Search

The recent explosion in question answering research produced a wealth of both factoid reading comprehension (RC) and commonsense reasoning datasets. Combining them presents a different kind of task: deciding not simply whether information is present in the text, but also whether a confident guess could be made for the missing information. We present QuAIL, the first RC dataset to combine text-based, world knowledge and unanswerable questions, and to provide question type annotation that would enable diagnostics of the reasoning strategies by a given QA system. QuAIL contains 15K multi-choice questions for 800 texts in 4 domains. Crucially, it offers both general and text-specific questions, unlikely to be found in pretraining data. We show that QuAIL poses substantial challenges to the current state-of-the-art systems, with a 30% drop in accuracy compared to the most similar existing dataset.

The field of NLP has made substantial progress in building meaning representations. However, an important aspect of linguistic meaning, social meaning, has been largely overlooked. We introduce the concept of social meaning to NLP and discuss how insights from sociolinguistics can inform work on representation learning in NLP. We also identify key challenges for this new line of research.

Knowledge Graph Embeddings (KGEs) have been intensively explored in recent years due to their promise for a wide range of applications. However, existing studies focus on improving the final model performance without acknowledging the computational cost of the proposed approaches, in terms of execution time and environmental impact. This paper proposes a simple yet effective KGE framework which can reduce the training time and carbon footprint by orders of magnitudes compared with state-of-the-art approaches, while producing competitive performance. We highlight three technical innovations: full batch learning via relational matrices, closed-form Orthogonal Procrustes Analysis for KGEs, and non-negative-sampling training. In addition, as the first KGE method whose entity embeddings also store full relation information, our trained models encode rich semantics and are highly interpretable. Comprehensive experiments and ablation studies involving 13 strong baselines and two standard datasets verify the effectiveness and efficiency of our algorithm., To appear at NAACL 2021

Masked language models have quickly become the de facto standard when processing text. Recently, several approaches have been proposed to further enrich word representations with external knowledge sources such as knowledge graphs. However, these models are devised and evaluated in a monolingual setting only. In this work, we propose a language-independent entity prediction task as an intermediate training procedure to ground word representations on entity semantics and bridge the gap across different languages by means of a shared vocabulary of entities. We show that our approach effectively injects new lexical-semantic knowledge into neural models, improving their performance on different semantic tasks in the zero-shot crosslingual setting. As an additional advantage, our intermediate training does not require any supplementary input, allowing our models to be applied to new datasets right away. In our experiments, we use Wikipedia articles in up to 100 languages and already observe consistent gains compared to strong baselines when predicting entities using only the English Wikipedia. Further adding extra languages lead to improvements in most tasks up to a certain point, but overall we found it non-trivial to scale improvements in model transferability by training on ever increasing amounts of Wikipedia languages.

BERT-based architectures currently give state-of-the-art performance on many NLP tasks, but little is known about the exact mechanisms that contribute to its success. In the current work, we focus on the interpretation of self-attention, which is one of the fundamental underlying components of BERT. Using a subset of GLUE tasks and a set of handcrafted features-of-interest, we propose the methodology and carry out a qualitative and quantitative analysis of the information encoded by the individual BERT's heads. Our findings suggest that there is a limited set of attention patterns that are repeated across different heads, indicating the overall model overparametrization. While different heads consistently use the same attention patterns, they have varying impact on performance across different tasks. We show that manually disabling attention in certain heads leads to a performance improvement over the regular fine-tuned BERT models., Accepted to EMNLP 2019

There is a growing body of work that proposes methods for mitigating bias in machine learning systems. These methods typically rely on access to protected attributes such as race, gender, or age. However, this raises two significant challenges: (1) protected attributes may not be available or it may not be legal to use them, and (2) it is often desirable to simultaneously consider multiple protected attributes, as well as their intersections. In the context of mitigating bias in occupation classification, we propose a method for discouraging correlation between the predicted probability of an individual’s true occupation and a word embedding of their name. This method leverages the societal biases that are encoded in word embeddings, eliminating the need for access to protected attributes. Crucially, it only requires access to individuals’ names at training time and not at deployment time. We evaluate two variations of our proposed method using a large-scale dataset of online biographies. We find that both variations simultaneously reduce race and gender biases, with almost no reduction in the classifier’s overall true positive rate.

