Daniel Hershcovich

Tenure-Track Assistant Professor at CoAStaL, Natural Language Processing section, Department of Computer Science, University of Copenhagen, Denmark.

I was previously a postdoc at CoAStaL, after completing my PhD at ELSC and CS at HUJI with Ari Rappoport and Omri Abend. I developed a parser for UCCA. Before that, I did my BSc in Mathematics and Computer Science at the Open University of Israel. During 2008-2019, I was also at Project Debater, IBM Research.

An announcement is out for a 3-year NLP postdoc on Scandinavian literature (deadline: 1 July 2022) with the Department of Nordic Studies and Linguistics. Contact me for more information!
Our paper Generalized Quantifiers as a Source of Error in Multilingual NLU Benchmarks was accepted to NAACL 2022!
Our paper Compositional Generalization in Multilingual Semantic Parsing over Wikidata was accepted to TACL!

Cross-cultural and cross-lingual adaptation of (and investigation with) natural language processing models.
Interpretable representation of natural language syntax, semantics and pragmatics.
Promoting welfare, sustainability and responsible behavior through language technology.

(on Google Scholar, on Semantic Scholar)

Compositional Generalization in Multilingual Semantic Parsing over Wikidata.
Ruixiang Cui, Rahul Aralikatte, Heather Lent and Daniel Hershcovich. TACL, 2022.

preprint abstract bib code and data

Semantic parsing (SP) allows humans to leverage vast knowledge resources through natural interaction. However, parsers are mostly designed for and evaluated on English resources, such as CFQ (Keysers et al., 2020), the current standard benchmark based on English data generated from grammar rules and oriented towards Freebase, an outdated knowledge base. We propose a method for creating a multilingual, parallel dataset of question-query pairs, grounded in Wikidata. We introduce such a dataset, which we call Multilingual Compositional Wikidata Questions (MCWQ), and use it to analyze the compositional generalization of semantic parsers in Hebrew, Kannada, Chinese and English. While within-language generalization is comparable across languages, experiments on zero-shot cross-lingual transfer demonstrate that cross-lingual compositional generalization fails, even with state-of-the-art pretrained multilingual encoders. Furthermore, our methodology, dataset and results will facilitate future research on SP in more realistic and diverse settings than has been possible with existing resources.
Towards Climate Awareness in NLP Research.
Daniel Hershcovich, Nicolas Webersinke, Mathias Kraus, Julia Anna Bingler and Markus Leippold.

preprint abstract bib

The climate impact of AI, and NLP research in particular, has become a serious issue given the enormous amount of energy that is increasingly being used for training and running computational models. Consequently, increasing focus is placed on efficient NLP. However, this important initiative lacks simple guidelines that would allow for systematic climate reporting of NLP research. We argue that this deficiency is one of the reasons why very few publications in NLP report key figures that would allow a more thorough examination of environmental impact. As a remedy, we propose a climate performance model card with the primary purpose of being practically usable with only limited information about experiments and the underlying computer hardware. We describe why this step is essential to increase awareness about the environmental impact of NLP research and, thereby, paving the way for more thorough discussions.
Generalized Quantifiers as a Source of Error in Multilingual NLU Benchmarks.
Ruixiang Cui, Daniel Hershcovich and Anders Søgaard. NAACL 2022.

preprint abstract bib

Logical approaches to representing language have developed and evaluated computational models of quantifier words since the 19th century, but today's NLU models still struggle to capture their semantics. We rely on Generalized Quantifier Theory for language-independent representations of the semantics of quantifier words, to quantify their contribution to the errors of NLU models. We find that quantifiers are pervasive in NLU benchmarks, and their occurrence at test time is associated with performance drops. Multilingual models also exhibit unsatisfying quantifier reasoning abilities, but not necessarily worse for non-English languages. To facilitate directly-targeted probing, we present an adversarial generalized quantifier NLI task (GQNLI) and show that pre-trained language models have a clear lack of robustness in generalized quantifier reasoning.
Evaluating Deep Taylor Decomposition for Reliability Assessment in the Wild.
Stephanie Brandl, Daniel Hershcovich and Anders Søgaard. ICWSM 2022.

abstract bib

We argue that we need to evaluate model interpretability methods in the wild, in situations where professionals make critical decisions, and models can potentially assist them. We present an in-the-wild evaluation of token attribution based on Deep Taylor Decomposition, with professional journalists performing reliability assessments. We find that using this method in conjunction with RoBERTa-Large, fine-tuned on the Gossip Corpus, led to faster and better human decision-making, as well as a more critical attitude toward news sources among the journalists. We present a comparison of human and model rationales, as well as a qualitative analysis of the journalists' experiences with machine-in-the-loop decision making.
Challenges and Strategies in Cross-Cultural NLP.
Daniel Hershcovich, Stella Frank, Heather Lent, Miryam de Lhoneux, Mostafa Abdou, Stephanie Brandl, Emanuele Bugliarello, Laura Cabello Piqueras, Ilias Chalkidis, Ruixiang Cui, Constanza Fierro, Katerina Margatina, Phillip Rust and Anders Søgaard. ACL 2022.

proceedings pdf preprint abstract bib slides poster

Various efforts in the Natural Language Processing (NLP) community have been made to accommodate linguistic diversity and serve speakers of many different languages. However, it is important to acknowledge that speakers and the content they produce and require, vary not just by language, but also by culture. Although language and culture are tightly linked, there are important differences. Analogous to cross-lingual and multilingual NLP, cross-cultural and multicultural NLP considers these differences in order to better serve users of NLP systems. We propose a principled framework to frame these efforts, and survey existing and potential strategies.
Scaling Creative Inspiration with Fine-Grained Functional Facets of Product Ideas.
Tom Hope, Ronen Tamari, Hyeonsu Kang, Daniel Hershcovich, Joel Chan, Aniket Kittur and Dafna Shahaf. CHI 2022.

