The ACL Anthology is a prime resource for research papers within computational linguistics and natural language processing, while continuing to be an open-source and community-driven project. Since Gildea et al. (2018) reported on its state and planned directions, the Anthology has seen major technical changes. We discuss what led to these changes and how they impact long-term maintainability and community engagement, describe which open-source data and software tools the Anthology currently provides, and provide a survey of literature that has used the Anthology as a main data source.

Recipe texts are an idiosyncratic form of instructional language that pose unique challenges for automatic understanding. One challenge is that a cooking step in one recipe can be explained in another recipe in different words, at a different level of abstraction, or not at all. Previous work has annotated correspondences between recipe instructions at the sentence level, often glossing over important correspondences between cooking steps across recipes. We present a novel and fully-parsed English recipe corpus, ARA (Aligned Recipe Actions), which annotates correspondences between individual actions across similar recipes with the goal of capturing information implicit for accurate recipe understanding. We represent this information in the form of recipe graphs, and we train a neural model for predicting correspondences on ARA. We find that substantial gains in accuracy can be obtained by taking fine-grained structural information about the recipes into account.

We present a comprehensive platform to run human-computer experiments where an agent instructs a human in Minecraft, a 3D blocksworld environment. This platform enables comparisons between different agents by matching users to agents. It performs extensive logging and takes care of all boilerplate, allowing to easily incorporate new agents to evaluate them. Our environment is prepared to evaluate any kind of instruction giving system, recording the interaction and all actions of the user. We provide example architects, a Wizard-of-Oz architect and set-up scripts to automatically download, build and start the platform.

When generating technical instructions, it is often convenient to describe complex objects in the world at different levels of abstraction. A novice user might need an object explained piece by piece, while for an expert, talking about the complex object (e.g. a wall or railing) directly may be more succinct and efficient. We show how to generate building instructions at different levels of abstraction in Minecraft. We introduce the use of hierarchical planning to this end, a method from AI planning which can capture the structure of complex objects neatly. A crowdsourcing evaluation shows that the choice of abstraction level matters to users, and that an abstraction strategy which balances low-level and high-level object descriptions compares favorably to ones which don’t.

When generating technical instructions, it is often necessary to describe an object that does not exist yet. For example, an NLG system which explains how to build a house needs to generate sentences like “build *a wall of height five to your left*” and “now build *a wall on the other side*.” Generating (indefinite) referring expressions to objects that do not exist yet is fundamentally different from generating the usual definite referring expressions, because the new object must be distinguished from an infinite set of possible alternatives. We formalize this problem and present an algorithm for generating such expressions, in the context of generating building instructions within the Minecraft video game.

We introduce the use of Poincaré embeddings to improve existing state-of-the-art approaches to domain-specific taxonomy induction from text as a signal for both relocating wrong hyponym terms within a (pre-induced) taxonomy as well as for attaching disconnected terms in a taxonomy. This method substantially improves previous state-of-the-art results on the SemEval-2016 Task 13 on taxonomy extraction. We demonstrate the superiority of Poincaré embeddings over distributional semantic representations, supporting the hypothesis that they can better capture hierarchical lexical-semantic relationships than embeddings in the Euclidean space.

De-identification is the task of detecting protected health information (PHI) in medical text. It is a critical step in sanitizing electronic health records (EHR) to be shared for research. Automatic de-identification classifiers can significantly speed up the sanitization process. However, obtaining a large and diverse dataset to train such a classifier that works well across many types of medical text poses a challenge as privacy laws prohibit the sharing of raw medical records. We introduce a method to create privacy-preserving shareable representations of medical text (i.e. they contain no PHI) that does not require expensive manual pseudonymization. These representations can be shared between organizations to create unified datasets for training de-identification models. Our representation allows training a simple LSTM-CRF de-identification model to an F1 score of 97.4%, which is comparable to a strong baseline that exposes private information in its representation. A robust, widely available de-identification classifier based on our representation could potentially enable studies for which de-identification would otherwise be too costly.

Corpora with language learner writing usually consist of essays, which are difficult to annotate reliably and to process automatically due to the high degree of freedom and the nature of learner language. We develop a task which mildly constrains learner utterances to facilitate consistent annotation and reliable automatic processing but at the same time does not prime learners with textual information. In this task, learners retell a comic strip. We present the resulting task-based corpus of stories written by learners of German. We designed the corpus to be able to serve multiple purposes: The corpus was manually annotated, including target hypotheses and syntactic structures. We achieve a very high inter-annotator agreement: κ = 0.765 for the annotation of minimal target hypotheses and κ = 0.507 for the extended target hypotheses. We attribute this to the design of our task and the annotation guidelines, which are based on those for the Falko corpus (Reznicek et al., 2012).

Incrementality is ubiquitous in human-human interaction and beneficial for human-computer interaction. It has been a topic of research in different parts of the NLP community, mostly with focus on the specific topic at hand even though incremental systems have to deal with similar challenges regardless of domain. In this survey, I consolidate and categorize the approaches, identifying similarities and differences in the computation and data, and show trade-offs that have to be considered. A focus lies on evaluating incremental systems because the standard metrics often fail to capture the incremental properties of a system and coming up with a suitable evaluation scheme is non-trivial.

Predictive incremental parsing produces syntactic representations of sentences as they are produced, e.g. by typing or speaking. In order to generate connected parses for such unfinished sentences, upcoming word types can be hypothesized and structurally integrated with already realized words. For example, the presence of a determiner as the last word of a sentence prefix may indicate that a noun will appear somewhere in the completion of that sentence, and the determiner can be attached to the predicted noun. We combine the forward-looking parser predictions with backward-looking N-gram histories and analyze in a set of experiments the impact on language models, i.e. stronger discriminative power but also higher data sparsity. Conditioning N-gram models, MaxEnt models or RNN-LMs on parser predictions yields perplexity reductions of about 6%. Our method (a) retains online decoding capabilities and (b) incurs relatively little computational overhead which sets it apart from previous approaches that use syntax for language modeling. Our method is particularly attractive for modular systems that make use of a syntax parser anyway, e.g. as part of an understanding pipeline where predictive parsing improves language modeling at no additional cost.

We present a corpus of time-aligned spoken data of Wikipedia articles as well as the pipeline that allows to generate such corpora for many languages. There are initiatives to create and sustain spoken Wikipedia versions in many languages and hence the data is freely available, grows over time, and can be used for automatic corpus creation. Our pipeline automatically downloads and aligns this data. The resulting German corpus currently totals 293h of audio, of which we align 71h in full sentences and another 86h of sentences with some missing words. The English corpus consists of 287h, for which we align 27h in full sentence and 157h with some missing words. Results are publically available.

We present the Hamburg Dependency Treebank (HDT), which to our knowledge is the largest dependency treebank currently available. It consists of genuine dependency annotations, i. e. they have not been transformed from phrase structures. We explore characteristics of the treebank and compare it against others. To exemplify the benefit of large dependency treebanks, we evaluate different parsers on the HDT. In addition, a set of tools will be described which help working with and searching in the treebank.