_id stringlengths 4 10 | text stringlengths 0 18.4k | title stringlengths 0 8.56k |
|---|---|---|
d246904453 | This paper presents FAMIE, a comprehensive and efficient active learning (AL) toolkit for multilingual information extraction. FAMIE is designed to address a fundamental problem in existing AL frameworks where annotators need to wait for a long time between annotation batches due to the time-consuming nature of model training and data selection at each AL iteration. This hinders the engagement, productivity, and efficiency of annotators. Based on the idea of using a small proxy network for fast data selection, we introduce a novel knowledge distillation mechanism to synchronize the proxy network with the main large model (i.e., BERT-based) to ensure the appropriateness of the selected annotation examples for the main model. Our AL framework can support multiple languages. The experiments demonstrate the advantages of FAMIE in terms of competitive performance and time efficiency for sequence labeling with AL. We publicly release our code (https://github.com/ nlp-uoregon/famie) and demo website (http://nlp.uoregon.edu:9000/). A demo video for FAMIE is provided at: https://youtu.be/I2i8n_jAyrY. | FAMIE: A Fast Active Learning Framework for Multilingual Information Extraction |
d10982256 | This is the version of an article from the following conference: Pizzato, Luiz Augusto Sangoi and Mollá-Aliod, Diego (2005) Extracting exact answers using a meta question answering system Australasian Language Technology Workshop (3rd : 2005) (10 -11 December 2005 : Sydney)Access to the published version:Abstract This work concerns a question answering tool that uses multiple Web search engines and Web question answering systems to retrieve snippets of text that may contain an exact answer for a natural language question. The method described here treats each Web information retrieval system in a unique manner in order to extract the best results they can provide. The results obtained suggest that our method is comparable with some of today's state-of-the-art systems. | Macquarie University ResearchOnline Extracting Exact Answers using a Meta Question Answering System |
d259370593 | Anxiety disorders are the most common of mental illnesses, but relatively little is known about how to detect them from language. The primary clinical manifestation of anxiety is worry associated cognitive distortions, which are likely expressed at the discourse-level of semantics. Here, we investigate the development of a modern linguistic assessment for degree of anxiety, specifically evaluating the utility of discourselevel information in addition to lexical-level large language model embeddings. We find that a combined lexico-discourse model outperforms models based solely on state-of-theart contextual embeddings (RoBERTa), with discourse-level representations derived from Sentence-BERT and DiscRE both providing additional predictive power not captured by lexical-level representations. Interpreting the model, we find that discourse patterns of causal explanations, among others, were used significantly more by those scoring high in anxiety, dovetailing with psychological literature.Franziska Burger, Mark A. Neerincx, and Willem-PaulBrinkman. 2021. Natural language processing for cognitive therapy: Extracting schemas from thought records. PLOS ONE, 16:1-24. | Discourse-Level Representations can Improve Prediction of Degree of Anxiety |
d248721945 | This paper describes the SLT-CDT-UoS group's submission to the first Special Task on Formality Control for Spoken Language Translation, part of the IWSLT 2022 Evaluation Campaign. Our efforts were split between two fronts: data engineering and altering the objective function for best hypothesis selection. We used language-independent methods to extract formal and informal sentence pairs from the provided corpora; using English as a pivot language, we propagated formality annotations to languages treated as zero-shot in the task; we also further improved formality controlling with a hypothesis re-ranking approach. On the test sets for English-to-German and Englishto-Spanish, we achieved an average accuracy of .935 within the constrained setting and .995 within unconstrained setting. In a zero-shot setting for English-to-Russian and English-to-Italian, we scored average accuracy of .590 for constrained setting and .659 for unconstrained. | Controlling Formality in Low-Resource NMT with Domain Adaptation and Re-Ranking: SLT-CDT-UoS at IWSLT2022 |
d10261368 | This paper describes the Howard University system for the language identification shared task of the Second Workshop on Computational Approaches to Code Switching. Our system is based on prior work on Swahili-English token-level language identification. Our system primarily uses character n-gram, prefix and suffix features, letter case and special character features along with previously existing tools. These are then combined with generated label probabilities of the immediate context of the token for the final system. | |
d222291089 | This paper describes our submission of the WMT 2020 Shared Task on Sentence Level Direct Assessment, Quality Estimation (QE). In this study, we empirically reveal the mismatching issue when directly adopting BERTScore(Zhang et al., 2020)to QE. Specifically, there exist lots of mismatching errors between source sentence and translated candidate sentence with token pairwise similarity. In response to this issue, we propose to expose explicit cross lingual patterns, e.g. word alignments and generation score, to our proposed zero-shot models. Experiments show that our proposed QE model with explicit cross-lingual patterns could alleviate the mismatching issue, thereby improving the performance. Encouragingly, our zero-shot QE method could achieve comparable performance with supervised QE method, and even outperforms the supervised counterpart on 2 out of 6 directions. We expect our work could shed light on the zero-shot QE model improvement. | Zero-Shot Translation Quality Estimation with Explicit Cross-Lingual Patterns |
d10282694 | We present a simple preordering approach for machine translation based on a featurerich logistic regression model to predict whether two children of the same node in the source-side parse tree should be swapped or not. Given the pair-wise children regression scores we conduct an efficient depth-first branch-and-bound search through the space of possible children permutations, avoiding using a cascade of classifiers or limiting the list of possible ordering outcomes. We report experiments in translating English to Japanese and Korean, demonstrating superior performance as (a) the number of crossing links drops by more than 10% absolute with respect to other state-of-the-art preordering approaches, (b) BLEU scores improve on 2.2 points over the baseline with lexicalised reordering model, and (c) decoding can be carried out 80 times faster. * This work was done during an internship of the first author at SDL Research, Cambridge. | Source-side Preordering for Translation using Logistic Regression and Depth-first Branch-and-Bound Search * |
d7126582 | The paper presents the work being done so far on the building of the Croatian Morphological Lexicon (CML). It has been collected since 2002 in the Institute of Linguistics, Faculty of Philosophy, University of Zagreb. The CML is planned to have two sub-lexicons: derivative/compositional and inflectional, both produced by a generator. The result of generation is lexicon as two distinct lists of generated combinations of morphemes and complete word-forms both with additional data that can be used in further processing. The inflectional component is presented more in detail in the second part of the paper. At the end, the several possible applications of CML are discussed. | Building the Croatian Morphological Lexicon |
d258588434 | Automatic image captioning evaluation is critical for benchmarking and promoting advances in image captioning research. Existing metrics only provide a single score to measure caption qualities, which are less explainable and informative. Instead, we humans can easily identify the problems of captions in details, e.g., which words are inaccurate and which salient objects are not described, and then rate the caption quality. | InfoMetIC: An Informative Metric for Reference-free Image Caption Evaluation |
d233407579 | A LightGBM model fed with target word lexical characteristics and features obtained from word frequency lists, psychometric data and bigram association measures has been optimized for the 2021 CMCL Shared Task on Eye-Tracking Data Prediction. It obtained the best performance of all teams on two of the five eye-tracking measures to predict, allowing it to rank first on the official challenge criterion and to outperform all deep-learning based systems participating in the challenge. | LAST at CMCL 2021 Shared Task: Predicting Gaze Data During Reading with a Gradient Boosting Decision Tree Approach |
d3574064 | We participate in the two event extraction tasks of BioNLP 2016 Shared Task: binary relation extraction of SeeDev task and localization relations extraction of Bacteria Biotope task. Convolutional neural network (CNN) is employed to model the sentences by convolution and maxpooling operation from raw input with word embedding. Then, full connected neural network is used to learn senior and significant features automatically. The proposed model mainly contains two modules: distributive semantic representation building, such as word embedding, POS embedding, distance embedding and entity type embedding, and CNN model training. The results with F-score of 0.370 and 0.478 in our participant tasks, which were evaluated on the test data set, show that our proposed method contributes to binary relation extraction effectively and can reduce the impact of artificial feature engineering through automatically feature learning. | DUTIR in BioNLP-ST 2016: Utilizing Convolutional Network and Distributed Representation to Extract Complicate Relations |
d13641515 | This paper presents a perceptual evaluation of a text to speech (TTS) synthesizer in Greek with respect to acoustic registration of enclitic stress and related naturalness and intelligibility. Based on acoustical measurements and observations of naturally recorded utterances, the corresponding output of a commercially available formant-based speech synthesizer was altered and the results were subjected to perceptual evaluation. Pitch curve, intensity, and duration of the syllable bearing enclitic stress, were acoustically manipulated, while a phonetically identical phrase contrasting only in stress served as control stimulus. Ten listeners judged the perceived naturalness and preference (in pairs) and the stress pattern of each variant of a base phrase. It was found that intensity modification adversely affected perceived naturalness while increasing perceived stress prominence. Duration modification had no appreciable effect. Pitch curve modification tended to produce an improvement in perceived naturalness and preference but the results failed to achieve statistical significance. The results indicated that the current prosodic module of the speech synthesizer reflects a good balance between prominence of stress assignment, intelligibility, and naturalness. | Perceptual evaluation of text-to-speech implementation of enclitic stress in Greek |