Normalization of clinical text involves linking different ways of talking about the same clinical concept to the same term in the standardized vocabulary. To date, very few annotated corpora for normalization have been available, and existing corpora so far have been limited in scope and only dealt with the normalization of diseases and disorders. In this paper, we describe the annotation methodology we developed in order to create a new manually annotated wide-coverage corpus for clinical concept normalization, the Medical Concept Normalization (MCN) corpus. In order to ensure wider coverage, we applied normalization to the text spans corresponding to the medical problems, treatments, and tests in the named entity corpus released for the fourth i2b2/VA shared task. In contrast to previous annotation efforts, we do not assign multiple concept labels to the named entities that do not map to a unique concept in the controlled vocabulary. Nor do we leave that named entity without a concept label. Instead, our normalization method that splits such named entities, resolving some of the core ambiguity issues. Lastly, we supply a sieve-based normalization baseline for MCN which combines MetaMap with multiple exact match components. The resulting corpus consists of 100 discharge summaries and provides normalization for the total of 10,919 concept mentions, using 3792 unique concepts from two controlled vocabularies. Our inter-annotator agreement is 67.69% pre-adjudication and 74.20% post-adjudication. Our sieve-based normalization baseline for MCN achieves 77% accuracy in cross-validation. We also detail the challenges of creating a normalization corpus, including the limitations deriving from both the mention span selection and the ambiguity and inconsistency within the current standardized terminologies. In order to facilitate the development of improved concept normalization methods, the MCN corpus will be publicly released to the research community in a shared task in 2019.

In this paper, we present a method for adversarial decomposition of text representation. This method can be used to decompose a representation of an input sentence into several independent vectors, each of them responsible for a specific aspect of the input sentence. We evaluate the proposed method on two case studies: the conversion between different social registers and diachronic language change. We show that the proposed method is capable of fine-grained controlled change of these aspects of the input sentence. It is also learning a continuous (rather than categorical) representation of the style of the sentence, which is more linguistically realistic. The model uses adversarial-motivational training and includes a special motivational loss, which acts opposite to the discriminator and encourages a better decomposition. Furthermore, we evaluate the obtained meaning embeddings on a downstream task of paraphrase detection and show that they significantly outperform the embeddings of a regular autoencoder., Accepted at NAACL 2019

Normalization maps clinical terms in medical notes to concepts in standardized medical vocabularies. To complement the traditional lexical transformation based approach, we propose a hybrid normalization system which incorporates a deep learning model to capture semantic similarity between different surface expressions of the same concept. When evaluating our system against the mentions which may be normalized to existing concepts in the ShARe/CLEF eHealth 2013 dataset, our hybrid system achieves 90.6% in accuracy and outperforms a strong exact match + edit distance baseline by 2.6%. The results suggest the potential of the deep learning model to further improve the performance of normalization by mapping concept mentions to concepts using semantic similarity.

We propose a context-aware neural network model for temporal information extraction. This model has a uniform architecture for event-event, event-timex and timex-timex pairs. A Global Context Layer (GCL), inspired by Neural Turing Machine (NTM), stores processed temporal relations in narrative order, and retrieves them for use when relevant entities come in. Relations are then classified in context. The GCL model has long-term memory and attention mechanisms to resolve irregular long-distance dependencies that regular RNNs such as LSTM cannot recognize. It does not require any new input features, while outperforming the existing models in literature. To our knowledge it is also the first model to use NTM-like architecture to process the information from global context in discourse-scale natural text processing. We are going to release the source code in the future.

