_id stringlengths 4 10 | text stringlengths 0 18.4k | title stringlengths 0 8.56k |
|---|---|---|
d9425719 | We report on the role of the Urdu grammar in the Parallel Grammar (ParGram) project(Butt et al., 1999;. 1 The ParGram project was designed to use a single grammar development platform and a unified methodology of grammar writing to develop large-scale grammars for typologically different languages. At the beginning of the project, three typologically similar European grammars were implemented. The addition of two Asian languages, Urdu and Japanese, has shown that the basic analysis decisions made for the European languages can be applied to typologically distinct languages. However, the Asian languages required the addition of a small number of new standard analyses to cover constructions and analysis techniques not found in the European languages. With these additional standards, the ParGram project can now be applied to other typologically distinct languages. | Urdu and the Parallel Grammar Project |
d14701624 | To resolve structural ambiguities in syntactic analysis of natural language, which are caused by prepositional phrase attachment, relative clause attachment, and so on, we developed an experimental system called tile Dependency Anal!lzcr. The system uses instances of dependency structures extracted froth a terminology dictionary as a knowledge ba.~e. Structural (attachment) ambiguity is represented by showing that a word has several words as c;tndidate modiliees. Tim system resolves such ambiguity as follows. First, it searches the knowledge base for modification relationships (dependencies) between the word and each of its possible modifiees, then assigns an order of preference to these relationships, and finally seieets the most preferable deper.dency. The knowledge base can be constructed semi-automatically, since the source of knowledge exists in the form of texts, and these sentences can be analyzed by the parser and transformed into dependency structures by the system. We are realizing knowledge bootstrapping by adding the outputs of the system to its knowledge base. | Dependency Analyzer: A Knowledge-Based Approach to Structural Disambiguation |
d8601453 | While the intuition that morphological preprocessing of languages in various applications can be beneficial appears to be often true, especially in the case of morphologically richer languages, it is not always the case. Previous work on translation between Nordic languages, including the morphologically rich Finnish, found that morphological analysis and preprocessing actually led to a decrease in translation quality below that of the unprocessed baseline.In this paper we investigate the proposition that the effect on translation quality depends on the kind of morphological preprocessing; and in particular that a specific kind of morphological preprocessing before translation could improve translation quality, a preprocessing that first transforms the source language to look more like the target, adapted from work on preprocessing via syntactically motivated reordering. We show that this is indeed the case in translating from Finnish, and that the results hold for different target languages and different morphological analysers. | Morphosyntactic Target Language Matching in Statistical Machine Translation |
d62651345 | Causal ambiguity in Natural Language: conceptual representation of ' parce que/ because' and ' puisque/ sincd | |
d252995552 | We present a comparative study of p(e)re- | Some Notes on p(e)re-Reduplication in Bulgarian and Ukrainian: A Corpus-based Study |
d38084426 | In France, Human Language Technologies (HLT) projects, though regularly supported by the French authorities, are usually limited in time and therefore do not offer long-term, continuous exploitation of their results. As an answer to these limitations and needs, the Technolangue programme, launched in 2002 by the three French funding bodies, namely the ministries of Research, Culture and Industry, aims at offering permanent infrastructures, allowing to capitalize on the results of industrial and academic R&D projects, both at national and international levels. The first section of this paper presents an overview of the Technolangue programme and some selected projects. In addition to its involvement in a number of LR projects, ELDA is coordinating two crucial projects in the Evaluation and Technology watch action lines of Technolangue, namely the organisation of an evaluation platform called EVALDA and the implementation of an HLT-oriented information portal known as Technolangue.net. As developed in this paper, the setting up of an evaluation infrastructure and the implementation of a HLT web portal both aim to fit one of the main objectives of the programme, i.e. the creation of permanent actions in the HLT field. | TECHNOLANGUE: A PERMANENT EVALUATION AND INFORMATION INFRASTRUCTURE |
d219308956 | ||
d238638482 | ||