d216553665 | We present an efficient method of utilizing pretrained language models, where we learn selective binary masks for pretrained weights in lieu of modifying them through finetuning. Extensive evaluations of masking BERT and RoBERTa on a series of NLP tasks show that our masking scheme yields performance comparable to finetuning, yet has a much smaller memory footprint when several tasks need to be inferred simultaneously. Through intrinsic evaluations, we show that representations computed by masked language models encode information necessary for solving downstream tasks. Analyzing the loss landscape, we show that masking and finetuning produce models that reside in minima that can be connected by a line segment with nearly constant test accuracy. This confirms that masking can be utilized as an efficient alternative to finetuning. | Masking as an Efficient Alternative to Finetuning for Pretrained Language Models |
d199373064 | This paper describes the categorisation of Irish MWEs, and the construction of the first version of a lexicon of Irish MWEs for NLP purposes (Ilfhocail, meaning 'Multiwords'), collected from a number of resources. For the purposes of quality assurance, 530 entries of this lexicon were examined and manually annotated for POS and MWE category. | Ilfhocail: A Lexicon of Irish MWEs |
d593533 | We define the crouching Dirichlet, hidden Markov model (CDHMM), an HMM for partof-speech tagging which draws state prior distributions for each local document context. This simple modification of the HMM takes advantage of the dichotomy in natural language between content and function words. In contrast, a standard HMM draws all prior distributions once over all states and it is known to perform poorly in unsupervised and semisupervised POS tagging. This modification significantly improves unsupervised POS tagging performance across several measures on five data sets for four languages. We also show that simply using different hyperparameter values for content and function word states in a standard HMM (which we call HMM+) is surprisingly effective. | Crouching Dirichlet, Hidden Markov Model: Unsupervised POS Tagging with Context Local Tag Generation |
d236478176 | Modern text simplification (TS) heavily relies on the availability of gold standard data to build machine learning models. However, existing studies show that parallel TS corpora contain inaccurate simplifications and incorrect alignments. Additionally, evaluation is usually performed by using metrics such as BLEU or SARI to compare system output to the gold standard. A major limitation is that these metrics do not match human judgements and the performance on different datasets and linguistic phenomena vary greatly. Furthermore, our research shows that the test and training subsets of parallel datasets differ significantly. In this work, we investigate existing TS corpora, providing new insights that will motivate the improvement of existing state-ofthe-art TS evaluation methods. Our contributions include the analysis of TS corpora based on existing modifications used for simplification and an empirical study on TS models performance by using better-distributed datasets. We demonstrate that by improving the distribution of TS datasets, we can build more robust TS models. | Investigating Text Simplification Evaluation |
d53617637 | This paper relies on the work developed by the research team IES_UNR (Argentina) and presents a pedagogical application of NooJ for the teaching and learning of Spanish as a foreign language. However, as this proposal specifically addresses learners of Spanish whose mother tongue is Italian, it also entailed vital collaboration with Mario Monteleone from the University of Salerno, Italy. The adjective was chosen on account of its lower frequency of occurrence in texts written in Spanish, and particularly in the Argentine Rioplatense variety, and with the aim of developing strategies to increase its use. The features that the adjective shares with other grammatical categories render it extremely productive and provide elements that enrich the learner's proficiency. The reference corpus contains the front pages of the Argentinian newspaper Clarín related to an emblematic historical moment, whose starting point is March 24, 1976, when a military coup began, and covers a thirty year period until March 24, 2006. The use of the linguistic resources created in NooJ for the automatic processing of texts written in Spanish accounts for the adjective in a relevant historical context for Argentina. | A Pedagogical Application of NooJ in Language Teaching: the Adjective in Spanish and Italian |
d3904790 | In this paper we report a behavioural experiment documenting that different lexicosyntactic formulations of the discourse relation of causation are deemed more or less acceptable by different categories of readers. We further report promising results for automatically selecting the formulation that is most appropriate for a given category of reader using supervised learning. This investigation is embedded within a longer term research agenda aimed at summarising scientific writing for lay readers using appropriate paraphrasing. | Reformulating Discourse Connectives for Non-Expert Readers |
d247613500 | Nowadays, pretrained language models (PLMs) have dominated the majority of NLP tasks. While, little research has been conducted on systematically evaluating the language abilities of PLMs. In this paper, we present a large-scale empirical study on genEral language ability evaluation of PLMs (ElitePLM). In our study, we design four evaluation dimensions, i.e., memory, comprehension, reasoning, and composition, to measure ten widely-used PLMs within five categories. Our empirical results demonstrate that: (1) PLMs with varying training objectives and strategies are good at different ability tests; (2) fine-tuning PLMs in downstream tasks is usually sensitive to the data size and distribution; (3) PLMs have excellent transferability between similar tasks. Moreover, the prediction results of PLMs in our experiments are released as an open resource for more deep and detailed analysis on the language abilities of PLMs. This paper can guide the future work to select, apply, and design PLMs for specific tasks. We have made all the details of experiments publicly available at https:/github.com/RUCAIBox/ElitePLM. evaluation of summaries. In Text summarization branches out, pages 74-81. Hugo Liu and Push Singh. 2004. Conceptnet-a practical commonsense reasoning tool-kit. BT technology journal, 22(4):211-226. | ElitePLM: An Empirical Study on General Language Ability Evaluation of Pretrained Language Models |
d247619149 | Information extraction suffers from its varying targets, heterogeneous structures, and demandspecific schemas. In this paper, we propose a unified text-to-structure generation framework, namely UIE, which can universally model different IE tasks, adaptively generate targeted structures, and collaboratively learn general IE abilities from different knowledge sources. Specifically, UIE uniformly encodes different extraction structures via a structured extraction language, adaptively generates target extractions via a schema-based prompt mechanism -structural schema instructor, and captures the common IE abilities via a large-scale pretrained text-to-structure model. Experiments show that UIE achieved the state-of-the-art performance on 4 IE tasks, 13 datasets, and on all supervised, low-resource, and few-shot settings for a wide range of entity, relation, event and sentiment extraction tasks and their unification. These results verified the effectiveness, universality, and transferability of UIE 1 . * Part of this work was done when Yaojie Lu and Qing Liu interned at Baidu. † Corresponding authors. | Unified Structure Generation for Universal Information Extraction |
d14922761 | Children with autism spectrum disorder often exhibit idiosyncratic patterns of behaviors and interests. In this paper, we focus on measuring the presence of idiosyncratic interests at the linguistic level in children with autism using distributional semantic models. We model the semantic space of children's narratives by calculating pairwise word overlap, and we compare the overlap found within and across diagnostic groups. We find that the words used by children with typical development tend to be used by other children with typical development, while the words used by children with autism overlap less with those used by children with typical development and even less with those used by other children with autism. These findings suggest that children with autism are veering not only away from the topic of the target narrative but also in idiosyncratic semantic directions potentially defined by their individual topics of interest. | Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal Detecting linguistic idiosyncratic interests in autism using distributional semantic models |
d1052307 | This paper deals with a seldom studied object/oblique alternation phenomenon in Japanese, which. We call this the bump alternation. This phenomenon, first discussed bySadanobu (1990), is similar to the English with/against alternation. For example, compare hit the wall with the bat [=immobile-as-direct-object frame] to hit the bat against the wall [=mobile-asdirect-object frame]). However, in the Japanese version, the case frame remains constant. Although we fundamentally question Sadanobu's acceptability judgment, we also claim that the causation type (i.e., whether the event is an instance of onset or extended causation;Talmy, 1988;2000)could make an improvement. An extended causative interpretation could improve the acceptability of the otherwise awkward immobile-as-direct-object frame. We examined this claim through a rating study, and the results showed an interaction between the Causation type (extended/onset) and the Object type (mobile/immobile) in the direction we predicted. We propose that a perspective shift on what is moving causes the "extended causation" advantage. | A Study of the Bump Alternation in Japanese from the Perspective of Extended/Onset Causation |
d248366286 | Relation extraction is a core problem for natural language processing in the biomedical domain. Recent research on relation extraction showed that prompt-based learning improves the performance on both fine-tuning on full training set and few-shot training. However, less effort has been made on domain-specific tasks where good prompt design can be even harder. In this paper, we investigate prompting for biomedical relation extraction, with experiments on the ChemProt dataset. We present a simple yet effective method to systematically generate comprehensive prompts that reformulate the relation extraction task as a cloze-test task under a simple prompt formulation. In particular, we experiment with different ranking scores for prompt selection. With BioMed-RoBERTa-base, our results show that prompting-based fine-tuning obtains gains by 14.21 F1 over its regular fine-tuning baseline, and 1.14 F1 over SciFive-Large, the current state-of-the-art on ChemProt. Besides, we find prompt-based learning requires fewer training examples to make reasonable predictions. The results demonstrate the potential of our methods in such a domain-specific relation extraction task. | Decorate the Examples: A Simple Method of Prompt Design for Biomedical Relation Extraction |
d258079354 | We investigate in this paper how distributions of occupations with respect to gender is reflected in pre-trained language models. Such distributions are not always aligned to normative ideals, nor do they necessarily reflect a descriptive assessment of reality. In this paper, we introduce an approach for measuring to what degree pre-trained language models are aligned to normative and descriptive occupational distributions. To this end, we use official demographic information about gender-occupation distributions provided by the national statistics agencies of France, Norway, United Kingdom, and the United States. We manually generate template-based sentences combining gendered pronouns and nouns with occupations, and subsequently probe a selection of ten language models covering the English, French, and Norwegian languages. The scoring system we introduce in this work is language independent, and can be used on any combination of template-based sentences, occupations, and languages. The approach could also be extended to other dimensions of national census data and other demographic variables. | Measuring Normative and Descriptive Biases in Language Models Using Census Data |