proceedings preprint abstract bib

Web-scale repositories of products, patents and scientific papers offer an opportunity for creating automated systems that scour millions of ideas and assist users in discovering inspirations and solutions. Yet the common representation of ideas is in the form of raw textual descriptions, lacking important structure that is required for supporting creative innovation. Prior work has pointed to the importance of functional structure -- capturing the mechanisms and purposes of inventions -- for allowing users to discover structural connections across ideas and creatively adapt existing technologies. However, the use of functional representations was either coarse and limited in expressivity, or dependent on curated knowledge bases with poor coverage and significant manual effort from users. To help bridge this gap and unlock the potential of large-scale idea mining, we propose a novel computational representation that automatically breaks up products into fine-grained functional facets. We train a model to extract these facets from a challenging real-world corpus of invention descriptions, and represent each product as a set of facet embeddings. We design similarity metrics that support granular matching between functional facets across ideas, and use them to build a novel functional search capability that enables expressive queries for mechanisms and purposes. We construct a graph capturing hierarchical relations between purposes and mechanisms across an entire corpus of products, and use the graph to help problem-solvers explore the design space around a focal problem and view related problem perspectives. In empirical user studies, our approach leads to a significant boost in search accuracy and in the quality of creative inspirations, outperforming strong baselines and state-of-art representations of product texts by 50-60%.
A Multilingual Benchmark for Probing Negation-Awareness with Minimal Pairs.
Mareike Hartmann, Miryam de Lhoneux, Daniel Hershcovich, Yova Kementchedjhieva, Lukas Nielsen, Chen Qiu and Anders Søgaard.
CoNLL 2021.

proceedings abstract bib

Negation is one of the most fundamental concepts in human cognition and language, and several natural language inference (NLI) probes have been designed to investigate pretrained language models' ability to detect and reason with negation. However, the existing probing datasets are limited to English only, and do not enable controlled probing of performance in the absence or presence of negation. In response, we present a multilingual (English, Bulgarian, German, French and Chinese) benchmark collection of NLI examples that are grammatical and correctly labeled, as a result of manual inspection and editing. We use the benchmark to probe the negation-awareness of multilingual language models and find that models that correctly predict examples with negation cues often fail to correctly predict their counter-examples {\em without} negation cues, even when the cues are irrelevant for semantic inference.
Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color.
Mostafa Abdou, Artur Kulmizev, Daniel Hershcovich, Stella Frank, Ellie Pavlick and Anders Søgaard.
CoNLL 2021.

proceedings preprint abstract bib

Pretrained language models have been shown to encode relational information, such as the relations between entities or concepts in knowledge-bases -- (Paris, Capital, France). However, simple relations of this type can often be recovered heuristically and the extent to which models implicitly reflect topological structure that is grounded in world, such as perceptual structure, is unknown. To explore this question, we conduct a thorough case study on color. Namely, we employ a dataset of monolexemic color terms and color chips represented in CIELAB, a color space with a perceptually meaningful distance metric. Using two methods of evaluating the structural alignment of colors in this space with text-derived color term representations, we find significant correspondence. Analyzing the differences in alignment across the color spectrum, we find that warmer colors are, on average, better aligned to the perceptual color space than cooler ones, suggesting an intriguing connection to findings from recent work on efficient communication in color naming. Further analysis suggests that differences in alignment are, in part, mediated by collocationality and differences in syntactic usage, posing questions as to the relationship between color perception and usage and context.
Great Service! Fine-grained Parsing of Implicit Arguments.
Ruixiang Cui and Daniel Hershcovich.
IWPT 2021.

proceedings preprint abstract bib code

Broad-coverage meaning representations in NLP mostly focus on explicitly expressed content. More importantly, the scarcity of datasets annotating diverse implicit roles limits empirical studies into their linguistic nuances. For example, in the web review "Great service!", the provider and consumer are implicit arguments of different types. We examine an annotated corpus of fine-grained implicit arguments (Cui and Hershcovich, 2020) by carefully re-annotating it, resolving several inconsistencies. Subsequently, we present the first transition-based neural parser that can handle implicit arguments dynamically, and experiment with two different transition systems on the improved dataset. We find that certain types of implicit arguments are more difficult to parse than others and that the simpler system is more accurate in recovering implicit arguments, despite having a lower overall parsing score, attesting current reasoning limitations of NLP models. This work will facilitate a better understanding of implicit and underspecified language, by incorporating it holistically into meaning representations.
Lexical Semantic Recognition.
Nelson F. Liu, Daniel Hershcovich, Michael Kranzlein, Nathan Schneider.
MWE 2021.

proceedings preprint abstract bib code

In lexical semantics, full-sentence segmentation and segment labeling of various phenomena are generally treated separately, despite their interdependence. We hypothesize that a unified lexical semantic recognition task is an effective way to encapsulate previously disparate styles of annotation, including multiword expression identification / classification and supersense tagging. Using the STREUSLE corpus, we train a neural CRF sequence tagger and evaluate its performance along various axes of annotation. As the label set generalizes that of previous tasks (PARSEME, DiMSUM), we additionally evaluate how well the model generalizes to those test sets, finding that it approaches or surpasses existing models despite training only on STREUSLE. Our work also establishes baseline models and evaluation metrics for integrated and accurate modeling of lexical semantics, facilitating future work in this area.
It’s the Meaning That Counts: The State of the Art in NLP and Semantics.
Daniel Hershcovich and Lucia Donatelli.
KI - Künstliche Intelligenz, 2021.