d15248370 | Clinical documents have been an emerging target of natural language applications. Information stored in documents created at clinical settings can be very useful for doctors or medical experts. However, the way these documents are created and stored is often a hindrance to accessing their content. In this paper, an automatic method for restoring the intended structure of Hungarian ophthalmology documents is described. The statements in these documents in their original form appeared under various subheadings. We successfully applied our method for reassigning the correct heading for each line based on its content. The results show that the categorization was correct for 81.99% of the statements in our testset, compared to a human categorization. | Restoring the intended structure of Hungarian ophthalmology documents |
d8533468 | ||
d235599185 | ||
d219310356 | ||
d12181252 | This paper introduces an indexing method based on static analysis of grammar rules and type signatures for typed feature structure grammars (TFSGs). The static analysis tries to predict at compile-time which feature paths will cause unification failure during parsing at run-time. To support the static analysis, we introduce a new classification of the instances of variables used in TFSGs, based on what type of structure sharing they create. The indexing actions that can be performed during parsing are also enumerated. Non-statistical indexing has the advantage of not requiring training, and, as the evaluation using large-scale HPSGs demonstrates, the improvements are comparable with those of statistical optimizations. Such statistical optimizations rely on data collected during training, and their performance does not always compensate for the training costs. § | Optimizing Typed Feature Structure Grammar Parsing through Non-Statistical Indexing |
d234258405 | A top-down parser for Minimalist grammars (MGs; Stabler, 2013) can successfully predict a variety of off-line processing preferences, via metrics linking parsing behavior to memory load (Kobele et al., 2013;Gerth, 2015;Graf et al., 2017). The increasing empirical coverage of this model is intriguing, given its close association to modern minimalist syntax. Recently however, Zhang (2017) has argued that this framework is unable to account for a set of complexity profiles reported for English and Mandarin Chinese stacked relative clauses. Based on these observations, this paper proposes extensions to this model implementing a notion of memory reactivation, in the form of memory metrics sensitive to repetitions of movement features. We then show how these metrics derive the correct predictions for the stacked RC processing contrasts. | A Minimalist Approach to Facilitatory Effects in Stacked Relative Clauses |
d7541636 | Information about the animacy of nouns is important for a wide range of tasks in NLP. In this paper, we present a method for determining the animacy of English nouns using WordNet and machine learning techniques.Our method firstly categorises the senses from WordNet using an annotated corpus and then uses this information in order to classify nouns for which the sense is not known. Our evaluation results show that the accuracy of the classification of a noun is around 97% and that animate entities are more difficult to identify than inanimate ones. | Learning to identify animate references |
d241583193 | ||
d18414856 | In this paper, we propose a scheme for anaphora annotation in Hindi Dependency Treebank. The goal is to identify and handle the challenges that arise in the annotation of reference relations in Hindi. We identify some of the issues related to anaphora annotation specific to Hindi such as distribution of markable span, sequential annotation, representation format, annotation of multiple referents etc. The scheme hence incorporates some characteristics specific to these issues in order to achieve a consistent annotation. Most significant among these characteristics is the head-modifier separation in referent selection. The modifier-modified dependency relations inside a markable is utilized for this headmodifier distinction. A part of the Hindi Dependency Treebank, of around 2500 sentences has been annotated with anaphoric relations and an inter-annotator study was carried out which shows a significant agreement over selection of the head referent using the proposed scheme as compared to MUC annotation format. The current annotation is done for a limited set of pronominal categories. | Anaphora Annotation in Hindi Dependency TreeBank |