d245131376 | Generic unstructured neural networks have been shown to struggle on out-of-distribution compositional generalization. Compositional data augmentation via example recombination has transferred some prior knowledge about compositionality to such black-box neural models for several semantic parsing tasks, but this often required task-specific engineering or provided limited gains.We present a more powerful data recombination method using a model called Compositional Structure Learner (CSL). CSL is a generative model with a quasi-synchronous context-free grammar backbone, which we induce from the training data. We sample recombined examples from CSL and add them to the fine-tuning data of a pre-trained sequence-tosequence model (T5). This procedure effectively transfers most of CSL's compositional bias to T5 for diagnostic tasks, and results in a model even stronger than a T5-CSL ensemble on two real world compositional generalization tasks. This results in new state-ofthe-art performance for these challenging semantic parsing tasks requiring generalization to both natural language variation and novel compositions of elements.Alfred V Aho and Jeffrey D Ullman. 1972. The theory of parsing, translation, and compiling, volume 1. Prentice-Hall Englewood Cliffs, NJ.Ekin Akyürek, Afra Feyza Akyürek, and Jacob Andreas. 2021. Learning to recombine and resample data for compositional generalization. | Improving Compositional Generalization with Latent Structure and Data Augmentation |
d1918444 | In this paper, we show that one benefit of FUG, the ability to state global conslralnts on choice separately from syntactic rules, is difficult in generation systems based on augmented context free grammars (e.g., Def'mite Clause Cn'anmm~). They require that such constraints be expressed locally as part of syntactic rules and therefore, duplicated in the grammar. Finally, we discuss a reimplementation of lUg that achieves the similar levels of efficiency as Rubinoff's adaptation of MUMBLE, a detcrministc language generator. | Functional Unification Grammar Revisited |
d233209910 | We study how masking and predicting tokens in an unsupervised fashion can give rise to linguistic structures and downstream performance gains. Recent theories have suggested that pretrained language models acquire useful inductive biases through masks that implicitly act as cloze reductions. While appealing, we show that the success of the random masking strategy used in practice cannot be explained by such cloze-like masks alone. We construct cloze-like masks using task-specific lexicons for three different classification datasets and show that the majority of pretrained performance gains come from generic masks that are not associated with the lexicon. To explain the empirical success of these generic masks, we demonstrate a correspondence between the masked language model (MLM) objective and existing methods for learning statistical dependencies in graphical models. Using this, we derive a method for extracting these learned statistical dependencies in MLMs and show that these dependencies encode useful inductive biases in the form of syntactic structures. In an unsupervised parsing evaluation, simply forming a minimum spanning tree on the implied statistical dependence structure outperforms a classic method for unsupervised parsing (58.74 vs. 55.91 UUAS). | On the Inductive Bias of Masked Language Modeling: From Statistical to Syntactic Dependencies |
d3299357 | In this paper, we present a comparative study of the four state-of-the-art sequential taggers applied on Magahi data for part-of-speech (POS) annotation . Magahi is one of the smaller Indo-Aryan languages spoken in Eastern state of Bihar in India. It is an extremely resource-poor language and it is the first attempt to develop some kind of Natural Language Processing (NLP) resource for the language.The four taggers that we test are -Support Vector Machines (SVM) based SVMTool, Hidden Markov Model (HMM) based TnT tagger, Maximum Entropy based MxPost tagger and Memory based MBT tagger. All these taggers are trained on a miniscule dataset of around 50,000 words using 33 tags from the BIS-tagset for Indian languages and tested on around 13,000 words. The performance of all these taggers are tested against a frequency-based baseline tagger. While all these taggers perform worse than on the English data, the best performance is given by the Maximum Entropy tagger after tuning of certain parameters. The paper discusses the result of the taggers and the ways in which the performance of the taggers could be improved for Magahi. | Developing a POS tagger for Magahi: A Comparative Study |
d247627719 | Identifying argument components from unstructured texts and predicting the relationships expressed among them are two primary steps of argument mining. The intrinsic complexity of these tasks demands powerful learning models. While pretrained Transformerbased Language Models (LM) have been shown to provide state-of-the-art results over different NLP tasks, the scarcity of manually annotated data and the highly domaindependent nature of argumentation restrict the capabilities of such models. In this work, we propose a novel transfer learning strategy to overcome these challenges. We utilize argumentation-rich social discussions from the ChangeMyView subreddit as a source of unsupervised, argumentative discourse-aware knowledge by finetuning pretrained LMs on a selectively masked language modeling task. Furthermore, we introduce a novel promptbased strategy for inter-component relation prediction that compliments our proposed finetuning method while leveraging on the discourse context. Exhaustive experiments show the generalization capability of our method on these two tasks over within-domain as well as out-of-domain datasets, outperforming several existing and employed strong baselines. 1 | Can Unsupervised Knowledge Transfer from Social Discussions Help Argument Mining? |
d248780537 | The rapid development of conversational assistants accelerates the study on conversational question answering (QA). However, the existing conversational QA systems usually answer users' questions with a single knowledge source, e.g., paragraphs or a knowledge graph, but overlook the important visual cues, let alone multiple knowledge sources of different modalities. In this paper, we hence define a novel research task, i.e., multimodal conversational question answering (MMCoQA), aiming to answer users' questions with multimodal knowledge sources via multi-turn conversations. This new task brings a series of research challenges, including but not limited to priority, consistency, and complementarity of multimodal knowledge. To facilitate the data-driven approaches in this area, we construct the first multimodal conversational QA dataset, named MMConvQA. Questions are fully annotated with not only natural language answers but also the corresponding evidence and valuable decontextualized self-contained questions. Meanwhile, we introduce an endto-end baseline model, which divides this complex research task into question understanding, multi-modal evidence retrieval, and answer extraction. Moreover, we report a set of benchmarking results, and the results indicate that there is ample room for improvement. | MMCoQA: Conversational Question Answering over Text, Tables, and Images |
d245877654 | We propose PromptBERT, a novel contrastive learning method for learning better sentence representation. We firstly analyze the drawback of current sentence embedding from original BERT and find that it is mainly due to the static token embedding bias and ineffective BERT layers. Then we propose the first prompt-based sentence embeddings method and discuss two prompt representing methods and three prompt searching methods to make BERT achieve better sentence embeddings. Moreover, we propose a novel unsupervised training objective by the technology of template denoising, which substantially shortens the performance gap between the supervised and unsupervised settings. Extensive experiments show the effectiveness of our method. Compared to SimCSE, PromptBert achieves 2.29 and 2.58 points of improvement based on BERT and RoBERTa in the unsupervised setting. Our code is available at https://github.com/kongds/ Prompt-BERT. | PromptBERT: Improving BERT Sentence Embeddings with Prompts |
d219301744 | ||
d254823170 | While the problem of hallucinations in neural machine translation has long been recognized, so far the progress on its alleviation is very little. Indeed, recently it turned out that without artificially encouraging models to hallucinate, previously existing methods fall short and even the standard sequence log-probability is more informative. It means that internal characteristics of the model can give much more information than we expect, and before using external models and measures, we first need to ask: how far can we go if we use nothing but the translation model itself ? We propose to use a method that evaluates the percentage of the source contribution to a generated translation. Intuitively, hallucinations are translations "detached" from the source, hence they can be identified by low source contribution. This method improves detection accuracy for the most severe hallucinations by a factor of 2 and is able to alleviate hallucinations at test time on par with the previous best approach that relies on external models. Next, if we move away from internal model characteristics and allow external tools, we show that using sentence similarity from cross-lingual embeddings further improves these results. We release the code of our experiments. 1 | Detecting and Mitigating Hallucinations in Machine Translation: Model Internal Workings Alone Do Well, Sentence Similarity Even Better |
d246016165 | Recent work has shown that an answer verification step introduced in Transformer-based answer selection models can significantly improve the state of the art in Question Answering. This step is performed by aggregating the embeddings of top k answer candidates to support the verification of a target answer. Although the approach is intuitive and sound, it still shows two limitations: (i) the supporting candidates are ranked only according to the relevancy with the question and not with the answer, and (ii) the support provided by the other answer candidates is suboptimal as these are retrieved independently of the target answer. In this paper, we address both drawbacks by proposing (i) a double reranking model, which, for each target answer, selects the best support; and (ii) a second neural retrieval stage designed to encode question and answer pair as the query, which finds more specific verification information. The results on well-known datasets for Answer Sentence Selection show significant improvement over the state of the art. | Double Retrieval and Ranking for Accurate Question Answering |