pdf online abstract bib

Semantics, the study of meaning, is central to research in Natural Language Processing (NLP) and many other fields connected to Artificial Intelligence. Nevertheless, how semantics is understood in NLP ranges from traditional, formal linguistic definitions based on logic and the principle of compositionality to more applied notions based on grounding meaning in real-world objects and real-time interaction. “Semantic” methods may additionally strive for meaningful representation of language that integrates broader aspects of human cognition and embodied experience, calling into question how adequate a representation of meaning based on linguistic signal alone is for current research agendas. We review the state of computational semantics in NLP and investigate how different lines of inquiry reflect distinct understandings of semantics and prioritize different layers of linguistic meaning. In conclusion, we identify several important goals of the field and describe how current research addresses them.
How far can we get with one GPU in 100 hours? CoAStaL at MultiIndicMT Shared Task.
Rahul Aralikatte, Héctor Ricardo Murrieta Bello, Daniel Hershcovich, Marcel Bollmann, Anders Søgaard.
MultiIndicMT Shared Task 2021.

proceedings abstract bib

This work shows that competitive translation results can be obtained in a constrained setting by incorporating the latest advances in memory and compute optimization. We train and evaluate large multilingual translation models using a single GPU for a maximum of 100 hours and get within 4-5 BLEU points of the top submission on the leaderboard. We also benchmark standard baselines on the PMI corpus and re-discover well-known shortcomings of translation systems and metrics.
Moses and the Character-Based Random Babbling Baseline: CoAStaL at AmericasNLP 2021 Shared Task.
Marcel Bollmann, Rahul Aralikatte, Héctor Murrieta Bello, Daniel Hershcovich, Miryam de Lhoneux, Anders Søgaard.
AmericasNLP Shared Task 2021.

proceedings abstract bib

We evaluated a range of neural machine translation techniques developed specifically for low-resource scenarios. Unsuccessfully. In the end, we submitted two runs: (i) a standard phrase-based model, and (ii) a random babbling baseline using character trigrams. We found that it was surprisingly hard to beat (i), in spite of this model being, in theory, a bad fit for polysynthetic languages; and more interestingly, that (ii) was better than several of the submitted systems, highlighting how difficult low-resource machine translation for polysynthetic languages is.
An autonomous debating system.
Noam Slonim, Yonatan Bilu, Carlos Alzate, Roy Bar-Haim, Ben Bogin, Francesca Bonin, Leshem Choshen, Edo Cohen-Karlik, Lena Dankin, Lilach Edelstein, Liat Ein-Dor, Roni Friedman-Melamed, Assaf Gavron, Ariel Gera, Martin Gleize, Shai Gretz, Dan Gutfreund, Alon Halfon, Daniel Hershcovich, Ron Hoory, Yufang Hou, Shay Hummel, Michal Jacovi, Charles Jochim, Yoav Kantor, Yoav Katz, David Konopnicki, Zvi Kons, Lili Kotlerman, Dalia Krieger, Dan Lahav, Tamar Lavee, Ran Levy, Naftali Liberman, Yosi Mass, Amir Menczel, Shachar Mirkin, Guy Moshkowich, Shila Ofek-Koifman, Matan Orbach, Ella Rabinovich, Ruty Rinott, Slava Shechtman, Dafna Sheinwald, Eyal Shnarch, Ilya Shnayderman, Aya Soffer, Artem Spector, Benjamin Sznajder, Assaf Toledo, Orith Toledo-Ronen, Elad Venezian and Ranit Aharonov.
Nature, 2021.

pdf online abstract bib website

Artificial intelligence (AI) is defined as the ability of machines to perform tasks that are usually associated with intelligent beings. Argument and debate are fundamental capabilities of human intelligence, essential for a wide range of human activities, and common to all human societies. The development of computational argumentation technologies is therefore an important emerging discipline in AI research1. Here we present Project Debater, an autonomous debating system that can engage in a competitive debate with humans. We provide a complete description of the system’s architecture, a thorough and systematic evaluation of its operation across a wide range of debate topics, and a detailed account of the system’s performance in its public debut against three expert human debaters. We also highlight the fundamental differences between debating with humans as opposed to challenging humans in game competitions, the latter being the focus of classical ‘grand challenges’ pursued by the AI research community over the past few decades. We suggest that such challenges lie in the ‘comfort zone’ of AI, whereas debating with humans lies in a different territory, in which humans still prevail, and for which novel paradigms are required to make substantial progress.
Does injecting linguistic structure into language models lead to better alignment with brain recordings?.
Mostafa Abdou, Ana Valeria Gonzalez, Mariya Toneva, Daniel Hershcovich and Anders Søgaard.