d1108658 | This study presents a method that assesses ESL learners' vocabulary usage to improve an automated scoring system of spontaneous speech responses by non-native English speakers. Focusing on vocabulary sophistication, we estimate the difficulty of each word in the vocabulary based on its frequency in a reference corpus and assess the mean difficulty level of the vocabulary usage across the responses (vocabulary profile).Three different classes of features were generated based on the words in a spoken response: coverage-related, average word rank and the average word frequency and the extent to which they influence human-assigned language proficiency scores was studied. Among these three types of features, the average word frequency showed the most predictive power. We then explored the impact of vocabulary profile features in an automated speech scoring context, with particular focus on the impact of two factors: genre of reference corpora and the characteristics of item-types.The contribution of the current study lies in the use of vocabulary profile as a measure of lexical sophistication for spoken language assessment, an aspect heretofore unexplored in the context of automated speech scoring. | The 7th Workshop on the Innovative Use of NLP for Building Educational Applications, pages 180-189, Vocabulary Profile as a Measure of Vocabulary Sophistication |
d21703708 | In this paper, we propose a neural network model for Part-Of-Speech (POS) tagging of User-Generated Content (UGC) such as Twitter, Facebook and Web forums. The proposed model is end-to-end and uses both character and word level representations. Character level representations are learned during the training of the model through a Convolutional Neural Network (CNN). For word level representations, we combine several pre-trainned embeddings (Word2Vec, FastText and GloVe). To deal with the issue of the poor availability of annotated social media data, we have implemented a Transfer Learning (TL) approach. We demonstrate the validity and genericity of our model on a POS tagging task by conducting our experiments on five social media languages (English, German, French, Italian and Spanish). | A Neural Network Model for Part-Of-Speech Tagging of Social Media Texts |
d52088192 | Neural machine translation (NMT) systems have demonstrated promising results in recent years. However, nontrivial amounts of manual effort are required for tuning network architectures, training configurations, and preprocessing settings such as byte pair encoding (BPE). In this study, we propose an evolution strategy based automatic tuning method for NMT. In particular, we apply the covariance matrix adaptation-evolution strategy (CMA-ES), and investigate a Pareto-based multi-objective CMA-ES to optimize the translation performance and computational time jointly. Experimental results show that the proposed method automatically finds NMT systems that outperform the initial manual setting. | Evolution Strategy Based Automatic Tuning of Neural Machine Translation Systems |
d1842024 | Data-driven (statistical) approaches have been playing an increasingly prominent role in parsing since the 1990s. In recent years, there has been a growing interest in dependency-based as opposed to constituency-based approaches to syntactic parsing, with application to a wide range of research areas and different languages. Graph-based and transition-based methods are the two dominant data-driven approaches to dependency parsing. In a graph-based model, it defines a space of candidate dependency trees for a given sentence. Each candidate tree is scored via a local or global scoring function. The parser (usually uses dynamic programming) outputs the highest-scored tree. In contrast, in a transition-based model, it defines a transition system for mapping a sentence to its dependency tree. It induces a model for predicting the next state transition, given the transition history. Given the induced model, the output parse tree is built deterministically upon the construction of the optimal transition sequence. | Recent Advances in Dependency Parsing |
d8842519 | Linguistic Knowledge Generator | |
d231695889 | ||
d219308599 | ||
d257767714 | Single document extractive text summarization produces a condensed version of a document by extracting salient sentences from the document. Most significant and diverse information can be obtained from a document by breaking it into topical clusters of sentences. The spectral clustering method is useful in text summarization because it does not assume any fixed shape of the clusters, and the number of clusters can automatically be inferred using the Eigen gap method. In our approach, we have used word embedding-based sentence representation and a spectral clustering algorithm to identify various topics covered in a Bengali document and generate an extractive summary by selecting salient sentences from the identified topics. We have compared our developed Bengali summarization system with several baseline extractive summarization systems. The experimental results show that the proposed approach performs better than some baseline Bengali summarization systems it is compared to. | Unsupervised Bengali Text Summarization Using Sentence Embedding and Spectral Clustering |