d252595697 | We present Chandojñānam, a web-based Sanskrit meter (chanda) identification and utilization system. In addition to the core functionality of identifying meters, it sports a friendly user interface to display the scansion, which is a graphical representation of the metrical pattern. The system supports identification of meters from uploaded images by using optical character recognition (OCR) engines in the backend. It is also able to process entire text files at a time. The text can be processed in two modes, either by treating it as a list of individual lines, or as a collection of verses. When a line or a verse does not correspond exactly to a known meter, Chandojñānam is capable of finding fuzzy (i.e., approximate and close) matches based on sequence matching. This opens up the scope of a meter based correction of erroneous digital corpora. The system is available for use at https://sanskrit.iitk.ac.in/jnanasangraha/chanda/, and the source code in the form of a Python library is made available at https | Chandojnanam: A Sanskrit Meter Identification and Utilization System |
d150149297 | Recently, a variety of unsupervised methods have been proposed that map pre-trained word embeddings of different languages into the same space without any parallel data. These methods aim to find a linear transformation based on the assumption that monolingual word embeddings are approximately isomorphic between languages. However, it has been demonstrated that this assumption holds true only on specific conditions, and with limited resources, the performance of these methods decreases drastically. To overcome this problem, we propose a new unsupervised multilingual embedding method that does not rely on such assumption and performs well under resource-poor scenarios, namely when only a small amount of monolingual data (i.e., 50k sentences) are available, or when the domains of monolingual data are different across languages. Our proposed model, which we call 'Multilingual Neural Language Models', shares some of the network parameters among multiple languages, and encodes sentences of multiple languages into the same space. The model jointly learns word embeddings of different languages in the same space, and generates multilingual embeddings without any parallel data or pre-training. Our experiments on word alignment tasks have demonstrated that, on the low-resource condition, our model substantially outperforms existing unsupervised and even supervised methods trained with 500 bilingual pairs of words. Our model also outperforms unsupervised methods given different-domain corpora across languages. Our code is publicly available 1 . | Unsupervised Multilingual Word Embedding with Limited Resources using Neural Language Models |
d248406121 | Summarizing sales calls is a routine task performed manually by salespeople. We present a production system which combines generative models fine-tuned for customer-agent setting, with a human-in-the-loop user experience for an interactive summary curation process. We address challenging aspects of dialogue summarization task in a real-world setting including long input dialogues, content validation, lack of labeled data and quality evaluation. We show how GPT-3 can be leveraged as an offline data labeler to handle training data scarcity and accommodate privacy constraints in an industrial setting. Experiments show significant improvements by our models in tackling the summarization and content validation tasks on public datasets. | An End-to-End Dialogue Summarization System for Sales Calls |
d2692228 | We describe an annotation scheme for syntactic information in the CHILDES database(MacWhinney, 2000), which contains several megabytes of transcribed dialogs between parents and children. The annotation scheme is based on grammatical relations (GRs) that are composed of bilexical dependencies (between a head and a dependent) labeled with the name of the relation involving the two words (such as subject, object and adjunct). We also discuss automatic annotation using our syntactic annotation scheme. | Adding Syntactic Annotations to Transcripts of Parent-Child Dialogs |
d218973828 | ||
d253553576 | Radiology report summarization (RRS) is a growing area of research. Given the Findings section of a radiology report, the goal is to generate a summary (called an Impression section) that highlights the key observations and conclusions of the radiology study. However, RRS currently faces essential limitations. First, many prior studies conduct experiments on private datasets, preventing the reproduction of results and fair comparisons across different systems and solutions. Second, most prior approaches are evaluated solely on chest Xrays. To address these limitations, we propose a dataset (MIMIC-RRS) involving three new modalities and seven new anatomies based on the MIMIC-III and MIMIC-CXR datasets. We then conduct extensive experiments to evaluate the performance of models both within and across modality-anatomy pairs in MIMIC-RRS. In addition, we evaluate their clinical efficacy via RadGraph, a factual correctness metric. | Toward Expanding the Scope of Radiology Report Summarization to Multiple Anatomies and Modalities |
d218571035 | We focus on the study of conversational recommendation in the context of multi-type dialogs, where the bots can proactively and naturally lead a conversation from a nonrecommendation dialog (e.g., QA) to a recommendation dialog, taking into account user's interests and feedback. To facilitate the study of this task, we create a human-to-human Chinese dialog dataset DuRecDial (about 10k dialogs, 156k utterances), which contains multiple sequential dialogs for every pair of a recommendation seeker (user) and a recommender (bot). In each dialog, the recommender proactively leads a multi-type dialog to approach recommendation targets and then makes multiple recommendations with rich interaction behavior. This dataset allows us to systematically investigate different parts of the overall problem, e.g., how to naturally lead a dialog, how to interact with users for recommendation. Finally we establish baseline results on DuRecDial for future studies. 1 | Towards Conversational Recommendation over Multi-Type Dialogs |
d227231846 | In the last few years, deep learning has proved to be a very effective paradigm to discover patterns in large data sets. Unfortunately, deep learning training on small data sets is not the best option because most of the time traditional machine learning algorithms could get better scores. Now, we can train the neural network on a large data set then fine-tune on a smaller data set using the transfer learning technique. In this paper, we present our system for NADI shared Task: Countrylevel Dialect Identification, Our system is based on fine-tuning of BERT and it achieves 22.85 F1-score on Test Set and our rank is 5th out of 18 teams. | Arabic Dialect Identification Using BERT Fine-Tuning |
d249605873 | In this paper, we present the first Entity Linking corpus for Icelandic. We describe our approach of using a multilingual entity linking model (mGENRE) in combination with Wikipedia API Search (WAPIS) to label our data and compare it to an approach using WAPIS only. We find that our combined method reaches 53.9% coverage on our corpus, compared to 30.9% using only WAPIS. We analyze our results and explain the value of using a multilingual system when working with Icelandic. Additionally, we analyze the data that remain unlabeled, identify patterns and discuss why they may be more difficult to annotate. | Building an Icelandic Entity Linking Corpus |
d16432357 | Social media provides a wealth of information regarding users' perspectives on issues, public figures and brands, but it can be a timeconsuming and labor-intensive process to develop data pipelines in which those perspectives are encoded, and to build visualizations that illuminate important developments. This paper describes a system for quickly developing a model of the conversation around an issue on Twitter, and a flexible visualization system that allows analysts to interactively explore key facets of the analysis. | An Unsupervised System for Visual Exploration of Twitter Conversations |
d7997131 | Efficient music information retrieval (MIR) require to have meta information about music along with content based information in the knowledge base. Discussion forums on music are rich sources of information gathered from a wider audience. Taking into consideration the nature of text in these web resources, the yield of relation extraction is quite dependent on resolving the entity references in the document. Among the few music forums dealing with Indian classical music, rasikas.org (rasikas, 2015) having rich information about artistes, raga and other music concepts is taken for our study. The forum posts generally contain anaphoric references to the main topic of the thread or any other entity in the discourse. In this paper we focus on coreference resolution for short discourse noisy text like that of forum posts. Since grammatical roles capture relation between mentions in a discourse, those features extracted from dependency parsing are widely explored along with semantic compatibility feature. On investigation of issues, the need for integrating known dependencies between features emerged. A Bayesian network with predefined network structure is evaluated, since a Bayesian belief network enacts a probabilistic rule based system. To the extent possible the superior behaviour of Bayesian network over SVM is analysed. | Coreference Resolution to Support IE from Indian Classical Music Forums |
d237571584 | While coreference resolution is defined independently of dataset domain, most models for performing coreference resolution do not transfer well to unseen domains. We consolidate a set of 8 coreference resolution datasets targeting different domains to evaluate the offthe-shelf performance of models. We then mix three datasets for training; even though their domain, annotation guidelines, and metadata differ, we propose a method for jointly training a single model on this heterogeneous data mixture by using data augmentation to account for annotation differences and sampling to balance the data quantities. We find that in a zeroshot setting, models trained on a single dataset transfer poorly while joint training yields improved overall performance, leading to better generalization in coreference resolution models. This work contributes a new benchmark for robust coreference resolution and multiple new state-of-the-art results. 1 | On Generalization in Coreference Resolution |
d362885 | We describe the DCU-MIXED and DCU-SVR submissions to the WMT-14 Quality Estimation task 1.1, predicting sentencelevel perceived post-editing effort. Feature design focuses on target-side features as we hypothesise that the source side has little effect on the quality of human translations, which are included in task 1.1 of this year's WMT Quality Estimation shared task. We experiment with features of the QuEst framework, features of our past work, and three novel feature sets. Despite these efforts, our two systems perform poorly in the competition. Follow up experiments indicate that the poor performance is due to improperly optimised parameters. | Target-Centric Features for Translation Quality Estimation |