Display Omitted The first study to assess the ease of adoption of the state-of-the-art clinical NLP systems.Five clinical NLP systems were carefully examined by four expert evaluators and eight end user evaluators.The results show that the adoptability of clinical NLP systems is generally unsatisfactory. ObjectiveIn recognition of potential barriers that may inhibit the widespread adoption of biomedical software, the 2014 i2b2 Challenge introduced a special track, Track 3 - Software Usability Assessment, in order to develop a better understanding of the adoption issues that might be associated with the state-of-the-art clinical NLP systems. This paper reports the ease of adoption assessment methods we developed for this track, and the results of evaluating five clinical NLP system submissions. Materials and methodsA team of human evaluators performed a series of scripted adoptability test tasks with each of the participating systems. The evaluation team consisted of four "expert evaluators" with training in computer science, and eight "end user evaluators" with mixed backgrounds in medicine, nursing, pharmacy, and health informatics. We assessed how easy it is to adopt the submitted systems along the following three dimensions: communication effectiveness (i.e., how effective a system is in communicating its designed objectives to intended audience), effort required to install, and effort required to use. We used a formal software usability testing tool, TURF, to record the evaluators' interactions with the systems and 'think-aloud' data revealing their thought processes when installing and using the systems and when resolving unexpected issues. ResultsOverall, the ease of adoption ratings that the five systems received are unsatisfactory. Installation of some of the systems proved to be rather difficult, and some systems failed to adequately communicate their designed objectives to intended adopters. Further, the average ratings provided by the end user evaluators on ease of use and ease of interpreting output are -0.35 and -0.53, respectively, indicating that this group of users generally deemed the systems extremely difficult to work with. While the ratings provided by the expert evaluators are higher, 0.6 and 0.45, respectively, these ratings are still low indicating that they also experienced considerable struggles. DiscussionThe results of the Track 3 evaluation show that the adoptability of the five participating clinical NLP systems has a great margin for improvement. Remedy strategies suggested by the evaluators included (1) more detailed and operation system specific use instructions; (2) provision of more pertinent onscreen feedback for easier diagnosis of problems; (3) including screen walk-throughs in use instructions so users know what to expect and what might have gone wrong; (4) avoiding jargon and acronyms in materials intended for end users; and (5) packaging prerequisites required within software distributions so that prospective adopters of the software do not have to obtain each of the third-party components on their own.

Annotation is one of the main vehicles for supplying knowledge to machine learning systems built to automate text processing tasks. In this chapter, we discuss how linguistic annotation is used in machine learning for different natural language processing (NLP) tasks. Specifically, we focus on how different layers of annotation are leveraged in tasks that aim to discover higher-level linguistic information. We present how machine learning fits into the annotation process in the MATTER cycle, discuss some common machine learning algorithms used in NLP, explain the fundamentals of feature selection, and explore methods for leveraging limited quantities of annotated data. We close with a case study of the 2012 i2b2 NLP shared task which targeted temporal information extraction, a higher-level task that requires a synthesis of information from multiple linguistic levels.

This paper describes the winning system for SemEval-2017 Task 6: #HashtagWars: Learning a Sense of Humor. Humor detection has up until now been predominantly addressed using feature-based approaches. Our system utilizes recurrent deep learning methods with dense embeddings to predict humorous tweets from the @midnight show #HashtagWars. In order to include both meaning and sound in the analysis, GloVe embeddings are combined with a novel phonetic representation to serve as input to an LSTM component. The output is combined with a character-based CNN model, and an XGBoost component in an ensemble model which achieves 0.675 accuracy on the evaluation data.

This paper describes a new shared task for humor understanding that attempts to eschew the ubiquitous binary approach to humor detection and focus on comparative humor ranking instead. The task is based on a new dataset of funny tweets posted in response to shared hashtags, collected from the ‘Hashtag Wars’ segment of the TV show @midnight. The results are evaluated in two subtasks that require the participants to generate either the correct pairwise comparisons of tweets (subtask A), or the correct ranking of the tweets (subtask B) in terms of how funny they are. 7 teams participated in subtask A, and 5 teams participated in subtask B. The best accuracy in subtask A was 0.675. The best (lowest) rank edit distance for subtask B was 0.872.

The ability to predict psychiatric readmission would facilitate the development of interventions to reduce this risk, a major driver of psychiatric health-care costs. The symptoms or characteristics of illness course necessary to develop reliable predictors are not available in coded billing data, but may be present in narrative electronic health record (EHR) discharge summaries. We identified a cohort of individuals admitted to a psychiatric inpatient unit between 1994 and 2012 with a principal diagnosis of major depressive disorder, and extracted inpatient psychiatric discharge narrative notes. Using these data, we trained a 75-topic Latent Dirichlet Allocation (LDA) model, a form of natural language processing, which identifies groups of words associated with topics discussed in a document collection. The cohort was randomly split to derive a training (70%) and testing (30%) data set, and we trained separate support vector machine models for baseline clinical features alone, baseline features plus common individual words and the above plus topics identified from the 75-topic LDA model. Of 4687 patients with inpatient discharge summaries, 470 were readmitted within 30 days. The 75-topic LDA model included topics linked to psychiatric symptoms (suicide, severe depression, anxiety, trauma, eating/weight and panic) and major depressive disorder comorbidities (infection, postpartum, brain tumor, diarrhea and pulmonary disease). By including LDA topics, prediction of readmission, as measured by area under receiver-operating characteristic curves in the testing data set, was improved from baseline (area under the curve 0.618) to baseline+1000 words (0.682) to baseline+75 topics (0.784). Inclusion of topics derived from narrative notes allows more accurate discrimination of individuals at high risk for psychiatric readmission in this cohort. Topic modeling and related approaches offer the potential to improve prediction using EHRs, if generalizability can be established in other clinical cohorts.