d52162266 | We introduce an adversarial method for producing high-recall explanations of neural text classifier decisions. Building on an existing architecture for extractive explanations via hard attention, we add an adversarial layer which scans the residual of the attention for remaining predictive signal. Motivated by the important domain of detecting personal attacks in social media comments, we additionally demonstrate the importance of manually setting a semantically appropriate "default" behavior for the model by explicitly manipulating its bias term. We develop a validation set of humanannotated personal attacks to evaluate the impact of these changes. | Extractive Adversarial Networks: High-Recall Explanations for Identifying Personal Attacks in Social Media Posts |
d17198322 | This paper presents the design and implementation details of an email synthesizer using two-stage stochastic natural language generation, where the first stage structures the emails according to sender style and topic structure, and the second stage synthesizes text content based on the particulars of an email structure element and the goals of a given communication for surface realization. The synthesized emails reflect sender style and the intent of communication, which can be further used as synthetic evidence for developing other applications. | Two-Stage Stochastic Email Synthesizer |
d18603314 | This paper describes the Linear A/Minoan digital corpus and the approaches we applied to develop it.We aim to set up a suitable study resource for Linear A and Minoan. | Minoan linguistic resources: The Linear A Digital Corpus |
d53643073 | This paper presents a rule-based system for disambiguating French locative verbs and their translation into Arabic. The disambiguation phase is based on the use of the French Verb dictionary of Dubois and Dubois Charlier (LVF) as a linguistic resource, from which a base of disambiguation rules is extracted. The extracted rules take the form of transducers which are subsequently applied to texts. The translation phase consists in translating the disambiguated locative verbs returned by the disambiguation phase. The translation takes into account the verb tense, as well as the inflected form of that verb. This phase is based on bilingual dictionaries that contain the different French locative verbs and their translation into Arabic. The experimentation and the evaluation are done using the linguistic platform NooJ, both a language resource development environment and a tool for automatic large corpora flow(Fehri, 2012).This work is licensed under a Creative Commons Attribution 4.0 International License. License details: http://creativecommons.org/licenses/by/4.0/ | A Rule-Based System for Disambiguating French Locative Verbs and Their Translation into Arabic |
d2747041 | Breaking away from traditional attempts at coreference resolution from discourseonly inputs, we try to do the same by constructing rich verb semantics from perceptual data, viz. a 2-D video. Using a bottom-up dynamic attention model and relative-motion-features between agents in the video, transitive verbs, their argument ordering etc. are learned through association with co-occurring adult commentary. This leads to learning of synonymous NP phrases as well as anaphora such as "it","each other" etc. This preliminary demonstration argues for a new approach to developmental NLP, with multi-modal semantics as the basis for computational language learning. | Discovering Coreference Using Image-Grounded Verb Models |
d16040431 | We present an on-going work on a software package that integrates discriminative machine learning with the open source WebAnnotator system of Tannier (2012). The WebAnnotator system allows users to annotate web pages within their browser with custom tag sets. Meanwhile, we integrate the WebAnnotator system with a machine learning package which enables automatic tagging of new web pages. We hope the software evolves into a useful information extraction tool for motivated hobbyists who have domain expertise on their task of interest but lack machine learning or programming knowledge. This paper presents the system architecture, including the WebAnnotator-based front-end and the machine learning component. The system is available under an open source license. | Creating Custom Taggers by Integrating Web Page Annotation and Machine Learning |
d10631637 | This work is concerned with the space of alignments searched by word alignment systems. We focus on situations where word re-ordering is limited by syntax. We present two new alignment spaces that limit an ITG according to a given dependency parse. We provide D-ITG grammars to search these spaces completely and without redundancy. We conduct a careful comparison of five alignment spaces, and show that limiting search with an ITG reduces error rate by 10%, while a D-ITG produces a 31% reduction. | A Comparison of Syntactically Motivated Word Alignment Spaces |