d237604959 | A method for creating a vision-and-language (V&L) model is to extend a language model through structural modifications and V&L pretraining. Such an extension aims to make a V&L model inherit the capability of natural language understanding (NLU) from the original language model. To see how well this is achieved, we propose to evaluate V&L models using an NLU benchmark (GLUE). We compare five V&L models, including singlestream and dual-stream models, trained with the same pre-training. Dual-stream models, with their higher modality independence achieved by approximately doubling the number of parameters, are expected to preserve the NLU capability better. Our main finding is that the dual-stream scores are not much different than the single-stream scores, contrary to expectation. Further analysis shows that pre-training causes the performance drop in NLU tasks with few exceptions. These results suggest that adopting a single-stream structure and devising the pre-training could be an effective method for improving the maintenance of language knowledge in V&L extensions. | Effect of Visual Extensions on Natural Language Understanding in Vision-and-Language Models |
d220445441 | ||
d259360453 | Reasoning has been a central topic in artificial intelligence from the beginning. The recent progress made on distributed representation and neural networks continues to improve the state-of-the-art performance of natural language inference. However, it remains an open question whether the models perform real reasoning to reach their conclusions or rely on spurious correlations. Adversarial attacks have proven to be an important tool to help evaluate the Achilles' heel of the victim models. In this study, we explore the fundamental problem of developing attack models based on logic formalism. We propose NatLogAttack to perform systematic attacks centring around natural logic, a classical logic formalism that is traceable back to Aristotle's syllogism and has been closely developed for natural language inference. The proposed framework renders both label-preserving and label-flipping attacks. We show that compared to the existing attack models, NatLogAttack generates better adversarial examples with fewer visits to the victim models. The victim models are found to be more vulnerable under the label-flipping setting. NatLogAttack provides a tool to probe the existing and future NLI models' capacity from a key viewpoint and we hope more logicbased attacks will be further explored for understanding the desired property of reasoning. 1 | NatLogAttack: A Framework for Attacking Natural Language Inference Models with Natural Logic |
d41422196 | Poor digital representation of minority languages further prevents their usability on digital media and devices. The Digital Language Diversity Project, a three-year project funded under the Erasmus+ programme, aims at addressing the problem of low digital representation of EU regional and minority languages by giving their speakers the intellectual an practical skills to create, share, and reuse online digital content. Availability of digital content and technical support to use it are essential prerequisites for the development of language-based digital applications, which in turn can boost digital usage of these languages. In this paper we introduce the project, its aims, objectives and current activities for sustaining digital usability of minority languages through adult education. | Fostering digital representation of EU regional and minority languages: the Digital Language Diversity Project |
d29944028 | OBLING: A TESTER. FOIL TR`ANSFORMATIONAL' GKAMMAKS | |
d9023840 | This paper describes work on a rule-based, open-source parser for Swedish. The central component is a wide-coverage grammar implemented in the GF formalism (Grammatical Framework), a dependently typed grammar formalism based on Martin-Löf type theory. GF has strong support for multilinguality and has so far been used successfully for controlled languages (Angelov and Ranta, 2009) and recent experiments have showed that it is also possible to use the framework for parsing unrestricted language. In addition to GF, we use two other main resources: the Swedish treebank Talbanken and the electronic lexicon SALDO. By combining the grammar with a lexicon extracted from SALDO we obtain a parser accepting all sentences described by the given rules. We develop and test this on examples from Talbanken. The resulting parser gives a full syntactic analysis of the input sentences. It will be highly reusable, freely available, and as GF provides libraries for compiling grammars to a number of programming languages, chosen parts of the the grammar may be used in various NLP applications. | Combining Language Resources Into A Grammar-Driven Swedish Parser |
d235195626 | Anglicisms are a challenge in German speech recognition. Due to their irregular pronunciation compared to native German words, automatically generated pronunciation dictionaries often contain incorrect phoneme sequences for Anglicisms. In this work, we propose a multitask sequence-to-sequence approach for grapheme-to-phoneme conversion to improve the phonetization of Anglicisms. We extended a grapheme-to-phoneme model with a classification task to distinguish Anglicisms from native German words. With this approach, the model learns to generate different pronunciations depending on the classification result. We used our model to create supplementary Anglicism pronunciation dictionaries to be added to an existing German speech recognition model. Tested on a special Anglicism evaluation set, we improved the recognition of Anglicisms compared to a baseline model, reducing the word error rate by a relative 1 % and the Anglicism error rate by a relative 3 %. With our experiment, we show that multitask learning can help solving the challenge of Anglicisms in German speech recognition. | Multitask Learning for Grapheme-to-Phoneme Conversion of Anglicisms in German Speech Recognition |
d145961136 | ||
d252873184 | Large-scale, high-quality corpora are critical for advancing research in coreference resolution. However, existing datasets vary in their definition of coreferences and have been collected via complex and lengthy guidelines that are curated for linguistic experts. These concerns have sparked a growing interest among researchers to curate a unified set of guidelines suitable for annotators with various backgrounds. In this work, we develop a crowdsourcing-friendly coreference annotation methodology, ezCoref, consisting of an annotation tool and an interactive tutorial. We use ezCoref to re-annotate 240 passages from seven existing English coreference datasets (spanning fiction, news, and multiple other domains) while teaching annotators only cases that are treated similarly across these datasets. 1 Surprisingly, we find that reasonable quality annotations were already achievable (>90% agreement between the crowd and expert annotations) even without extensive training. On carefully analyzing the remaining disagreements, we identify the presence of linguistic cases that our annotators unanimously agree upon but lack unified treatments (e.g., generic pronouns, appositives) in existing datasets. We propose the research community should revisit these phenomena when curating future unified annotation guidelines. OntoNotes: Maybe we need a [CIA] version of the Miranda warning: You have the right to conceal your coup intentions, because we may rat on you. ARRAU: Maybe [we]e1 need [a [CIA] version of [the Miranda warning]]: [You]e4 have [the right to conceal [[your]e5 [coup] intentions]], because [we]e6 may rat on [you]e7. Crowd (this work): Maybe [we]e1 need [a [CIA] version of [the [Miranda] warning]]: [You]e3 have [the right] to conceal [[your]e3 coup intentions], because [we]e1may rat on [you]e3. | ezCoref: Towards Unifying Annotation Guidelines for Coreference Resolution |
d251741522 | User simulators (USs) are commonly used to train task-oriented dialogue systems (DSs) via reinforcement learning. The interactions often take place on semantic level for efficiency, but there is still a gap from semantic actions to natural language, which causes a mismatch between training and deployment environment. Incorporating a natural language generation (NLG) module with USs during training can partly deal with this problem. However, since the policy and NLG of USs are optimised separately, these simulated user utterances may not be natural enough in a given context. In this work, we propose a generative transformerbased user simulator (GenTUS). GenTUS consists of an encoder-decoder structure, which means it can optimise both the user policy and natural language generation jointly. GenTUS generates both semantic actions and natural language utterances, preserving interpretability and enhancing language variation. In addition, by representing the inputs and outputs as word sequences and by using a large pre-trained language model we can achieve generalisability in feature representation. We evaluate Gen-TUS with automatic metrics and human evaluation. Our results show that GenTUS generates more natural language and is able to transfer to an unseen ontology in a zero-shot fashion. In addition, its behaviour can be further shaped with reinforcement learning opening the door to training specialised user simulators. | GenTUS: Simulating User Behaviour and Language in Task-oriented Dialogues with Generative Transformers |
d248157393 | We propose a method to protect the privacy of search engine users by decomposing the queries using semantically related and unrelated distractor terms. Instead of a single query, the search engine receives multiple decomposed query terms. Next, we reconstruct the search results relevant to the original query term by aggregating the search results retrieved for the decomposed query terms. We show that the word embeddings learnt using a distributed representation learning method can be used to find semantically related and distractor query terms. We derive the relationship between the obfuscity achieved through the proposed query anonymisation method and the reconstructability of the original search results using the decomposed queries. We analytically study the risk of discovering the search engine users' information intents under the proposed query obfuscation method, and empirically evaluate its robustness against clustering-based attacks. Our experimental results show that the proposed method can accurately reconstruct the search results for user queries, without compromising the privacy of the search engine users. | Query Obfuscation by Semantic Decomposition |
d219310201 | ||
d235377430 | End-to-end simultaneous speech translation (SST), which directly translates speech in one language into text in another language in realtime, is useful in many scenarios but has not been fully investigated. In this work, we propose RealTranS, an end-to-end model for SST. To bridge the modality gap between speech and text, RealTranS gradually downsamples the input speech with interleaved convolution and unidirectional Transformer layers for acoustic modeling, and then maps speech features into text space with a weighted-shrinking operation and a semantic encoder. Besides, to improve the model performance in simultaneous scenarios, we propose a blank penalty to enhance the shrinking quality and a Wait-K-Stride-N strategy to allow local reranking during decoding. Experiments on public and widely-used datasets show that RealTranS with the Wait-K-Stride-N strategy outperforms prior end-to-end models as well as cascaded models in diverse latency settings. | RealTranS: End-to-End Simultaneous Speech Translation with Convolutional Weighted-Shrinking Transformer |