Background The Sixth Informatics for Integrating Biology and the Bedside (i2b2) Natural Language Processing Challenge for Clinical Records focused on the temporal relations in clinical narratives. The organizers provided the research community with a corpus of discharge summaries annotated with temporal information, to be used for the development and evaluation of temporal reasoning systems. 18 teams from around the world participated in the challenge. During the workshop, participating teams presented comprehensive reviews and analysis of their systems, and outlined future research directions suggested by the challenge contributions. Methods The challenge evaluated systems on the information extraction tasks that targeted: (1) clinically significant events, including both clinical concepts such as problems, tests, treatments, and clinical departments, and events relevant to the patient's clinical timeline, such as admissions, transfers between departments, etc; (2) temporal expressions, referring to the dates, times, durations, or frequencies phrases in the clinical text. The values of the extracted temporal expressions had to be normalized to an ISO specification standard; and (3) temporal relations, between the clinical events and temporal expressions. Participants determined pairs of events and temporal expressions that exhibited a temporal relation, and identified the temporal relation between them. Results For event detection, statistical machine learning (ML) methods consistently showed superior performance. While ML and rule based methods seemed to detect temporal expressions equally well, the best systems overwhelmingly adopted a rule based approach for value normalization. For temporal relation classification, the systems using hybrid approaches that combined ML and heuristics based methods produced the best results.

One of the major goals in automated argumentation mining is to uncover the argument structure present in argumentative text. In order to determine this structure, one must understand how different individual components of the overall argument are linked. General consensus in this field dictates that the argument components form a hierarchy of persuasion, which manifests itself in a tree structure. This work provides the first neural network-based approach to argumentation mining, focusing on the two tasks of extracting links between argument components, and classifying types of argument components. In order to solve this problem, we propose to use a joint model that is based on a Pointer Network architecture. A Pointer Network is appealing for this task for the following reasons: 1) It takes into account the sequential nature of argument components; 2) By construction, it enforces certain properties of the tree structure present in argument relations; 3) The hidden representations can be applied to auxiliary tasks. In order to extend the contribution of the original Pointer Network model, we construct a joint model that simultaneously attempts to learn the type of argument component, as well as continuing to predict links between argument components. The proposed joint model achieves state-of-the-art results on two separate evaluation corpora, achieving far superior performance than a regular Pointer Network model. Our results show that optimizing for both tasks, and adding a fully-connected layer prior to recurrent neural network input, is crucial for high performance., Comment: 10 pages; under review for ICLR

Precise evaluation metrics are important for assessing progress in high-level language generation tasks such as machine translation or image captioning. Historically, these metrics have been evaluated using correlation with human judgment. However, human-derived scores are often alarmingly inconsistent and are also limited in their ability to identify precise areas of weakness. In this paper, we perform a case study for metric evaluation by measuring the effect that systematic sentence transformations (e.g. active to passive voice) have on the automatic metric scores. These sentence “corruptions” serve as unit tests for precisely measuring the strengths and weaknesses of a given metric. We find that not only are human annotations heavily inconsistent in this study, but that the Metric Unit TesT analysis is able to capture precise shortcomings of particular metrics (e.g. comparing passive and active sentences) better than a simple correlation with human judgment can.

In this paper, we explore the inherent tension between corpus data and linguistic theory that aims to model it, with particular reference to the dynamic and variable nature of the lexicon. We explore the process through which modeling of the data is accomplished, presenting itself as a sequence of conflicting stages of discovery. First-stage data analysis informs the model, whereas the seeming chaos of organic data inevitably violates our theoretical assumptions. But in the end, it is restrictions apparent in the data that call for postulating structure within a revised theoretical model. We show the complete cycle using two case studies and discuss the implications.