d2409730 | Discriminative training in query spelling correction is difficult due to the complex internal structures of the data. Recent work on query spelling correction suggests a two stage approach a noisy channel model that is used to retrieve a number of candidate corrections, followed by discriminatively trained ranker applied to these candidates. The ranker, however, suffers from the fact the low recall of the first, suboptimal, search stage. | A Discriminative Model for Query Spelling Correction with Latent Structural SVM |
d46602 | This paper presents a translation model that is based on tree sequence alignment, where a tree sequence refers to a single sequence of subtrees that covers a phrase. The model leverages on the strengths of both phrase-based and linguistically syntax-based method. It automatically learns aligned tree sequence pairs with mapping probabilities from word-aligned biparsed parallel texts. Compared with previous models, it not only captures non-syntactic phrases and discontinuous phrases with linguistically structured features, but also supports multi-level structure reordering of tree typology with larger span. This gives our model stronger expressive power than other reported models. Experimental results on the NIST MT-2005 Chinese-English translation task show that our method statistically significantly outperforms the baseline systems. I J I I J J T f T e J J T f T e I J J I I J J 1) | A Tree Sequence Alignment-based Tree-to-Tree Translation Model |
d7149590 | Open IE methods extract structured propositions from text. However, these propositions are neither consolidated nor generalized, and querying them may lead to insufficient or redundant information. This work suggests an approach to organize open IE propositions using entailment graphs. The entailment relation unifies equivalent propositions and induces a specific-to-general structure. We create a large dataset of gold-standard proposition entailment graphs, and provide a novel algorithm for automatically constructing them. Our analysis shows that predicate entailment is extremely context-sensitive, and that current lexical-semantic resources do not capture many of the lexical inferences induced by proposition entailment. | Focused Entailment Graphs for Open IE Propositions |
d60223 | In this paper, we propose a selective combination approach of pivot and direct statistical machine translation (SMT) models to improve translation quality. We work with Persian-Arabic SMT as a case study. We show positive results (from 0.4 to 3.1 BLEU on different direct training corpus sizes) in addition to a large reduction of pivot translation model size. | Selective Combination of Pivot and Direct Statistical Machine Translation Models |
d1237242 | It is suggested in this paper that two-level morphology theory (Kay, Koskenniemi) can be extended to include morphological tone. This extension treats phonological features as I/O tapes for Finite State Transducers in a parallel sequential incrementation (PSI) architecture; phonological processes (e.g. assimilation) are seen as variants of an elementary unification operation over feature tapes (linear unification phonology, LUP). The phenomena analysed are tone terracing with tone-spreading (horizontal assimilation), downstep, upstep, downdrift, upsweep in two West African languages, Tem (Togo) and Baule (C6te d'Ivoire). It is shown that an FST acccount leads to more insightful definitions of the basic phenomena than other approaches (e.g. phonological rules or metrical systems). | FINITE STATE PROCESSING OF TONE SYSTEMS |
d262376165 | State-of-the-art statistical approaches to the Coreference Resolution task rely on sophisticated modeling, but very few (10-20) simple features. In this paper we propose to extend the standard feature set substantially, incorporating more linguistic knowledge. To investigate the usability of linguistically motivated features, we evaluate our system for a variety of machine learners on the standard dataset (MUC-7) with the traditional learning set-up (Soon et al., 2001). | Coreference Resolution with and without Linguistic Knowledge |
d61580390 | In extant phrase-based statistical machine translation (SMT) systems, the translation model relies on word-to-word alignments, which serve as constraints for the subsequent heuristic extraction and scoring processes. Word alignments are usually inferred in a probabilistic framework; yet, only one single best alignment is retained, as if alignments were deterministically produced. In this paper, we explore ways to take into account the entire alignment matrix, where each alignment link is scored by its probability. By comparison with previous attempts, we use an exponential model to compute these probabilities, which enables us to achieve significant improvements on the NIST MT'09 Arabic-English translation task. | Discriminative Weighted Alignment Matrices For Statistical Machine Translation |
