can be formulated as a sequential decision-making problem, which can be solved by Reinforcement Learning (RL) algorithms. The predominant RL paradigm for summarisation learns a cross-input policy, which requires considerable time, data and parameter tuning due to the huge search spaces and the delayed rewards. Learning input-specific RL policies is a more efficient alternative, but so far depends on handcrafted rewards, which are difficult to design and yield poor performance. We propose RELIS, a novel RL paradigm that learns a reward function with Learning-to-Rank (L2R) algorithms at training time and uses this reward function to train an input-specific RL policy at test time. We prove that RELIS guarantees to generate near-optimal summaries with appropriate L2R and RL algorithms. Empirically, we evaluate our approach on extractive multi-document summarisation. We show that RELIS reduces the training time by two orders of magnitude compared to the state-of-the-art models while performing on par with them.

We learn user preferences from ratings and reviews by using multi-task learning (MTL) of rating prediction and summarization of item reviews. Reviews of an item tend to describe detailed user preferences (e.g., the cast, genre, or screenplay of a movie). A summary of such a review or a rating describes an overall user experience of the item. Our objective is to learn latent vectors which are shared across rating prediction and review summary generation. Additionally, the learned latent vectors and the generated summary act as explanations for the recommendation. Our MTL-based approach J3R uses a multi-layer perceptron for rating prediction, combined with pointer-generator networks with attention mechanism for the summarization component. We provide empirical evidence for joint learning of rating prediction and summary generation being beneficial for recommendation by conducting experiments on the Yelp dataset and six domains of the Amazon 5-core dataset. Additionally, we provide two ways of explanations visualizing (a) the user vectors on different topics of a domain, computed from our J3R approach and (b) a ten-word review summary of a review and the attention highlights generated on the review based on the user-item vectors

The 2017 Fake News Challenge Stage 1 (FNC-1) shared task addressed a stance classification task as a crucial first step towards detecting fake news. To date, there is no in-depth analysis paper to critically discuss FNC-1’s experimental setup, reproduce the results, and draw conclusions for next-generation stance classification methods. In this paper, we provide such an in-depth analysis for the three top-performing systems. We first find that FNC-1’s proposed evaluation metric favors the majority class, which can be easily classified, and thus overestimates the true discriminative power of the methods. Therefore, we propose a new F1-based metric yielding a changed system ranking. Next, we compare the features and architectures used, which leads to a novel feature-rich stacked LSTM model that performs on par with the best systems, but is superior in predicting minority classes. To understand the methods’ ability to generalize, we derive a new dataset and perform both in-domain and cross-domain experiments. Our qualitative and quantitative study helps interpreting the original FNC-1 scores and understand which features help improving performance and why. Our new dataset and all source code used during the reproduction study are publicly available for future research.

We propose a method to perform automatic document summarisation without using reference summaries.Instead, our method interactively learns from users' preferences.The merit of preference-based interactive summarisation is that preferences are easier for users to provide than reference summaries.Existing preference-based interactive learning methods suffer from high sample complexity, i.e. they need to interact with the oracle for many rounds in order to converge.In this work, we propose a new objective function, which enables us to leverage active learning, preference learning and reinforcement learning techniques in order to reduce the sample complexity.Both simulation and real-user experiments suggest that our method significantly advances the state of the art.Our source code is freely available at https://github.com/UKPLab/emnlp2018-april.

There exists an ever-growing set of data-centric systems that allow data scientists of varying skill levels to interactively manipulate, analyze and explore large structured data sets. However, there are currently not many systems that allow data scientists and novice users to interactively explore large unstructured text document collections from heterogeneous sources. In this demo paper, we present a new system for interactive text summarization called Sherlock. The task of automatically producing textual summaries is an important step to understand a collection of multiple topic-related documents. It has many real-world applications in journalism, medicine, and many more. However, none of the existing summarization systems allow users to provide feedback at interactive speed. We therefore integrate a new approximate summarization model into Sherlock that can guarantee interactive speeds even for large text collections to keep the user engaged in the process.

In this paper, we propose an extractive multi-document summarization (MDS) system using joint optimization and active learning for content selection grounded in user feedback.Our method interactively obtains user feedback to gradually improve the results of a state-of-the-art integer linear programming (ILP) framework for MDS.Our methods complement fully automatic methods in producing highquality summaries with a minimum number of iterations and feedbacks.We conduct multiple simulation-based experiments and analyze the effect of feedbackbased concept selection in the ILP setup in order to maximize the user-desired content in the summary.

Concept-map-based multi-document summarization is a variant of traditional summarization that produces structured summaries in the form of concept maps. In this work, we propose a new model for the task that addresses several issues in previous methods. It learns to identify and merge coreferent concepts to reduce redundancy, determines their importance with a strong supervised model and finds an optimal summary concept map via integer linear programming. It is also computationally more efficient than previous methods, allowing us to summarize larger document sets. We evaluate the model on two datasets, finding that it outperforms several approaches from previous work.

Coherent extracts are a novel type of summary combining the advantages of manually created abstractive summaries, which are fluent but difficult to evaluate, and low-quality automatically created extractive summaries, which lack coherence and structure. We use a corpus of heterogeneous documents to address the issue that information seekers usually face – a variety of different types of information sources. We directly extract information from these, but minimally redact and meaningfully order it to form a coherent text. Our qualitative and quantitative evaluations show that quantitative results are not sufficient to judge the quality of a summary and that other quality criteria, such as coherence, should also be taken into account. We find that our manually created corpus is of high quality and that it has the potential to bridge the gap between reference corpora of abstracts and automatic methods producing extracts. Our corpus is available to the research community for further development.

We introduce the task of detecting cross-lingual marketing blunders, which occur if a trade name resembles an inappropriate or negatively connotated word in a target language. To this end, we suggest a formal task definition and a semi-automatic method based the propagation of pragmatic labels from Wiktionary across sense-disambiguated translations. Our final tool assists users by providing clues for problematic names in any language, which we simulate in two experiments on detecting previously occurred marketing blunders and identifying relevant clues for established international brands. We conclude the paper with a suggested research roadmap for this new task. To initiate further research, we publish our online demo along with the source code and data at http://uby.ukp.informatik.tu-darmstadt.de/blunder/.