Abstractive text summarization aims at compressing the information of a long source document into a rephrased, condensed summary. Despite advances in modeling techniques, abstractive summarization models still suffer from several key challenges: (i) layout bias: they overfit to the style of training corpora; (ii) limited abstractiveness: they are optimized to copying n-grams from the source rather than generating novel abstractive summaries; (iii) lack of transparency: they are not interpretable. In this work, we propose a framework based on document-level structure induction for summarization to address these challenges. To this end, we propose incorporating latent and explicit dependencies across sentences in the source document into end-to-end single-document summarization models. Our framework complements standard encoder-decoder summarization models by augmenting them with rich structure-aware document representations based on implicitly learned (latent) structures and externally-derived linguistic (explicit) structures. We show that our summarization framework, trained on the CNN/DM dataset, improves the coverage of content in the source documents, generates more abstractive summaries by generating more novel n-grams, and incorporates interpretable sentence-level structures, while performing on par with standard baselines.

Traditional named entity recognition models use gazetteers (lists of entities) as features to improve performance. Although modern neural network models do not require such hand-crafted features for strong performance, recent work has demonstrated their utility for named entity recognition on English data. However, designing such features for low-resource languages is challenging, because exhaustive entity gazetteers do not exist in these languages. To address this problem, we propose a method of “soft gazetteers” that incorporates ubiquitously available information from English knowledge bases, such as Wikipedia, into neural named entity recognition models through cross-lingual entity linking. Our experiments on four low-resource languages show an average improvement of 4 points in F1 score.

Current natural language processing models work well on a single task, yet they often fail to continuously learn new tasks without forgetting previous ones as they are re-trained throughout their lifetime, a challenge known as lifelong learning. State-of-the-art lifelong language learning methods store past examples in episodic memory and replay them at both training and inference time. However, as we show later in our experiments, there are three significant impediments: (1) needing unrealistically large memory module to achieve good performance, (2) suffering from negative transfer, (3) requiring multiple local adaptation steps for each test example that significantly slows down the inference speed. In this paper, we identify three common principles of lifelong learning methods and propose an efficient meta-lifelong framework that combines them in a synergistic fashion. To achieve sample efficiency, our method trains the model in a manner that it learns a better initialization for local adaptation. Extensive experiments on text classification and question answering benchmarks demonstrate the effectiveness of our framework by achieving state-of-the-art performance using merely 1% memory size and narrowing the gap with multi-task learning. We further show that our method alleviates both catastrophic forgetting and negative transfer at the same time.

Cross-lingual entity linking (XEL) is the task of finding referents in a target-language knowledge base (KB) for mentions extracted from source-language texts. The first step of (X)EL is candidate generation, which retrieves a list of plausible candidate entities from the target-language KB for each mention. Approaches based on resources from Wikipedia have proven successful in the realm of relatively high-resource languages, but these do not extend well to low-resource languages with few, if any, Wikipedia pages. Recently, transfer learning methods have been shown to reduce the demand for resources in the low-resource languages by utilizing resources in closely related languages, but the performance still lags far behind their high-resource counterparts. In this paper, we first assess the problems faced by current entity candidate generation methods for low-resource XEL, then propose three improvements that (1) reduce the disconnect between entity mentions and KB entries, and (2) improve the robustness of the model to low-resource scenarios. The methods are simple, but effective: We experiment with our approach on seven XEL datasets and find that they yield an average gain of 16.9% in Top-30 gold candidate recall, compared with state-of-the-art baselines. Our improved model also yields an average gain of 7.9% in in-KB accuracy of end-to-end XEL.1

Most state-of-the-art models for named entity recognition (NER) rely on the availability of large amounts of labeled data, making them challenging to extend to new, lower-resourced languages. However, there are now many proposed solutions to this problem involving either cross-lingual transfer learning, which learns from other highly resourced languages, or active learning, which efficiently selects effective training data based on model predictions. In this paper, we ask the question: given this recent progress, and some amount of human annotation, what is the most effective method for efficiently creating high-quality entity recognizers in under-resourced languages? Based on extensive experimentation using both simulated and real human annotation, we settle on a recipe of starting with a cross-lingual transferred model, then performing targeted annotation of only uncertain entity spans in the target language, minimizing annotator effort. Results demonstrate that cross-lingual transfer is a powerful tool when very little data can be annotated, but an entity-targeted annotation strategy can achieve competitive accuracy quickly, with just one-tenth of training data.