d259833830 | Although Question Answering (QA) have advanced to the human-level language skills in NLP tasks, there is still a problem: the QA model gets confused when there are similar sentences or paragraphs. Existing studies focus on enhancing the text understanding of the candidate answers to improve the overall performance of the QA models. However, since these methods focus on re-ranking queries or candidate answers, they fail to resolve the confusion when many generated answers are similar to the expected answer. To address these issues, we propose a novel contrastive learning framework called ContrastiveQA that alleviates the confusion problem in answer extraction. We propose a supervised method where we generate positive and negative samples from the candidate answers and the given answer, respectively. We thus introduce ContrastiveQA, which uses contrastive learning with sampling data to reduce incorrect answers. Experimental results on four QA benchmarks show the effectiveness of the proposed method. | Enhancing text comprehension for Question Answering with Contrastive Learning |
d233365003 | ||
d86987 | Even in a massive corpus such as the Web, a substantial fraction of extractions appear infrequently. This paper shows how to assess the correctness of sparse extractions by utilizing unsupervised language models. The REALM system, which combines HMMbased and n-gram-based language models, ranks candidate extractions by the likelihood that they are correct. Our experiments show that REALM reduces extraction error by 39%, on average, when compared with previous work.Because REALM pre-computes language models based on its corpus and does not require any hand-tagged seeds, it is far more scalable than approaches that learn models for each individual relation from handtagged data. Thus, REALM is ideally suited for open information extraction where the relations of interest are not specified in advance and their number is potentially vast. | Sparse Information Extraction: Unsupervised Language Models to the Rescue |
d7100691 | Predicting the success of literary works is a curious question among publishers and aspiring writers alike. We examine the quantitative connection, if any, between writing style and successful literature. Based on novels over several different genres, we probe the predictive power of statistical stylometry in discriminating successful literary works, and identify characteristic stylistic elements that are more prominent in successful writings. Our study reports for the first time that statistical stylometry can be surprisingly effective in discriminating highly successful literature from less successful counterpart, achieving accuracy up to 84%. Closer analyses lead to several new insights into characteristics of the writing style in successful literature, including findings that are contrary to the conventional wisdom with respect to good writing style and readability. | Success with Style: Using Writing Style to Predict the Success of Novels |
d7294125 | Today's event ordering research is heavily dependent on annotated corpora. Current corpora influence shared evaluations and drive algorithm development. Partly due to this dependence, most research focuses on partial orderings of a document's events. For instance, the TempEval competitions and the TimeBank only annotate small portions of the event graph, focusing on the most salient events or on specific types of event pairs (e.g., only events in the same sentence). Deeper temporal reasoners struggle with this sparsity because the entire temporal picture is not represented. This paper proposes a new annotation process with a mechanism to force annotators to label connected graphs. It generates 10 times more relations per document than the TimeBank, and our TimeBank-Dense corpus is larger than all current corpora. We hope this process and its dense corpus encourages research on new global models with deeper reasoning. | An Annotation Framework for Dense Event Ordering |
d17315163 | We introduce initial groundwork for estimating suicide risk and mental health in a deep learning framework. By modeling multiple conditions, the system learns to make predictions about suicide risk and mental health at a low false positive rate. Conditions are modeled as tasks in a multitask learning (MTL) framework, with gender prediction as an additional auxiliary task. We demonstrate the effectiveness of multi-task learning by comparison to a well-tuned single-task baseline with the same number of parameters. Our best MTL model predicts potential suicide attempt, as well as the presence of atypical mental health, with AUC > 0.8. We also find additional large improvements using multi-task learning on mental health tasks with limited training data. | Multi-Task Learning for Mental Health using Social Media Text |
d11278768 | Invited Talk Content Recognition in Dialogue | |
d17304416 | This paper presents latent semantic grammars for the unsupervised induction of English grammar. Latent semantic grammars were induced by applying singular value decomposition to n-gram by context-feature matrices. Parsing was used to evaluate performance. Experiments with context, projectivity, and prior distributions show the relative performance effects of these kinds of prior knowledge. Results show that prior distributions, projectivity, and part of speech information are not necessary to beat the right branching baseline. | TextGraphs-2: Graph-Based Algorithms for Natural Language Processing |