preprint abstract bib code

Neuroscientists evaluate deep neural networks for natural language processing as possible candidate models for how language is processed in the brain. These models are often trained without explicit linguistic supervision, but have been shown to learn some linguistic structure in the absence of such supervision (Manning et al., 2020), potentially questioning the relevance of symbolic linguistic theories in modeling such cognitive processes (Warstadt and Bowman, 2020). We evaluate across two fMRI datasets whether language models align better with brain recordings, if their attention is biased by annotations from syntactic or semantic formalisms. Using structure from dependency or minimal recursion semantic annotations, we find alignments improve significantly for one of the datasets. For another dataset, we see more mixed results. We present an extensive analysis of these results. Our proposed approach enables the evaluation of more targeted hypotheses about the composition of meaning in the brain, expanding the range of possible scientific inferences a neuroscientist could make, and opens up new opportunities for cross-pollination between computational neuroscience and linguistics.
Joint Semantic Analysis with Document-Level Cross-Task Coherence Rewards.
Rahul Aralikatte, Mostafa Abdou, Heather Lent, Daniel Hershcovich and Anders Søgaard.
AAAI 2021 (long paper).

preprint abstract bib code

Coreference resolution and semantic role labeling are NLP tasks that capture different aspects of semantics, indicating respectively, which expressions refer to the same entity, and what semantic roles expressions serve in the sentence. However, they are often closely interdependent, and both generally necessitate natural language understanding. Do they form a coherent abstract representation of documents? We present a neural network architecture for joint coreference resolution and semantic role labeling for English, and train graph neural networks to model the 'coherence' of the combined shallow semantic graph. Using the resulting coherence score as a reward for our joint semantic analyzer, we use reinforcement learning to encourage global coherence over the document and between semantic annotations. This leads to improvements on both tasks in multiple datasets from different domains, and across a range of encoders of different expressivity, calling, we believe, for a more holistic approach for semantics in NLP.
Cross-lingual Semantic Representation for NLP with UCCA.
Omri Abend, Dotan Dvir, Daniel Hershcovich, Jakob Prange and Nathan Schneider.
COLING 2020 tutorial.

proceedings abstract bib slides videos

This is an introductory tutorial to UCCA (Universal Conceptual Cognitive Annotation), a cross-linguistically applicable framework for semantic representation, with corpora annotated in English, German and French, and ongoing annotation in Russian and Hebrew. UCCA builds on extensive typological work and supports rapid annotation. The tutorial will provide a detailed introduction to the UCCA annotation guidelines, design philosophy and the available resources; and a comparison to other meaning representations. It will also survey the existing parsing work, including the findings of three recent shared tasks, in SemEval and CoNLL, that addressed UCCA parsing. Finally, the tutorial will present recent applications and extensions to the scheme, demonstrating its value for natural language processing in a range of languages and domains.
Refining Implicit Argument Annotation For UCCA.
Ruixiang Cui and Daniel Hershcovich.
DMR 2020.

pdf proceedings preprint abstract bib slides data

Predicate-argument structure analysis is a central component in meaning representations of text. The fact that some arguments are not explicitly mentioned in a sentence gives rise to ambiguity in language understanding, and renders it difficult for machines to interpret text correctly. However, only few resources represent implicit roles for NLU, and existing studies in NLP only make coarse distinctions between categories of arguments omitted from linguistic form. To better understand the behaviour of implicit roles and their characteristics, in this paper, we design a typology for fine-grained implicit argument annotation on top of Universal Conceptual Cognitive Annotation's foundational layer. The proposed implicit argument categorisation is driven by theories of implicit role interpretation and consists of six types: Deictic, Generic, Genre-based, Type-identifiable, Non-specific, and Iterated-set. We exemplify our design by revisiting part of the UCCA EWT corpus, providing a new dataset annotated with the refinement layer, and making a comparative analysis with other schemes both in terms of quantity and quality.
Comparison by Conversion: Reverse-Engineering UCCA from Syntax and Lexical Semantics.
Daniel Hershcovich, Nathan Schneider, Dotan Dvir, Jakob Prange, Miryam de Lhoneux and Omri Abend.
COLING 2020.

pdf proceedings preprint abstract bib poster converter code parser code

Building robust natural language understanding systems will require a clear characterization of whether and how various linguistic meaning representations complement each other. To perform a systematic comparative analysis, we evaluate the mapping between meaning representations from different frameworks using two complementary methods: (i) a rule-based converter, and (ii) a supervised delexicalized parser that parses to one framework using only information from the other as features. We apply these methods to convert the STREUSLE corpus (with syntactic and lexical semantic annotations) to UCCA (a graph-structured full-sentence meaning representation). Both methods yield surprisingly accurate target representations, close to fully supervised UCCA parser quality---indicating that UCCA annotations are partially redundant with STREUSLE annotations. Despite this substantial convergence between frameworks, we find several important areas of divergence.
MRP 2020: The Second Shared Task on Cross-Framework and Cross-Lingual Meaning Representation Parsing.
Stephan Oepen, Omri Abend, Lasha Abzianidze, Johan Bos, Jan Hajic, Daniel Hershcovich, Bin Li, Tim O'Gorman, Nianwen Xue and Daniel Zeman.
CoNLL 2020 shared task.

pdf proceedings abstract bib slides website

The 2020 Shared Task at the Conference for Computational Language Learning (CoNLL) was devoted to Meaning Representation Parsing (MRP) across frameworks and languages. Extending a similar setup from the previous year, five distinct approaches to the representation of sentence meaning in the form of directed graphs were represented in the English training and evaluation data for the task, packaged in a uniform graph abstraction and serialization; for four of these representation frameworks, additional training and evaluation data was provided for one additional language per framework. The task received submissions from eight teams, of which two do not participate in the official ranking because they arrived after the closing deadline or made use of additional training data. All technical information regarding the task, including system submissions, official results, and links to supporting resources and software are available from the task web site at: http://mrp.nlpl.eu
HUJI-KU at MRP 2020: Two Transition-based Neural Parsers.
Ofir Arviv, Ruixiang Cui and Daniel Hershcovich.
CoNLL 2020 shared task.