We analyze the RI-TIMEXes in temporally annotated corpora and propose two hypotheses regarding the normalization of RI-TIMEXes in the clinical narrative domain: the anchor point hypothesis and the anchor relation hypothesis. We annotate the RI-TIMEXes in three corpora to study the characteristics of RI-TMEXes in different domains. This informed the design of our RI-TIMEX normalization system for the clinical domain, which consists of an anchor point classifier, an anchor relation classifier and a rule-based RI-TIMEX text span parser. We experiment with different feature sets and perform error analysis for each system component. The annotation confirmed the hypotheses that we can simplify the RI-TIMEXes normalization task using two multi-label classifiers. Our system achieves anchor point classification, anchor relation classification and rule-based parsing accuracy of 74.68%, 87.71% and 57.2% (82.09% under relaxed matching criteria) respectively on the held-out test set of the 2012 i2b2 temporal relation challenge. Experiments with feature sets reveals some interesting findings such as the verbal tense feature does not inform the anchor relation classification in clinical narratives as much as the tokens near the RI-TIMEX. Error analysis shows that underrepresented anchor point and anchor relation classes are difficult to detect. We formulate the RI-TIMEX normalization problem as a pair of multi-label classification problems. Considering only the RI-TIMEX extraction and normalization, the system achieves statistically significant improvement over the RI-TIMEX results of the best systems in the 2012 i2b2 challenge., Comment: Draft version

This paper demonstrates the effectiveness of a Long Short-Term Memory language model in our initial efforts to generate unconstrained rap lyrics. The goal of this model is to generate lyrics that are similar in style to that of a given rapper, but not identical to existing lyrics: this is the task of ghostwriting. Unlike previous work, which defines explicit templates for lyric generation, our model defines its own rhyme scheme, line length, and verse length. Our experiments show that a Long Short-Term Memory language model produces better “ghostwritten” lyrics than a baseline model.

This paper describes TwitterHawk, a system for sentiment analysis of tweets which participated in the SemEval-2015 Task 10, Subtasks A through D. The system performed competitively, most notably placing 1 st in topicbased sentiment classification (Subtask C) and ranking 4 th out of 40 in identifying the sentiment of sarcastic tweets. Our submissions in all four subtasks used a supervised learning approach to perform three-way classification to assign positive, negative, or neutral labels. Our system development efforts focused on text pre-processing and feature engineering, with a particular focus on handling negation, integrating sentiment lexicons, parsing hashtags, and handling expressive word modifications and emoticons. Two separate classifiers were developed for phrase-level and tweetlevel sentiment classification. Our success in aforementioned tasks came in part from leveraging the Subtask B data and building a single tweet-level classifier for Subtasks B, C and D.

Accurate knowledge of a patient's disease state and trajectory is critical in a clinical setting. Modern electronic healthcare records contain an increasingly large amount of data, and the ability to automatically identify the factors that influence patient outcomes stand to greatly improve the efficiency and quality of care. We examined the use of latent variable models (viz. Latent Dirichlet Allocation) to decompose free-text hospital notes into meaningful features, and the predictive power of these features for patient mortality. We considered three prediction regimes: (1) baseline prediction, (2) dynamic (time-varying) outcome prediction, and (3) retrospective outcome prediction. In each, our prediction task differs from the familiar time-varying situation whereby data accumulates; since fewer patients have long ICU stays, as we move forward in time fewer patients are available and the prediction task becomes increasingly difficult. We found that latent topic-derived features were effective in determining patient mortality under three timelines: in-hospital, 30 day post-discharge, and 1 year post-discharge mortality. Our results demonstrated that the latent topic features important in predicting hospital mortality are very different from those that are important in post-discharge mortality. In general, latent topic features were more predictive than structured features, and a combination of the two performed best. The time-varying models that combined latent topic features and baseline features had AUCs that reached 0.85, 0.80, and 0.77 for in-hospital, 30 day post-discharge and 1 year post-discharge mortality respectively. Our results agreed with other work suggesting that the first 24 hours of patient information are often the most predictive of hospital mortality. Retrospective models that used a combination of latent topic features and structured features achieved AUCs of 0.96, 0.82, and 0.81 for in-hospital, 30 day, and 1-year mortality prediction. Our work focuses on the dynamic (time-varying) setting because models from this regime could facilitate an on-going severity stratification system that helps direct care-staff resources and inform treatment strategies., Intel Corporation. Science and Technology Center for Big Data, National Library of Medicine (U.S.) (Biomedical Informatics Research Training Grant NIH/NLM 2T15 LM007092-22)