d209061963 | We present a scalable, open-source platform that "distills" a potentially large text collection into a knowledge graph. Our platform takes documents stored in Apache Solr and scales out the Stanford CoreNLP toolkit via Apache Spark integration to extract mentions and relations that are then ingested into the Neo4j graph database. The raw knowledge graph is then enriched with facts extracted from an external knowledge graph. The complete product can be manipulated by various applications using Neo4j's native Cypher query language: We present a subgraph-matching approach to align extracted relations with external facts and show that fact verification, locating textual support for asserted facts, detecting inconsistent and missing facts, and extracting distantly-supervised training data can all be performed within the same framework. | Knowledge Graph Construction from Unstructured Text with Applications to Fact Verification and Beyond |
d243865640 | Specialized number representations in NLP have shown improvements on numerical reasoning tasks like arithmetic word problems and masked number prediction. But humans also use numeracy to make better sense of world concepts, e.g., you can seat 5 people in your room but not 500. Does a better grasp of numbers improve a model's understanding of other concepts and words? This paper studies the effect of using six different number encoders on the task of masked word prediction (MWP), as a proxy for evaluating literacy. To support this investigation, we develop Wiki-Convert, a 900,000 sentence dataset annotated with numbers and units, to avoid conflating nominal and ordinal number occurrences. We find a significant improvement in MWP for sentences containing numbers, that exponent embeddings are the best number encoders, yielding over 2 points jump in prediction accuracy over a BERT baseline, and that these enhanced literacy skills also generalize to contexts without annotated numbers. We release all code at https://git.io/JuZXn. | Numeracy enhances the Literacy of Language Models |
d52036 | Publication bias refers to the phenomenon that statistically significant, "positive" results are more likely to be published than non-significant, "negative" results. Currently, researchers have to manually identify negative results in a large number of publications in order to examine publication biases. This paper proposes an NLP approach for automatically classifying negated sentences in biomedical abstracts as either reporting negative findings or not. Using multinomial naïve Bayes algorithm and bag-ofwords features enriched by parts-ofspeeches and constituents, we built a classifier that reached 84% accuracy based on 5-fold cross validation on a balanced data set. | Classifying Negative Findings in Biomedical Publications |
d19007354 | We present an elegant and extensible model that is capable of providing semantic interpretations for an unusually wide range of textual tables in documents. Unlike the few existing table analysis models, which largely rely on relatively ad hoc heuristics, our linguistically-oriented approach is systematic and grammar based, which allows our model (1) to be concise and yet (2) recognize a wider range of data models than others, and (3) disambiguate to a significantly finer extent the underlying semantic interpretation of the table in terms of data models drawn from relation database theory. To accomplish this, the model introduces Viterbi parsing under two-dimensional stochastic CFGs. The cleaner grammatical approach facilitates not only greater coverage, but also grammar extension and maintenance, as well as a more direct and declarative link to semantic interpretation, for which we also introduce a new, cleaner data model. In disambiguation experiments on recognizing relevant data models of unseen web tables from different domains, a blind evaluation of the model showed 60% precision and 80% recall. | A Grammatical Approach to Understanding Textual Tables using Two-Dimensional SCFGs |
d246015914 | Multi-hop knowledge graph (KG) reasoning has been widely studied in recent years to provide interpretable predictions on missing links with evidential paths. Most previous works use reinforcement learning (RL) based methods that learn to navigate the path towards the target entity. However, these methods suffer from slow and poor convergence, and they may fail to infer a certain path when there is a missing edge along the path. Here we present SQUIRE, the first Sequence-to-sequence based multi-hop reasoning framework, which utilizes an encoder-decoder Transformer structure to translate the query to a path. Our framework brings about two benefits: (1) It can learn and predict in an end-to-end fashion, which gives better and faster convergence; (2) Our transformer model does not rely on existing edges to generate the path, and has the flexibility to complete missing edges along the path, especially in sparse KGs. Experiments on standard and sparse KGs show that our approach yields significant improvement over prior methods, while converging 4x-7x faster. | SQUIRE: A Sequence-to-sequence Framework for Multi-hop Knowledge Graph Reasoning |