Existing recurrent neural language models often fail to capture higher-level structure present in text: for example, rhyming patterns present in poetry. Much prior work on poetry generation uses manually defined constraints which are satisfied during decoding using either specialized decoding procedures or rejection sampling. The rhyming constraints themselves are typically not learned by the generator. We propose an alternate approach that uses a structured discriminator to learn a poetry generator that directly captures rhyming constraints in a generative adversarial setup. By causing the discriminator to compare poems based only on a learned similarity matrix of pairs of line ending words, the proposed approach is able to successfully learn rhyming patterns in two different English poetry datasets (Sonnet and Limerick) without explicitly being provided with any phonetic information

Transformers have a potential of learning longer-term dependency, but are limited by a fixed-length context in the setting of language modeling. We propose a novel neural architecture Transformer-XL that enables learning dependency beyond a fixed length without disrupting temporal coherence. It consists of a segment-level recurrence mechanism and a novel positional encoding scheme. Our method not only enables capturing longer-term dependency, but also resolves the context fragmentation problem. As a result, Transformer-XL learns dependency that is 80% longer than RNNs and 450% longer than vanilla Transformers, achieves better performance on both short and long sequences, and is up to 1,800+ times faster than vanilla Transformers during evaluation. Notably, we improve the state-of-the-art results of bpc/perplexity to 0.99 on enwiki8, 1.08 on text8, 18.3 on WikiText-103, 21.8 on One Billion Word, and 54.5 on Penn Treebank (without finetuning). When trained only on WikiText-103, Transformer-XL manages to generate reasonably coherent, novel text articles with thousands of tokens. Our code, pretrained models, and hyperparameters are available in both Tensorflow and PyTorch.

It has been previously noted that neural machine translation (NMT) is very sensitive to domain shift. In this paper, we argue that this is a dual effect of the highly lexicalized nature of NMT, resulting in failure for sentences with large numbers of unknown words, and lack of supervision for domain-specific words. To remedy this problem, we propose an unsupervised adaptation method which fine-tunes a pre-trained out-of-domain NMT model using a pseudo-in-domain corpus. Specifically, we perform lexicon induction to extract an in-domain lexicon, and construct a pseudo-parallel in-domain corpus by performing word-for-word back-translation of monolingual in-domain target sentences. In five domains over twenty pairwise adaptation settings and two model architectures, our method achieves consistent improvements without using any in-domain parallel sentences, improving up to 14 BLEU over unadapted models, and up to 2 BLEU over strong back-translation baselines.

This paper presents the submission by the CMU-01 team to the SIGMORPHON 2019 task 2 of Morphological Analysis and Lemmatization in Context. This task requires us to produce the lemma and morpho-syntactic description of each token in a sequence, for 107 treebanks. We approach this task with a hierarchical neural conditional random field (CRF) model which predicts each coarse-grained feature (eg. POS, Case, etc.) independently. However, most treebanks are under-resourced, thus making it challenging to train deep neural models for them. Hence, we propose a multi-lingual transfer training regime where we transfer from multiple related languages that share similar typology.

For languages with no annotated resources, unsupervised transfer of natural language processing models such as named-entity recognition (NER) from resource-rich languages would be an appealing capability. However, differences in words and word order across languages make it a challenging problem. To improve mapping of lexical items across languages, we propose a method that finds translations based on bilingual word embeddings. To improve robustness to word order differences, we propose to use self-attention, which allows for a degree of flexibility with respect to word order. We demonstrate that these methods achieve state-of-the-art or competitive NER performance on commonly tested languages under a cross-lingual setting, with much lower resource requirements than past approaches. We also evaluate the challenges of applying these methods to Uyghur, a low-resource language.

This paper introduces the surface construction labeling (SCL) task, which expands the coverage of Shallow Semantic Parsing (SSP) to include frames triggered by complex constructions. We present DeepCx, a neural, transition-based system for SCL. As a test case for the approach, we apply DeepCx to the task of tagging causal language in English, which relies on a wider variety of constructions than are typically addressed in SSP. We report substantial improvements over previous tagging efforts on a causal language dataset. We also propose ways DeepCx could be extended to still more difficult constructions and to other semantic domains once appropriate datasets become available.

Much work in Natural Language Processing (NLP) has been for resource-rich languages, making generalization to new, less-resourced languages challenging. We present two approaches for improving generalization to low-resourced languages by adapting continuous word representations using linguistically motivated subword units: phonemes, morphemes and graphemes. Our method requires neither parallel corpora nor bilingual dictionaries and provides a significant gain in performance over previous methods relying on these resources. We demonstrate the effectiveness of our approaches on Named Entity Recognition for four languages, namely Uyghur, Turkish, Bengali and Hindi, of which Uyghur and Bengali are low resource languages, and also perform experiments on Machine Translation. Exploiting subwords with transfer learning gives us a boost of +15.2 NER F1 for Uyghur and +9.7 F1 for Bengali. We also show improvements in the monolingual setting where we achieve (avg.) +3 F1 and (avg.) +1.35 BLEU.