pdf preprint proceedings abstract bib slides TUPA code HIT-SCIR code multitask code

This paper describes the HUJI-KU system submission to the shared task on Cross-Framework Meaning Representation Parsing (MRP) at the 2020 Conference for Computational Language Learning (CoNLL), employing TUPA and the HIT-SCIR parser, which were, respectively, the baseline system and winning system in the 2019 MRP shared task. Both are transition-based parsers using BERT contextualized embeddings. We generalized TUPA to support the newly-added MRP frameworks and languages, and experimented with multitask learning with the HIT-SCIR parser. We reached 4th place in both the cross-framework and cross-lingual tracks.
Køpsala: Transition-Based Graph Parsing via Efficient Training and Effective Encoding.
Daniel Hershcovich, Miryam de Lhoneux, Artur Kulmizev, Elham Pejhan and Joakim Nivre.
IWPT 2020 shared task.

pdf preprint proceedings abstract bib slides code

We present Køpsala, the Copenhagen-Uppsala system for the Enhanced Universal Dependencies Shared Task at IWPT 2020. Our system is a pipeline consisting of off-the-shelf models for everything but enhanced graph parsing, and for the latter, a transition-based graph parser adapted from Che et al. (2019). We train a single enhanced parser model per language, using gold sentence splitting and tokenization for training, and rely only on tokenized surface forms and multilingual BERT for encoding. While a bug introduced just before submission resulted in a severe drop in precision, its post-submission fix would bring us to 4th place in the official ranking, according to average ELAS. Our parser demonstrates that a unified pipeline is effective for both Meaning Representation Parsing and Enhanced Universal Dependencies.
Universal Semantic Parsing with Neural Networks.
Daniel Hershcovich.
PhD thesis, Hebrew University of Jerusalem, 2019.

pdf abstract bib slides

A major scientific effort is dedicated to natural language understanding, which aims to be able to comprehend text, reason about it, and act upon it in an intelligent way. While specific use-cases or benchmarks can be solved with relatively simple systems, which either ignore word order ("bag-of-words" models) or treat it as a simple linear structure (such as the popular sequence-to-sequence framework allowing neural networks to learn tasks in an end-to-end fashion), understanding human language in general requires a hierarchical representation of meaning. Constructing this representation from text has been the goal of an extensive line of work in semantic parsing. While many semantic representation schemes have been proposed, they share many of their basic distinctions, such as between predicates (relations, states and events) and arguments (participants). This thesis focuses on a particular semantic representation scheme called Universal Conceptual Cognitive Annotation (UCCA), whose main design principles are support for all major linguistic semantic phenomena, cross-linguistic applicability, stability across translations, ease of annotation (even by those who are not experts in linguistics), and a modular architecture supporting multiple layers of semantic annotation. A fully automatic parser is presented, and evaluated on multiple languages (English, French and German). The parser, titled "TUPA" (transition-based UCCA parser), is able to learn very general graph structures: directed acyclic graphs over token sequences with non-terminal nodes for complex units, where these may cover discontinuous terminal yields. This general class of graphs covers the structures annotated in UCCA, as well as other representation schemes. TUPA is implemented as a transition-based parser, whose transition system supports these structural properties. Its transition classifier is a neural network equipped with a bidirectional long short-term memory (BiLSTM) module for calculating feature representations for the input. In an extensive comparison to conversion-based methods, as well as other classifier implementations, TUPA is shown to outperform all baselines in the task of UCCA parsing in both in-domain and out-of-domain settings in three languages. The parser is subsequently applied to two other semantic representation schemes, DM and AMR, and to syntactic dependencies in the Universal Dependencies (UD) scheme. This demonstrates that the flexible parser is usable not just for UCCA parsing. Furthermore, training TUPA in a multitask setting on all of these schemes improves its UCCA parsing accuracy, by effectively learning generalizations across the different representations: a shared model is thus able to apply semantic distinctions in one task, which have been learned for another. Finally, in an empirical comparison of the content of semantic and syntactic representations, we discover several aspects of divergence, i.e., differences in the content captured by these schemes. These have profound impact on the potential contribution of syntax to semantic parsing, and on the usefulness of each of the approaches for semantic tasks in natural language processing. I see semantic parsing as a means for computers to learn language. While different representations focus on different distinctions and do so with formally different structures, they share an overall goal, which is to support natural language processing applications, such as classifying text into categories, tagging it for linguistic properties, performing inference and reasoning, and generating new text according to some constraints (e.g., machine translation). The combined datasets annotated in every representation are an invaluable resource, which, used effectively, can greatly boost our achievements in language understanding and processing.
MRP 2019: Cross-Framework Meaning Representation Parsing.
Stephan Oepen, Omri Abend, Jan Hajič, Daniel Hershcovich, Marco Kuhlmann, Tim O'Gorman, Nianwen Xue, Jayeol Chun, Milan Straka and Zdeňka Urešová.
CoNLL 2019 shared task.

pdf proceedings abstract bib slides website

The 2019 Shared Task at the Conference for Computational Language Learning (CoNLL) was devoted to Meaning Representation Parsing (MRP) across frameworks. Five distinct approaches to the representation of sentence meaning in the form of directed graph were represented in the training and evaluation data for the task, packaged in a uniform abstract graph representation and serialization. The task received submissions from eighteen teams, of which five do not participate in the official ranking because they arrived after the closing deadline, made use of additional training data, or involved one of the task co-organizers. All technical information regarding the task, including system submissions, official results, and links to supporting resources and software are available from the task web site at: http://mrp.nlpl.eu
TUPA at MRP 2019: A Multi-Task Baseline System.
Daniel Hershcovich and Ofir Arviv.
CoNLL 2019 shared task.