d18775155 | When we write a report or an explanation on a newly-developed technology for readers including laypersons, it is very important to compose a title that can stimulate their interest in the technology. However, it is difficult for inexperienced authors to come up with an appealing title. In this research, we developed a support system for revising titles. We call it "title revision wizard". The wizard provides a guidance on revising draft title to compose a title meeting three key points, and support tools for coming up with and elaborating on comprehensible or appealing phrases. In order to test the effect of our title revision wizard, we conducted a questionnaire survey on the effect of the titles with or without using the wizard on the interest of lay readers. The survey showed that the wizard is effective and helpful for the authors who cannot compose appealing titles for lay readers by themselves.Key Point 1 (for Wording)Instead of technical terms, use synonymous plain terms even where the plain term is not synonymous with the technical term in a precise sense. Key Point 2 (for Content) Describe what the technology is for, rather than what the technology does. Key Point 3 (for Content) Describe the advantages of the technology, rather than the method of realizing the technology. | A Support System for Revising Titles to Stimulate the Lay Reader's Interest in Technical Achievements |
d247447330 | We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model was developed by continuing the pre-training phase of RobBERT [3] by using 2.8M Dutch COVID-19 related tweets posted in 2021. In order to test the performance of the model and compare it to RobBERT, the two models were tested on two tasks: (1) binary vaccine hesitancy detection and (2) detection of arguments for vaccine hesitancy. For both tasks, not only Twitter but also Facebook data was used to show cross-genre performance. In our experiments, CoNTACT showed statistically significant gains over RobBERT in all experiments for task 1. For task 2, we observed substantial improvements in virtually all classes in all experiments. An error analysis indicated that the domain adaptation yielded better representations of domain-specific terminology, causing CoNTACT to make more accurate classification decisions. | |
d233219869 | The problem of answering questions using knowledge from pre-trained language models (LMs) and knowledge graphs (KGs) presents two challenges: given a QA context (question and answer choice), methods need to (i) identify relevant knowledge from large KGs, and (ii) perform joint reasoning over the QA context and KG. Here we propose a new model, QA-GNN, which addresses the above challenges through two key innovations: (i) relevance scoring, where we use LMs to estimate the importance of KG nodes relative to the given QA context, and (ii) joint reasoning, where we connect the QA context and KG to form a joint graph, and mutually update their representations through graph-based message passing. We evaluate QA-GNN on the CommonsenseQA and OpenBookQA datasets, and show its improvement over existing LM and LM+KG models, as well as its capability to perform interpretable and structured reasoning, e.g., correctly handling negation in questions. | QA-GNN: Reasoning with Language Models and Knowledge Graphs for Question Answering |
d199480526 | As with many text generation tasks, the focus of recent progress on story generation has been in producing texts that are perceived to "make sense" as a whole. There are few automated metrics that address this dimension of story quality even on a shallow lexical level. To initiate investigation into such metrics, we apply a simple approach to identifying word relations that contribute to the 'narrative sense' of a story. We use this approach to comparatively analyze the output of a few notable story generation systems in terms of these relations. We characterize differences in the distributions of relations according to their strength within each story. | Identifying Sensible Lexical Relations in Generated Stories |
d6813475 | The SENSEVAL-3 task to perform word-sense disambiguation of WordNet glosses was designed to encourage development of technology to make use of standard lexical resources. The task was based on the availability of sensedisambiguated hand-tagged glosses created in the eXtended WordNet project. The hand-tagged glosses provided a "gold standard" for judging the performance of automated disambiguation systems. Seven teams participated in the task, with a total of 10 runs. Scoring these runs as an "all-words" task, along with considerable discussions among participants, provided more insights than just the underlying technology. The task identified several issues about the nature of the WordNet sense inventory and the underlying use of wordnet design principles, particularly the significance of WordNet-style relations. | SENSEVAL-3 TASK Word-Sense Disambiguation of WordNet Glosses |
d6249954 | Japanese has many noun phrase patterns of the type A no B consisting of two nouns A and B with an adnominal particle no. As the semantic relations between the two nouns in the noun phrase are not made explicit, the interpretation of the phrases depends mainly on the semantic characteristics of the nouns. This paper describes the semantic diversity of A no B and a method of semantic analysis for such phrases based on feature unification. | |
d11624810 | In this paper we address the problem of skewed class distribution in implicit discourse relation recognition. We examine the performance of classifiers for both binary classification predicting if a particular relation holds or not and for multi-class prediction. We review prior work to point out that the problem has been addressed differently for the binary and multi-class problems. We demonstrate that adopting a unified approach can significantly improve the performance of multi-class prediction. We also propose an approach that makes better use of the full annotations in the training set when downsampling is used. We report significant absolute improvements in performance in multi-class prediction, as well as significant improvement of binary classifiers for detecting the presence of implicit Temporal, Comparison and Contingency relations. | Addressing Class Imbalance for Improved Recognition of Implicit Discourse Relations |
d14555476 | In this paper, we experiment with a resource consisting of metaphorically annotated proverbs on the task of word-level metaphor recognition. We observe that existing feature sets do not perform well on this data. We design a novel set of features to better capture the peculiar nature of proverbs and we demonstrate that these new features are significantly more effective on the metaphorically dense proverb data. | Learning to Identify Metaphors from a Corpus of Proverbs |
d11238798 | This paper discusses the issue of suitability of software used for the teaching of Machine Translation. It considers requirements to such software, and describes a set of tools that have initially been created as developer environment of an APTrans MT system but can easily be included in the learning environment for MT training. The tools are user-friendly and feature modularity and reusability. | An MT Learning Environment for Computational Linguistics Students |
d252907395 | We make decisions by reacting to changes in the real world, particularly the emergence and disappearance of impermanent entities such as restaurants, services, and events. Because we want to avoid missing out on opportunities or making fruitless actions after those entities have disappeared, it is important to know when entities disappear as early as possible. We thus tackle the task of detecting disappearing enti- | Early Discovery of Disappearing Entities in Microblogs |
d252624549 | The spread of misinformation has become a major concern to our society, and social media is one of its main culprits. Evidently, health misinformation related to vaccinations has slowed down global efforts to fight the COVID-19 pandemic. Studies have shown that fake news spreads substantially faster than real news on social media networks. One way to limit this fast dissemination is by assessing information sources in a semi-automatic way. To this end, we aim to identify users who are prone to spread fake news in Arabic Twitter. Such users play an important role in spreading misinformation and identifying them has the potential to control the spread. We construct an Arabic dataset on Twitter users, which consists of 1,546 users, of which 541 are prone to spread fake news (based on our definition). We use features extracted from users' recent tweets, e.g., linguistic, statistical, and profile features, to predict whether they are prone to spread fake news or not. To tackle the classification task, multiple learning models are employed and evaluated. Empirical results reveal promising detection performance, where an F1 score of 0.73 was achieved by the logistic regression model. Moreover, when tested on a benchmark English dataset, our approach has outperformed the current state-of-the-art for this task. | Detecting Users Prone to Spread Fake News on Arabic Twitter |
d239020549 | Machine Translation of Canadian Court Decisions | |
d16303315 | The Arabic language is a collection of dialectal variants along with the standard form, Modern Standard Arabic (MSA). MSA is used in official Settings while the dialectal variants (DA) correspond to the native tongue of the Arabic speakers. Arabic speakers typically code switch between DA and MSA, which is reflected extensively in written online social media. Automatic processing such Arabic genre is very difficult for automated NLP tools since the linguistic difference between MSA and DA is quite profound. However, no annotated resources exist for marking the regions of such switches in the utterance. In this paper, we present a simplified Set of guidelines for detecting code switching in Arabic on the word/token level. We use these guidelines in annotating a corpus that is rich in DA with frequent code switching to MSA. We present both a quantitative and qualitative analysis of the annotations. | Simplified guidelines for the creation of Large Scale Dialectal Arabic Annotations |
d5743794 | Unambiguous Non-Terminally Separated (UNTS) grammars have properties that make them attractive for grammatical inference. However, these properties do not state the maximal performance they can achieve when they are evaluated against a gold treebank that is not produced by an UNTS grammar. In this paper we investigate such an upper bound. We develop a method to find an upper bound for the unlabeled F 1 performance that any UNTS grammar can achieve over a given treebank. Our strategy is to characterize all possible versions of the gold treebank that UNTS grammars can produce and to find the one that optimizes a metric we define. We show a way to translate this score into an upper bound for the F 1. In particular, we show that the F 1 parsing score of any UNTS grammar can not be beyond 82.2% when the gold treebank is the WSJ10 corpus. | Upper Bounds for Unsupervised Parsing with Unambiguous Non-Terminally Separated Grammars |
d238744065 | Prompting has recently been shown as a promising approach for applying pre-trained language models to perform downstream tasks. We present Multi-Stage Prompting, a simple and automatic approach for leveraging pre-trained language models to translation tasks. To better mitigate the discrepancy between pre-training and translation, MSP divides the translation process via pre-trained language models into multiple separate stages: the encoding stage, the re-encoding stage, and the decoding stage. During each stage, we independently apply different continuous prompts for allowing pretrained language models better shift to translation tasks. We conduct extensive experiments on three translation tasks. Experiments show that our method can significantly improve the translation performance of pre-trained language models. 1 | MSP: Multi-Stage Prompting for Making Pre-trained Language Models Better Translators |