Objective To evaluate state-of-the-art unsupervised methods on the word sense disambiguation (WSD) task in the clinical domain. In particular, to compare graph-based approaches relying on a clinical knowledge base with bottom-up topic-modeling-based approaches. We investigate several enhancements to the topic-modeling techniques that use domain-specific knowledge sources. Materials and methods The graph-based methods use variations of PageRank and distance-based similarity metrics, operating over the Unified Medical Language System (UMLS). Topic-modeling methods use unlabeled data from the Multiparameter Intelligent Monitoring in Intensive Care (MIMIC II) database to derive models for each ambiguous word. We investigate the impact of using different linguistic features for topic models, including UMLS-based and syntactic features. We use a sense-tagged clinical dataset from the Mayo Clinic for evaluation. Results The topic-modeling methods achieve 66.9% accuracy on a subset of the Mayo Clinic's data, while the graph-based methods only reach the 40–50% range, with a most-frequent-sense baseline of 56.5%. Features derived from the UMLS semantic type and concept hierarchies do not produce a gain over bag-of-words features in the topic models, but identifying phrases from UMLS and using syntax does help. Discussion Although topic models outperform graph-based methods, semantic features derived from the UMLS prove too noisy to improve performance beyond bag-of-words. Conclusions Topic modeling for WSD provides superior results in the clinical domain; however, integration of knowledge remains to be effectively exploited.

In this paper, we introduce a methodology for annotating compositional operations in natural language text and describe the Generative Lexicon Mark-up Language (GLML), a mark-up language inspired by the Generative Lexicon model, for identifying such relations. While most annotation systems capture surface relationships, GLML captures the “compositional history” of the argument selection relative to the predicate. We provide a brief overview of GL before moving on to our proposed methodology for annotating with GLML. There are three main tasks described in the paper. The first one is based on atomic semantic types and the other two exploit more fine-grained meaning parameters encoded in the Qualia Structure roles: (i) Argument Selection and Coercion Annotation for the SemEval-2010 competition; (ii) Qualia Selection in modification constructions; (iii) Type selection in modification constructions and verb-noun combinations involving dot objects. We explain what each task comprises and include the XML format for annotated sample sentences. We show that by identifying and subsequently annotating the typing and subtyping shifts in these constructions, we gain an insight into the workings of the general mechanisms of composition.

Temporal information in clinical narratives plays an important role in patients’ diagnosis, treatment and prognosis. In order to represent narrative information accurately, medical natural language processing (MLP) systems need to correctly identify and interpret temporal information. To promote research in this area, the Informatics for Integrating Biology and the Bedside (i2b2) project developed a temporally annotated corpus of clinical narratives. This corpus contains 310 de-identified discharge summaries, with annotations of clinical events, temporal expressions and temporal relations. This paper describes the process followed for the development of this corpus and discusses annotation guideline development, annotation methodology, and corpus quality.

Domain-specific knowledge is often recorded by experts in the form of unstructured text. For example, in the medical domain, clinical notes from electronic health records contain a wealth of information. Similar practices are found in other domains. The challenge we discuss in this paper is how to identify and extract part names from technicians repair notes, a noisy unstructured text data source from General Motors' archives of solved vehicle repair problems, with the goal to develop a robust and dynamic reasoning system to be used as a repair adviser by service technicians.In the present work, we discuss two approaches to this problem. We present an algorithm for ontology-guided entity disambiguation that uses existing knowledge sources such as domain-specific ontologies and other structured data. We illustrate its use in automotive domain, using GM parts ontology and the unit structure of repair manuals text to build context models, which are then used to disambiguate mentions of part-related entities in the text. We also describe extraction of part names with a small amount of annotated data using Hidden Markov Models (HMM) with shrinkage, achieving an f-score of approximately 80%. Next we used linear-chain Conditional Random Fields (CRF) in order to model observation dependencies present in the repair notes. Using CRF did not lead to improved performance, but a slight improvement over the HMM results was obtained by using a weighted combination of the HMM and CRF models.