pdf proceedings abstract bib poster slides code

This paper describes the TUPA system submission to the shared task on CrossFramework Meaning Representation Parsing (MRP) at the 2019 Conference for Computational Language Learning (CoNLL). TUPA provides a baseline point of comparison and is not considered in the official ranking of participating systems. While originally developed for UCCA only, TUPA has been generalized to support all MRP frameworks included in the task, and trained using multi-task learning to parse them all with a shared model. It is a transition-based parser with a BiLSTM encoder, augmented with BERT contextualized embeddings.
Rewarding Coreference Resolvers for Being Consistent with World Knowledge.
Rahul Aralikatte, Heather Lent, Ana Valeria Gonzalez, Daniel Hershcovich, Chen Qiu, Anders Sandholm, Michael Ringgaard and Anders Søgaard.
EMNLP-IJCNLP 2019 (short paper).

pdf preprint proceedings abstract bib code

Unresolved coreference is a bottleneck for relation extraction, and high-quality coreference resolvers may produce an output that makes it a lot easier to extract knowledge triples. We show how to improve coreference resolvers by forwarding their input to a relation extraction system and reward the resolvers for producing triples that are found in knowledge bases. Since relation extraction systems can rely on different forms of supervision and be biased in different ways, we obtain the best performance, improving over the state of the art, using multi-task reinforcement learning.
The Language of Legal and Illegal Activity on the Darknet.
Leshem Choshen*, Dan Eldad*, Daniel Hershcovich*, Elior Sulem* and Omri Abend.
ACL 2019 (long paper).

pdf preprint proceedings abstract bib slides code

The non-indexed parts of the Internet (the Darknet) have become a haven for both legal and illegal anonymous activity. Given the magnitude of these networks, scalably monitoring their activity necessarily relies on automated tools, and notably on NLP tools. However, little is known about what characteristics texts communicated through the Darknet have, and how well off-the-shelf NLP tools do on this domain. This paper tackles this gap and performs an in-depth investigation of the characteristics of legal and illegal text in the Darknet, comparing it to a clear net website with similar content as a control condition. Taking drug-related websites as a test case, we find that texts for selling legal and illegal drugs have several linguistic characteristics that distinguish them from one another, as well as from the control condition, among them the distribution of POS tags, and the coverage of their named entities in Wikipedia.
Argument Invention from First Principles.
Yonatan Bilu, Ariel Gera, Daniel Hershcovich, Benjamin Sznajder, Dan Lahav, Guy Moshkowich, Anael Malet, Assaf Gavron and Noam Slonim.
ACL 2019 (long paper).

pdf preprint proceedings abstract bib

Competitive debaters often find themselves facing a challenging task -- how to debate a topic they know very little about, with only minutes to prepare, and without access to books or the Internet? What they often do is rely on ''first principles'', commonplace arguments which are relevant to many topics, and which they have refined in past debates. In this work we aim to explicitly define a taxonomy of such principled recurring arguments, and, given a controversial topic, to automatically identify which of these arguments are relevant to the topic. As far as we know, this is the first time that this approach to argument invention is formalized and made explicit in the context of NLP. The main goal of this work is to show that it is possible to define such a taxonomy. While the taxonomy suggested here should be thought of as a ''first attempt'' it is nonetheless coherent, covers well the relevant topics and coincides with what professional debaters actually argue in their speeches, and facilitates automatic argument invention for new topics.
SemEval 2019 Task 1: Cross-lingual Semantic Parsing with UCCA.
Daniel Hershcovich, Zohar Aizenbud, Leshem Choshen, Elior Sulem, Ari Rappoport and Omri Abend.
SemEval 2019 shared task.

pdf preprint proceedings abstract bib slides website

We present the SemEval 2019 shared task on Universal Conceptual Cognitive Annotation (UCCA) parsing in English, German and French, and discuss the participating systems and results. UCCA is a cross-linguistically applicable framework for semantic representation, which builds on extensive typological work and supports rapid annotation. UCCA poses a challenge for existing parsing techniques, as it exhibits reentrancy (resulting in DAG structures), discontinuous structures and non-terminal nodes corresponding to complex semantic units. The shared task has yielded improvements over the state-of-the-art baseline in all languages and settings. Full results can be found in the task's website.
Syntactic Interchangeability in Word Embedding Models.
Daniel Hershcovich, Assaf Toledo, Alon Halfon and Noam Slonim.
RepEval 2019.

pdf preprint proceedings abstract bib poster slide code

Nearest neighbors in word embedding models are commonly observed to be semantically similar, but the relations between them can vary greatly. We investigate the extent to which word embedding models preserve syntactic interchangeability, as reflected by distances between word vectors, and the effect of hyper-parameters---context window size in particular. We use part of speech (POS) as a proxy for syntactic interchangeability, as generally speaking, words with the same POS are syntactically valid in the same contexts. We also investigate the relationship between interchangeability and similarity as judged by commonly-used word similarity benchmarks, and correlate the result with the performance of word embedding models on these benchmarks. Our results will inform future research and applications in the selection of word embedding model, suggesting a principle for an appropriate selection of the context window size parameter depending on the use-case.
Content Differences in Syntactic and Semantic Representations.
Daniel Hershcovich, Omri Abend and Ari Rappoport.
NAACL 2019 (long paper).