Neural models have shown several state-of-the-art performances on Semantic Role Labeling (SRL). However, the neural models require an immense amount of semantic-role corpora and are thus not well suited for low-resource languages or domains. The paper proposes a semi-supervised semantic role labeling method that outperforms the state-of-the-art in limited SRL training corpora. The method is based on explicitly enforcing syntactic constraints by augmenting the training objective with a syntactic-inconsistency loss component and uses SRL-unlabeled instances to train a joint-objective LSTM. On CoNLL-2012 English section, the proposed semi-supervised training with 1%, 10% SRL-labeled data and varying amounts of SRL-unlabeled data achieves +1.58, +0.78 F1, respectively, over the pre-trained models that were trained on SOTA architecture with ELMo on the same SRL-labeled data. Additionally, by using the syntactic-inconsistency loss on inference time, the proposed model achieves +3.67, +2.1 F1 over pre-trained model on 1%, 10% SRL-labeled data, respectively.

This paper explores extending shallow semantic parsing beyond lexical-unit triggers, using causal relations as a test case. Semantic parsing becomes difficult in the face of the wide variety of linguistic realizations that causation can take on. We therefore base our approach on the concept of constructions from the linguistic paradigm known as Construction Grammar (CxG). In CxG, a construction is a form/function pairing that can rely on arbitrary linguistic and semantic features. Rather than codifying all aspects of each construction’s form, as some attempts to employ CxG in NLP have done, we propose methods that offload that problem to machine learning. We describe two supervised approaches for tagging causal constructions and their arguments. Both approaches combine automatically induced pattern-matching rules with statistical classifiers that learn the subtler parameters of the constructions. Our results show that these approaches are promising: they significantly outperform naïve baselines for both construction recognition and cause and effect head matches.

Language of cause and effect captures an essential component of the semantics of a text. However, causal language is also intertwined with other semantic relations, such as temporal precedence and correlation. This makes it difficult to determine when causation is the primary intended meaning. This paper presents BECauSE 2.0, a new version of the BECauSE corpus with exhaustively annotated expressions of causal language, but also seven semantic relations that are frequently co-present with causation. The new corpus shows high inter-annotator agreement, and yields insights both about the linguistic expressions of causation and about the process of annotating co-present semantic relations.

Event extraction has become one of the most important topics in information extraction, but to date, there is very limited work on leveraging cross-lingual training to boost performance. We propose a new event extraction approach that trains on multiple languages using a combination of both language-dependent and language-independent features, with particular focus on the case where target domain training data is of very limited size. We show empirically that multilingual training can boost performance for the tasks of event trigger extraction and event argument extraction on the Chinese ACE 2005 dataset.

This paper describes a suite of tools for extracting conventionalized metaphors in English, Spanish, Farsi, and Russian. The method depends on three significant resources for each language: a corpus of conventionalized metaphors, a table of conventionalized conceptual metaphors (CCM table), and a set of extraction rules. Conventionalized metaphors are things like “escape from poverty” and “burden of taxation”. For each metaphor, the CCM table contains the metaphorical source domain word (such as “escape”) the target domain word (such as “poverty”) and the grammatical construction in which they can be found. The extraction rules operate on the output of a dependency parser and identify the grammatical configurations (such as a verb with a prepositional phrase complement) that are likely to contain conventional metaphors. We present results on detection rates for conventional metaphors and analysis of the similarity and differences of source domains for conventional metaphors in the four languages.

Fast and effective automated indexing is critical for search and personalized services. Key phrases that consist of one or more words and represent the main concepts of the document are often used for the purpose of indexing. In this paper, we investigate the use of additional semantic features and pre-processing steps to improve automatic key phrase extraction. These features include the use of signal words and freebase categories. Some of these features lead to significant improvements in the accuracy of the results. We also experimented with 2 forms of document pre-processing that we call light filtering and co-reference normalization. Light filtering removes sentences from the document, which are judged peripheral to its main content. Co-reference normalization unifies several written forms of the same named entity into a unique form. We also needed a “Gold Standard” ― a set of labeled documents for training and evaluation. While the subjective nature of key phrase selection precludes a true “Gold Standard”, we used Amazon's Mechanical Turk service to obtain a useful approximation. Our data indicates that the biggest improvements in performance were due to shallow semantic features, news categories, and rhetorical signals (nDCG 78.47% vs. 68.93%). The inclusion of deeper semantic features such as Freebase sub-categories was not beneficial by itself, but in combination with pre-processing, did cause slight improvements in the nDCG scores.