d252118402 | We introduce AARGH, an end-to-end taskoriented dialog system combining retrieval and generative approaches in a single model, aiming at improving dialog management and lexical diversity of outputs. The model features a new response selection method based on an action-aware training objective and a simplified single-encoder retrieval architecture which allow us to build an end-to-end retrievalenhanced generation model where retrieval and generation share most of the parameters.On the MultiWOZ dataset, we show that our approach produces more diverse outputs while maintaining or improving state tracking and context-to-response generation performance, compared to state-of-the-art baselines. | AARGH! End-to-end Retrieval-Generation for Task-Oriented Dialog |
d9468172 | Hierarchical A * (HA * ) uses of a hierarchy of coarse grammars to speed up parsing without sacrificing optimality. HA * prioritizes search in refined grammars using Viterbi outside costs computed in coarser grammars. We present Bridge Hierarchical A * (BHA * ), a modified Hierarchial A * algorithm which computes a novel outside cost called a bridge outside cost. These bridge costs mix finer outside scores with coarser inside scores, and thus constitute tighter heuristics than entirely coarse scores. We show that BHA * substantially outperforms HA * when the hierarchy contains only very coarse grammars, while achieving comparable performance on more refined hierarchies. | Hierarchical A * Parsing with Bridge Outside Scores |
d22605062 | This paper presents a statistical summary of the use of the Transformational Question Answering (TQA) system by the City of White Plains Planning Department during the year 1978.A complete record of the 788 questions submitted to the system that year is included, as are separate listings of some of the problem inputs. Tables summarizing the performance of the system are also included and discussed. In general, performance of the system was sufficiently good that we believe that the approach being followed is a viable one, and are continuing to develop and extend the system. | Operating Statistics for The Transformational Question Answering System |
d16714688 | Natural Language Processing continues to grow in popularity in a range of research and commercial applications, yet managing the wide array of potential NLP components remains a difficult problem. This paper describes CURATOR, an NLP management framework designed to address some common problems and inefficiencies associated with building NLP process pipelines; and EDISON, an NLP data structure library in Java that provides streamlined interactions with CURATOR and offers a range of useful supporting functionality. | An NLP Curator (or: How I Learned to Stop Worrying and Love NLP Pipelines) |
d21745395 | Meaning Representation (AMR) annotations are often assumed to closely mirror dependency syntax, but AMR explicitly does not require this, and the assumption has never been tested. To test it, we devise an expressive framework to align AMR graphs to dependency graphs, which we use to annotate 200 AMRs. Our annotation explains how 97% of AMR edges are evoked by words or syntax. Previously existing AMR alignment frameworks did not allow for mapping AMR onto syntax, and as a consequence they explained at most 23%. While we find that there are indeed many cases where AMR annotations closely mirror syntax, there are also pervasive differences. We use our annotations to test a baseline AMR-to-syntax aligner, finding that this task is more difficult than AMRto-string alignment; and to pinpoint errors in an AMR parser. We make our data and code freely available for further research on AMR parsing and generation, and the relationship of AMR to syntax. | A Structured Syntax-Semantics Interface for English-AMR Alignment |
d259165390 | Reading comprehension is a crucial skill in many aspects of education, including language learning, cognitive development, and fostering early literacy skills in children. Automated answer-aware reading comprehension question generation has significant potential to scale up learner support in educational activities. One key technical challenge in this setting is that there can be multiple questions, sometimes very different from each other, with the same answer; a trained question generation method may not necessarily know which question human educators would prefer. To address this challenge, we propose 1) a data augmentation method that enriches the training dataset with diverse questions given the same context and answer and 2) an overgenerate-and-rank method to select the best question from a pool of candidates. We evaluate our method on the FairytaleQA dataset, showing a 5% absolute improvement in ROUGE-L over the best existing method. We also demonstrate the effectiveness of our method in generating harder, "implicit" questions, where the answers are not contained in the context as text spans. | Improving Reading Comprehension Question Generation with Data Augmentation and Overgenerate-and-rank |
d11829220 | As an initial effort to identify universal and language-specific factors that influence the behavior of distributional models, we have formulated a distributionally determined word similarity network model, implemented it for eleven different languages, and compared the resulting networks. In the model, vertices constitute words and two words are linked if they occur in similar contexts. The model is found to capture clear isomorphisms across languages in terms of syntactic and semantic classes, as well as functional categories of abstract discourse markers. Language specific morphology is found to be a dominating factor for the accuracy of the model. | Cross-lingual comparison between distributionally determined word similarity networks |
d39887248 | We present an ongoing project for the creation of a single central terminology database for all the institutions, agencies and other bodies of the European Union. The background, objectives, benefits and main features of the system are briefly introduced, followed by a presentation of the solutions proposed to resolve the complex validation issues addressed by a project which involves interaction between many institutions with different internal validation processes as well as access from the general public. | Validation and Quality Control Issues in a new Web-Based, Interactive Terminology Database for the Institutions and Agencies of the European Union Background to the IATE Project |
d53080574 | Sentiment analysis has immense implications in modern businesses through user-feedback mining. Large product-based enterprises like Samsung and Apple make crucial business decisions based on the large quantity of user reviews and suggestions available in different e-commerce websites and social media platforms like Amazon and Facebook. Sentiment analysis caters to these needs by summarizing user sentiment behind a particular object. In this paper, we present a novel approach of incorporating the neighboring aspects related information into the sentiment classification of the target aspect using memory networks. Our method outperforms the state of the art by 1.6% on average in two distinct domains. | IARM: Inter-Aspect Relation Modeling with Memory Networks in Aspect-Based Sentiment Analysis |
d226283996 | ||
d15748326 | The central concern of terminology, a component of the general documentation process, is concept analysis, an activity which is becoming recognized as fundamental as term banks evolve into knowledge bases. We propose that concept analysis can be facilitated by knowledge engineering technology, and describe a generic knowledge acquisition tool called CODE (Conceptually Oriented Design Environment) that has been successfully used in two terminology applications: 1) a bilingual vocabulary project with the Terminology Directorate of the Secretary of State of Canada, and 2) a software documentation project with Bell Northern Research. We conclude with some implications of computerassisted concept analysis for terminology. | CONCEPT ANALYSIS AND TERMINOLOGY: A KNOWLEDGE-BASED APPROACH TO DOCUMENTATION |
d226940747 | In recent years, statistical machine translation (SMT) has been widely deployed in translators' workflow with significant improvement of productivity. However, prior to invoking an SMT system to translate an unknown text, an SMT engine needs to be built. As such, building speed of the engine is essential for the translation workflow, i.e., the sooner an engine is built, the sooner it will be exploited.With the increase of the computational capabilities of recent technology the building time for an SMT engine has decreased substantially. For example, cloud-based SMT providers, such as KantanMT, can built high-quality, ready-to-use, custom SMT engines in less than a couple of days. To speed-up furthermore this process we look into optimizing the word alignment process that takes place during building the SMT engine. Namely, we substitute the word alignment tool used by KantanMT pipeline -Giza++ -with a more efficient one, i.e., fast_align.In this work we present the design and the implementation of the KantanMT pipeline that uses fast_align in place of Giza++. We also conduct a comparison between the two word alignment tools with industry data and report on our findings. Up to our knowledge, such extensive empirical evaluation of the two tools has not been done before. | Improving KantanMT Training Efficiency with fast align |
d237267360 | Unsupervised representation learning of words from large multilingual corpora is useful for downstream tasks such as word sense disambiguation, semantic text similarity, and information retrieval. The representation precision of log-bilinear fastText models is mostly due to their use of subword information. | One Size Does Not Fit All: Finding the Optimal Subword Sizes for FastText Models across Languages * |
d235254332 | MultiWOZ (Budzianowski et al., 2018)is one of the most popular multi-domain taskoriented dialog datasets, containing 10K+ annotated dialogs covering eight domains. It has been widely accepted as a benchmark for various dialog tasks, e.g., dialog state tracking (DST), natural language generation (NLG) and end-to-end (E2E) dialog modeling. In this work, we identify an overlooked issue with dialog state annotation inconsistencies in the dataset, where a slot type is tagged inconsistently across similar dialogs leading to confusion for DST modeling. We propose an automated correction for this issue, which is present in 70% of the dialogs. Additionally, we notice that there is significant entity bias in the dataset (e.g., "cambridge" appears in 50% of the destination cities in the train domain). The entity bias can potentially lead to named entity memorization in generative models, which may go unnoticed as the test set suffers from a similar entity bias as well. We release a new test set with all entities replaced with unseen entities. Finally, we benchmark joint goal accuracy (JGA) of the state-of-theart DST baselines on these modified versions of the data. Our experiments show that the annotation inconsistency corrections lead to 7-10% improvement in JGA. On the other hand, we observe a 29% drop in JGA when models are evaluated on the new test set with unseen entities. | Annotation Inconsistency and Entity Bias in MultiWOZ |