pdf preprint proceedings abstract bib supplementary material poster code data examples

Syntactic analysis plays an important role in semantic parsing, but this role remains a topic of ongoing debate. The debate has been constrained by the scarcity of empirical comparative studies between syntactic and semantic schemes, which hinders the development of parsing methods informed by the details of target schemes and constructions. We target this gap, and take Universal Dependencies (UD) and UCCA as a test case. After abstracting away from differences of convention or formalism, we find that most content divergences can be ascribed to: (1) UCCA's distinction between a Scene and a non-Scene; (2) UCCA's distinction between primary relations, secondary ones and participants; (3) different treatment of multi-word expressions, and (4) different treatment of inter-clause linkage. We further discuss the long tail of cases where the two schemes take markedly different approaches. Finally, we show that the proposed comparison methodology can be used for fine-grained evaluation of UCCA parsing, highlighting both challenges and potential sources for improvement. The substantial differences between the schemes suggest that semantic parsers are likely to benefit downstream text understanding applications beyond their syntactic counterparts.
Universal Dependency Parsing with a General Transition-Based DAG Parser.
Daniel Hershcovich, Omri Abend and Ari Rappoport.
CoNLL 2018 UD Shared Task.

pdf preprint proceedings abstract bib poster code

This paper presents our experiments with applying TUPA to the CoNLL 2018 UD shared task. TUPA is a general neural transition-based DAG parser, which we use to present the first experiments on recovering enhanced dependencies as part of the general parsing task. TUPA was designed for parsing UCCA, a cross-linguistic semantic annotation scheme, exhibiting reentrancy, discontinuity and non-terminal nodes. By converting UD trees and graphs to a UCCA-like DAG format, we train TUPA almost without modification on the UD parsing task. The generic nature of our approach lends itself naturally to multitask learning.
Multitask Parsing Across Semantic Representations.
Daniel Hershcovich, Omri Abend and Ari Rappoport.
ACL 2018 (long paper).

pdf preprint proceedings abstract bib supplementary material poster code data

The ability to consolidate information of different types is at the core of intelligence, and has tremendous practical value in allowing learning for one task to benefit from generalizations learned for others. In this paper we tackle the challenging task of improving semantic parsing performance, taking UCCA parsing as a test case, and AMR, SDP and Universal Dependencies (UD) parsing as auxiliary tasks. We experiment on three languages, using a uniform transition-based system and learning architecture for all parsing tasks. Despite notable conceptual, formal and domain differences, we show that multitask learning significantly improves UCCA parsing in both in-domain and out-of-domain settings.
A Transition-Based Directed Acyclic Graph Parser for UCCA.
Daniel Hershcovich, Omri Abend and Ari Rappoport.
ACL 2017 (long paper), Outstanding Paper Award.

pdf preprint proceedings abstract bib supplementary material slides poster code demo data

We present the first parser for UCCA, a cross-linguistically applicable framework for semantic representation, which builds on extensive typological work and supports rapid annotation. UCCA poses a challenge for existing parsing techniques, as it exhibits reentrancy (resulting in DAG structures), discontinuous structures and non-terminal nodes corresponding to complex semantic units. To our knowledge, the conjunction of these formal properties is not supported by any existing parser. Our transition-based parser, which uses a novel transition set and features based on bidirectional LSTMs, has value not just for UCCA parsing: its ability to handle more general graph structures can inform the development of parsers for other semantic DAG structures, and in languages that frequently use discontinuous structures.
Automatic Claim Negation: Why, How and When.
Yonatan Bilu, Daniel Hershcovich, Noam Slonim.
NAACL HLT 2015 (long paper).

proceedings abstract bib

The main goal of argumentation mining is to analyze argumentative structures within an argument-rich document, and reason about their composition. Recently, there is also interest in the task of simply detecting claims (sometimes called conclusion) in general documents. In this work we ask how this set of detected claims can be augmented further, by adding to it the negation of each detected claim. This presents two NLP problems: how to automatically negate a claim, and when such a negated claim can plausibly be used. We present first steps into solving both these problems, using a rule-based approach for the former and a statistical one towards the latter.
Context Dependent Claim Detection.
Ran Levy, Yonatan Bilu, Daniel Hershcovich, Ehud Aharoni and Noam Slonim.
COLING 2014 (long paper).

proceedings abstract bib

While discussing a concrete controversial topic, most humans will find it challenging to swiftly raise a diverse set of convincing and relevant claims that should set the basis of their arguments. Here, we formally define the challenging task of automatic claim detection in a given context and discuss its associated unique difficulties. Further, we outline a preliminary solution to this task, and assess its performance over annotated real world data, collected specifically for that purpose over hundreds of Wikipedia articles. We report promising results of a supervised learning approach, which is based on a cascade of classifiers designed to properly handle the skewed data which is inherent to the defined task. These results demonstrate the viability of the introduced task.
Claims on Demand–an Initial Demonstration of a System for Automatic Detection and Polarity Identification of Context Dependent Claims in Massive Corpora.
Ehud Aharoni, Carlos Alzate, Roy Bar-Haim, Yonatan Bilu, Lena Dankin, Iris Eiron, Daniel Hershcovich, Shay Hummel.
COLING 2014 (system demonstration).