Large scale parallel data generation for new language pairs requires intensive human effort and availability of experts. It becomes immensely difficult and costly to provide Statistical Machine Translation (SMT) systems for most languages due to the paucity of expert translators to provide parallel data. Even if experts are present, it appears infeasible due to the impending costs. In this paper we propose Active Crowd Translation (ACT), a new paradigm where active learning and crowd-sourcing come together to enable automatic translation for low-resource language pairs. Active learning aims at reducing cost of label acquisition by prioritizing the most informative data for annotation, while crowd-sourcing reduces cost by using the power of the crowds to make do for the lack of expensive language experts. We experiment and compare our active learning strategies with strong baselines and see significant improvements in translation quality. Similarly, our experiments with crowd-sourcing on Mechanical Turk have shown that it is possible to create parallel corpora using non-experts and with sufficient quality assurance, a translation system that is trained using this corpus approaches expert quality.

Producing machine translation (MT) for the many minority languages in the world is a serious challenge. Minority languages typically have few resources for building MT systems. For many minor languages there is little machine readable text, few knowledgeable linguists, and little money available for MT development. For these reasons, our research programs on minority language MT have focused on leveraging to the maximum extent two resources that are available for minority languages: linguistic structure and bilingual informants. All natural languages contain linguistic structure. And although the details of that linguistic structure vary from language to language, language universals such as context-free syntactic structure and the paradigmatic structure of inflectional morphology, allow us to learn the specific details of a minority language. Similarly, most minority languages possess speakers who are bilingual with the major language of the area. This paper discusses our efforts to utilize linguistic structure and the translation information that bilingual informants can provide in three sub-areas of our rapid development MT program: morphology induction, syntactic transfer rule learning, and refinement of imperfect learned rules.

Context-Based Machine TranslationTM (CBMT) is a new paradigm for corpus-based translation that requires no parallel text. Instead, CBMT relies on a lightweight translation model utilizing a fullform bilingual dictionary and a sophisticated decoder using long-range context via long n-grams and cascaded overlapping. The translation process is enhanced via in-language substitution of tokens and phrases, both for source and target, when top candidates cannot be confirmed or resolved in decoding. Substitution utilizes a synonym and near-synonym generator implemented as a corpus-based unsupervised learning process. Decoding requires a very large target-language-only corpus, and while substitution in target can be performed using that same corpus, substitution in source requires a separate (and smaller) source monolingual corpus. Spanish-to-English CBMT was tested on Spanish newswire text, achieving a BLEU score of 0.6462 in June 2006, the highest BLEU reported for any language pair. Further testing also shows that quality increases above the reported score as the target corpus size increases and as dictionary coverage of source words and phrases becomes more complete.

This paper compares a manually written MT grammar and a grammar learned automatically from an English-Spanish elicitation corpus with the ultimate purpose of automatically refining the translation rules. The experiment described here shows that the kind of automatic refinement operations required to correct a translation not only varies depending on the type of error, but also on the type of grammar. This paper describes the two types of grammars and gives a detailed error analysis of their output, indicating what kinds of refinements are required in each case.

Many corpus-based Machine Translation (MT) systems generate a number of partial translations which are then pieced together rather than immediately producing one overall translation. While this makes them more robust to ill-formed input, they are subject to disfluencies at phrasal translation boundaries even for well-formed input. We address this “boundary friction” problem by introducing a method that exploits overlapping phrasal translations and the increased confidence in translation accuracy they imply. We specify an efficient algorithm for producing translations using overlap. Finally, our empirical analysis indicates that this approach produces higher quality translations than the standard method of combining non-overlapping fragments generated by our Example-Based MT (EBMT) system in a peak-to-peak comparison.

Machine Translation of minority languages presents unique challenges, including the paucity of bilingual training data and the unavailability of linguistically-trained speakers. This paper focuses on a machine learning approach to transfer-based MT, where data in the form of translations and lexical alignments are elicited from bilingual speakers, and a seeded version-space learning algorithm formulates and refines transfer rules. A rule-generalization lattice is defined based on LFG-style f-structures, permitting generalization operators in the search for the most general rules consistent with the elicited data. The paper presents these methods and illustrates examples.

NICE is a machine translation project for low-density languages. We are building a tool that will elicit a controlled corpus from a bilingual speaker who is not an expert in linguistics. The corpus is intended to cover major typological phenomena, as it is designed to work for any language. Using implicational universals, we strive to minimize the number of sentences that each informant has to translate. From the elicited sentences, we learn transfer rules with a version space algorithm. Our vision for MT in the future is one in which systems can be quickly trained for new languages by native speakers, so that speakers of minor languages can participate in education, health care, government, and internet without having to give up their languages.

Knowledge-based interlingual machine translation systems produce semantically accurate translations, but typically require massive knowledge acquisition. This paper describes KANT, a system that reduces this requirement to produce practical, scalable, and accurate KBMT applications. First, the set of requirements is discussed, then the full KANT architecture is illustrated, and finally results from a fully implemented prototype are presented.