d616564 | OverviewMany research efforts have been devoted to develop robust statistical modeling techniques for many NLP tasks. Our field is now moving towards more complex tasks (e.g. RTE, QA), which require to complement these methods with a semantically rich representation based on world and linguistic knowledge (i.e. annotated linguistic data). In this tutorial we show several approaches to extract this knowledge from Wikipedia. This resource has attracted the attention of much work in the AI community, mainly because it provides semi-structured information and a large amount of manual annotations. The purpose of this tutorial is to introduce Wikipedia as a resource to the NLP community and to provide an introduction for NLP researchers both from a scientific and a practical (i.e. data acquisition and processing issues) perspective.OutlineThe tutorial is divided into three main parts:1. Extracting world knowledge from Wikipedia. We review methods aiming at extracting fully structured world knowledge from the content of the online encyclopedia. We show how to take categories, hyperlinks and infoboxes as building blocks for a semantic network with unlabeled relations between the concepts. The task of taxonomy induction then boils down to labeling the relations between these concepts, e.g. with isa, part-of, instance-of, located-in, etc. relations.Leveraging linguistic knowledge from Wikipedia.Wikipedia provides shallow markup annotations which can be interpreted as manual annotations of linguistic phenomena. These 'annotations' include word boundaries, word senses, named entities, translations of concepts in many languages. Furthermore, Wikipedia can be used as a multilingual comparable corpus.Future directions.Knowledge derived from Wikipedia has the potential to become a resource as important for NLP as WordNet. Also the Wikipedia edit history provides a repository of linguistic knowledge which is to be exploited. Potential applications of the knowledge implicitly encoded in the edit history include spelling corrections, natural language generation, text summarization, etc.Target audienceThis tutorial is designed for students and researchers in Computer Science and Computational Linguistics.No prior knowledge of information extraction topics is assumed. | Extracting World and Linguistic Knowledge from Wikipedia |
d3366048 | In this paper, we describe the results of sentiment analysis on tweets in three Indian languages -Bengali, Hindi, and Tamil. We used the recently released SAIL dataset(Patra et al., 2015), and obtained state-of-the-art results in all three languages. Our features are simple, robust, scalable, and language-independent. Further, we show that these simple features provide better results than more complex and language-specific features, in two separate classification tasks. Detailed feature analysis and error analysis have been reported, along with learning curves for Hindi and Bengali. | Sentiment Analysis of Tweets in Three Indian Languages |
d15098707 | As the amount of spoken communications accessible by computers increases, searching and browsing is becoming crucial for utilizing such material for gathering information. It is desirable for multimedia content analysis systems to handle various formats of data and to serve varying user needs while presenting a simple and consistent user interface. In this paper, we present a research system for searching and browsing spoken communications. The system uses core technologies such as speaker segmentation, automatic speech recognition, transcription alignment, keyword extraction and speech indexing and retrieval to make spoken communications easy to navigate. The main focus is on telephone conversations and teleconferences with comparisons to broadcast news. | A System for Searching and Browsing Spoken Communications |
d254685579 | Masked language modeling (MLM) has been widely used for pre-training effective bidirectional representations, but incurs substantial training costs. In this paper, we propose a novel concept-based curriculum masking (CCM) method to efficiently pre-train a language model. CCM has two key differences from existing curriculum learning approaches to effectively reflect the nature of MLM. First, we introduce a carefully-designed linguistic difficulty criterion that evaluates the MLM difficulty of each token. Second, we construct a curriculum that gradually masks words related to the previously masked words by retrieving a knowledge graph. Experimental results show that CCM significantly improves pre-training efficiency. Specifically, the model trained with CCM shows comparative performance with the original BERT on the General Language Understanding Evaluation benchmark at half of the training cost. Code is available at https://github.com/KoreaMGLEE/Conceptbased-curriculum-masking. | Efficient Pre-training of Masked Language Model via Concept-based Curriculum Masking |
d258959259 | One noticeable trend in metaphor detection is the embrace of linguistic theories such as the metaphor identification procedure (MIP) for model architecture design. While MIP clearly defines that the metaphoricity of a lexical unit is determined based on the contrast between its contextual meaning and its basic meaning, existing work does not strictly follow this principle, typically using the aggregated meaning to approximate the basic meaning of target words. In this paper, we propose a novel metaphor detection method, which models the basic meaning of the word based on literal annotation from the training set, and then compares this with the contextual meaning in a target sentence to identify metaphors. Empirical results show that our method outperforms the state-of-theart method significantly by 1.0% in F1 score. Moreover, our performance even reaches the theoretical upper bound on the VUA18 benchmark for targets with basic annotations, which demonstrates the importance of modelling basic meanings for metaphor detection. | Metaphor Detection via Explicit Basic Meanings Modelling |
d15545820 | Information about words--their pronunciation, syntax and meaning--is a crucial and costly part of human language technology. Many questions remain about the best way to express and use such lexical information. Nevertheless, much of this information is common to all current approaches, and therefore the effort to collect it can usefully be shared. The Linguistic Data Consortium (LDC) has undertaken to provide such common lexical information for the community of HLT researchers. The purpose of this paper is to sketch the various LDC lexical projects now underway or planned, and to solicit feedback from the community of HLT researchers. | Lexicons for Human Language Technology |
d1896571 | Morphological tokenization has been used in machine translation for morphologically complex languages to reduce lexical sparsity. Unfortunately, when translating into a morphologically complex language, recombining segmented tokens to generate original word forms is not a trivial task, due to morphological, phonological and orthographic adjustments that occur during tokenization. We review a number of detokenization schemes for Arabic, such as rule-based and table-based approaches and show their limitations. We then propose a novel detokenization scheme that uses a character-level discriminative string transducer to predict the original form of a segmented word. In a comparison to a stateof-the-art approach, we demonstrate slightly better detokenization error rates, without the need for any hand-crafted rules. We also demonstrate the effectiveness of our approach in an English-to-Arabic translation task. | Reversing Morphological Tokenization in English-to-Arabic SMT |
d925842 | Generative probabilistic models have been used for content modelling and template induction, and are typically trained on small corpora in the target domain. In contrast, vector space models of distributional semantics are trained on large corpora, but are typically applied to domaingeneral lexical disambiguation tasks. We introduce Distributional Semantic Hidden Markov Models, a novel variant of a hidden Markov model that integrates these two approaches by incorporating contextualized distributional semantic vectors into a generative model as observed emissions. Experiments in slot induction show that our approach yields improvements in learning coherent entity clusters in a domain. In a subsequent extrinsic evaluation, we show that these improvements are also reflected in multi-document summarization. | Probabilistic Domain Modelling With Contextualized Distributional Semantic Vectors |
d15309087 | This study points out that Yòu ( ) and Hái (✁) have their own prominent semantic features and syntactic patterns compared with each other. The differences reflect in the combination with verbs 1 . Hái ( ✁ ) has absolute superiority in collocation with V+Bu (✂)+V, which tends to express [durative]. Yòu ( ) has advantages in collocations with V+Le (✄)+V and derogatory verbs. Yòu ( )+V+Le (✄)+V tends to express [repetition], and Yòu ( )+derogatory verbs tends to express [repetition, derogatory]. We also find that the two words represent different semantic features when they match with grammatical aspect markers Le (✄ ), Zhe (☎ ) and Guo ( ✆ ). Different distributions have a close relation with their semantic features. This study is based on the investigation of the large-scale corpus and data statistics, applying methods of corpus linguistics, computational linguistics and semantic background model, etc. We also described and explained the language facts.PACLIC 29249 | A Corpus-based Comparatively Study on the Semantic Features and Syntactic patterns of Yòu/Hái in Mandarin Chinese |
d247451050 | In linguistics, a sememe is defined as the minimum semantic unit of languages. Sememe knowledge bases (KBs), which are built by manually annotating words with sememes, have been successfully applied to various NLP tasks. However, existing sememe KBs only cover a few languages, which hinders the wide utilization of sememes. To address this issue, the task of sememe prediction for Ba-belNet synsets (SPBS) is presented, aiming to build a multilingual sememe KB based on BabelNet, a multilingual encyclopedia dictionary. By automatically predicting sememes for a BabelNet synset, the words in many languages in the synset would obtain sememe annotations simultaneously. However, previous SPBS methods have not taken full advantage of the abundant information in BabelNet. In this paper, we utilize the multilingual synonyms, multilingual glosses and images in Ba-belNet for SPBS. We design a multimodal information fusion model to encode and combine this information for sememe prediction. Experimental results show the substantial outperformance of our model over previous methods (about 10 MAP and F1 scores). All the code and data of this paper can be obtained at https://github.com/thunlp/MSGI. | Sememe Prediction for BabelNet Synsets using Multilingual and Multimodal Information |
d235489736 | Neural models trained for next utterance generation in dialogue task learn to mimic the n-gram sequences in the training set with training objectives like negative log-likelihood (NLL) or cross-entropy. Such commonly used training objectives do not foster generating alternate responses to a context. But, the effects of minimizing an alternate training objective that fosters a model to generate alternate response and score it on semantic similarity has not been well studied. We hypothesize that a language generation model can improve on its diversity by learning to generate alternate text during training and minimizing a semantic loss as an auxiliary objective. We explore this idea on two different sized data sets on the task of next utterance generation in goal oriented dialogues. We make two observations (1) minimizing a semantic objective improved diversity in responses in the smaller data set (Frames) but only as-good-as minimizing the NLL in the larger data set (Mul-tiWoZ) (2) large language model embeddings can be more useful as a semantic loss objective than as initialization for token embeddings. | A Brief Study on the Effects of Training Generative Dialogue Models with a Semantic loss |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.