proceedings abstract bib

While discussing a concrete controversial topic, most humans will find it challenging to swiftly raise a diverse set of convincing and relevant claims that should set the basis of their arguments. Here, we demonstrate the initial capabilities of a system that, given a controversial topic, can automatically pinpoint relevant claims in Wikipedia, determine their polarity with respect to the given topic, and articulate them per the user's request.
A Benchmark Dataset for Automatic Detection of Claims and Evidence in the Context of Controversial Topics.
Ehud Aharoni, Anatoly Polnarov, Tamar Lavee, Daniel Hershcovich, Ran Levy, Ruty Rinott, Dan Gutfreund, Noam Slonim.
Workshop on Argumentation Mining at ACL 2014.

proceedings abstract bib

We describe a novel and unique argumentative structure dataset. This corpus consists of data extracted fro m hundreds of Wikipedia articles using a meticulously monitored manual annotation process. The result is 2,683 argument elements, collected in the context of 33 controversial topics, organized under a simple claim-evidence structure. The obtained data are publicly available for academic research.
Verification of Transactional Memory in POWER8.
Alon Adir, Dave Goodman, Daniel Hershcovich, Oz Hershkovitz, Bryan Hickerson, Karen Holtz, Wisam Kadry, Anatoly Koyfman, John Ludden, Charles Meissner, Amir Nahir, Randall R Pratt, Mike Schiffli, Brett St Onge, Brian Thompto, Elena Tsanko, Avi Ziv.
DAC 2014.

pdf abstract bib

Transactional memory is a promising mechanism for synchronizing concurrent programs that eliminates locks at the expense of hardware complexity. Transactional memory is a hard feature to verify. First, transactions comprise several instructions that must be observed as a single global atomic operation. In addition, there are many reasons a transaction can fail. This results in a high level of non-determinism which must be tamed by the verification methodology. This paper describes the innovation that was applied to tools and methodology in pre-silicon simulation, acceleration and post-silicon in order to verify transactional memory in the IBM POWER8 processor core.

Broad-Coverage Semantic Parsing: A Transition-Based Approach
The Israeli Seminar on Computational Linguistics, ISCOL 2016 (05/2016)
Broad-Coverage Transition-Based UCCA Parsing
Hebrew University, Computer Science Learning Club (11/2016)
A Transition-Based Directed Acyclic Graph Parser for Universal Conceptual Cognitive Annotation
Technion, Computational Data Science Seminar (06/2017)
A Transition-Based Directed Acyclic Graph Parser for Universal Conceptual Cognitive Annotation
University of Washington, The Paul G. Allen Center for Computer Science & Engineering (07/2017)
A Transition-Based Directed Acyclic Graph Parser for Universal Conceptual Cognitive Annotation
Tel Aviv University, NLP Seminar (01/2018)
Universal Semantic Parsing with Neural Networks
Georgetown University (06/2019)
Universal Meaning Representation Parsing
Seminar in Computational Linguistics, Department of Linguistics and Philology, Uppsala University (11/2019)
Meaning Representation Parsing
DIKU Bits, Department of Computer Science, University of Copenhagen (02/2020)
Meaning Representation and Parsing in Natural Language Processing
Seminar at the Department of Food and Resource Economics, University of Copenhagen (05/2021)
Challenges and Strategies in Cross-Cultural NLP
NLP Workshop, IT University of Copenhagen (03/2022)
Cultural and Environmental Considerations in Natural Language Processing
Seminar at University of Haifa, Department of Information Systems (03/2022)
Challenges and Strategies in Cross-Cultural NLP
LIIR Journal Club, KU Leuven (04/2022)
Argument Mining for Green Nutrition
SODAS-Climate meeting, University of Copenhagen (05/2022)

Human Language from a Computational Perspective (TA): 2014, 2015, 2016, 2017

materials

Natural Language Processing: 2019, 2020

materials
Introduction to Machine Learning: 2020

Workshop co-chair for EMNLP 2022.
Co-organized ISCOL 2017, the annual meeting of the Israeli seminar on computational linguistics, as well as SemEval 2019 Task 1 on Cross-lingual Semantic Parsing with UCCA, and the CoNLL 2019 and 2020 shared tasks on Cross-Framework Meaning Representation Parsing.
Co-presented a tutorial on Cross-lingual Semantic Representation for NLP with UCCA at COLING 2020.
Guest editor for the Künstliche Intelligenz Special Issue on NLP and Semantics.
Area chair for NAACL-HLT 2021, ACL-IJCNLP 2021 and *SEM 2021.
Action editor for ACL Rolling Review since October 2021.
Reviewer for ACL (2015, 2016, 2017, 2019, 2020), EMNLP (2017, 2018: Best Reviewer Award, 2019, 2020: Outstanding Reviewer), IJCNLP 2017, COLING (2018, 2020), NAACL-HLT 2019, *SEM (2019, 2020), DMR (2019, 2020), NoDaLiDa (2019, 2021), EurNLP 2019, AAAI 2020, IJCAI 2020, IWPT (2020, 2021) AACL-IJCNLP 2020, CoNLL 2020, TLT 2020, EACL 2021 and the journals Computational Linguistics, Computer Speech & Language and Language Resources and Evaluation.

I practice Kendo (Japanese fencing). I was even at the World Championships in Japan (2015) and Korea (2018).

Daniel Hershcovich
Department of Computer Science
University of Copenhagen
Lyngbyvej 2
DK-2100 Copenhagen Ø, Denmark
d... @di.ku.dk

News

Research Interests

Publications

Talks

Teaching

Hebrew University of Jerusalem

University of Copenhagen

Community

Miscellaneous

Contact Info