In this paper we introduce a methodology for annotating compositional operations in natural language text, and describe a mark-up language, GLML, based on Generative Lexicon, for identifying such relations. While most annotation systems capture surface relationships, GLML captures the "compositional history" of the argument selection relative to the predicate. We provide a brief overview of GL before moving on to our proposed methodology for annotating with GLML. There are three main tasks described in the paper: (i) Compositional mechanisms of argument selection; (ii) Qualia in modification constructions; (iii) Type selection in modification of dot objects. We explain what each task includes and provide a description of the annotation interface. We also include the XML format for GLML including examples of annotated sentences.

Sense inventories for polysemous predicates are often comprised by a number of related senses. In this paper, we examine different types of relations within sense inventories and give a qualitative analysis of the effects they have on decisions made by the annotators and annotator error. We also discuss some common traps and pitfalls in design of sense inventories. We use the data set developed specifically for the task of annotating sense distinctions dependent predominantly on semantics of the arguments and only to a lesser extent on syntactic frame.

In this paper we consider the problem of identifying and classifying discourse coherence relations. We report initial results over the recently released Discourse GraphBank (Wolf and Gibson, 2005). Our approach considers, and determines the contributions of, a variety of syntactic and lexico-semantic features. We achieve 81% accuracy on the task of discourse relation type classification and 70% accuracy on relation identification.

We present an overview of TARSQI, a modular system for automatic temporal annotation that adds time expressions, events and temporal relations to news texts.

In this paper, we introduce a model for sense assignment which relies on assigning senses to the contexts within which words appear, rather than to the words themselves. We argue that word senses as such are not directly encoded in the lexicon of the language. Rather, each word is associated with one or more stereotypical syntagmatic patterns, which we call selection contexts. Each selection context is associated with a meaning, which can be expressed in any of various formal or computational manifestations. We present a formalism for encoding contexts that help to determine the semantic contribution of a word in an utterance. Further, we develop a methodology through which such stereotypical contexts for words and phrases can be identified from very large corpora, and subsequently structured in a selection context dictionary, encoding both stereotypical syntactic and semantic information. We present some preliminary results.

The mapping of lexical meanings to wordforms is a major feature of natural languages. While usage pressures might assign short words to frequent meanings (Zipf’s law of abbreviation), the need for a productive and open-ended vocabulary, local constraints on sequences of symbols, and various other factors all shape the lexicons of the world’s languages. Despite their importance in shaping lexical structure, the relative contributions of these factors have not been fully quantified. Taking a coding-theoretic view of the lexicon and making use of a novel generative statistical model, we define upper bounds for the compressibility of the lexicon under various constraints. Examining corpora from 7 typologically diverse languages, we use those upper bounds to quantify the lexicon’s optimality and to explore the relative costs of major constraints on natural codes. We find that (compositional) morphology and graphotactics can sufficiently account for most of the complexity of natural codes—as measured by code length., Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, ISBN:978-1-954085-46-6

Analysing whether neural language models encode linguistic information has become popular in NLP. One method of doing so, which is frequently cited to support the claim that models like BERT encode syntax, is called probing; probes are small supervised models trained to extract linguistic information from another model’s output. If a probe is able to predict a particular structure, it is argued that the model whose output it is trained on must have implicitly learnt to encode it. However, drawing a generalisation about a model’s linguistic knowledge about a specific phenomena based on what a probe is able to learn may be problematic: in this work, we show that semantic cues in training data means that syntactic probes do not properly isolate syntax. We generate a new corpus of semantically nonsensical but syntactically well-formed Jabberwocky sentences, which we use to evaluate two probes trained on normal data. We train the probes on several popular language models (BERT, GPT-2, and RoBERTa), and find that in all settings they perform worse when evaluated on these data, for one probe by an average of 15.4 UUAS points absolute. Although in most cases they still outperform the baselines, their lead is reduced substantially, e.g. by 53% in the case of BERT for one probe. This begs the question: what empirical scores constitute knowing syntax?, Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, ISBN:978-1-954085-46-6