d18795816 | We report the results we obtained at the subtask B (Message Polarity Classification) of Se-mEval 2014 Task 9. The features used for representing the messages were basically trigrams of characters, trigrams of PoS and a number of words selected by means of a graph mining tool. Our approach performed slightly below the overall average, except when a corpus of tweets with sarcasm was evaluated, in which we performed quite well obtaining around 6% above the overall average. | BUAP: Polarity Classification of Short Texts |
d7830261 | We present a knowledge-based coreference resolution system for noun phrases in Hungarian texts. The system is used as a module in an automated psychological text processing project. Our system uses rules that rely on knowledge from the morphological, syntactic and semantic output of a deep parser and semantic relations form the Hungarian WordNet ontology. We also use rules that rely on Binding Theory, research results in Hungarian psycholinguistics, current research on proper name coreference identification and our own heuristics. We describe the constraints-and-preferences algorithm in detail that attempts to find coreference information for proper names, common nouns, pronouns and zero pronouns in texts. We present evaluation results for our system on a corpus manually annotated with coreference relations. Precision of the resolution of various coreference types reaches up to 80%, while overall recall is 63%. We also present an investigation of the various error types our system produced, along with an analysis of the results. | Knowledge-based Coreference Resolution for Hungarian |
d180318706 | Automated processing of clinical texts is commonly faced with various less exposed, and not so regularly discussed linguistically complex problems that need to be addressed. One of these issues concerns the usage of figurative language. Figurative language implies the use of words that go beyond their ordinary meaning, a linguistically complex and challenging problem and also a problem that causes great difficulty for the field of natural language processing (NLP). The problem is equally prevalent in both general language and also in various sublanguages, such as clinical medicine. Therefore we believe that a comprehensive model of e.g. clinical language processing needs to account for figurative language usage, and this paper provides a description, and preliminary results towards this goal. Since the empirical, clinical data used in the study is limited in size, there is no formal distinction made between different sub-classifications of figurative language. e.g., metaphors, idioms or simile. We illustrate several types of figurative expressions in the clinical discourse and apply a rather quantitative and corpus-based level analysis. The main research questions that this paper asks are whether there are traces of figurative language (or at least a subset of such types) in patient-doctor and patient-nurse interactions, how can they be found in a convenient way and whether these are transferred in the electronic health records and to what degree. | Figurative Language in Swedish Clinical Texts |
d41159378 | According toLi & Thompson (1981), Chinese serial verb constructions consisting of two verb phrases denoting two separate events can be classified into those having alternating, consecutive, purpose or circumstance relations. These classifications may overlap and a serial verb construction may be ambiguous between different interpretations. It has been argued in a recent study(Chan 1996)that there exists an entailment relation between the different interpretations of an ambiguous serial verb construction. In the present study, it is argued that because of the entailment relations between the different interpretations, the truth conditional definition of ambiguity has to be modified if it is to be applied to Chinese serial verb constructions which are ambiguous. | The Truth-Conditional Treatment of Ambiguity and Chinese Serial Verb Constructions |
d229365930 | ||
d14277325 | This paper describes a system for the unsupervised learning of morphological suffixes and stems from word lists. The system is composed of a generative probability model and a novel search algorithm. By examining morphologically rich subsets of an input lexicon, the search identifies highly productive paradigms. Quantitative results are shown by measuring the accuracy of the morphological relations identified. Experiments in English and Polish, as well as comparisons with other recent unsupervised morphology learning algorithms demonstrate the effectiveness of this technique. | Unsupervised Learning of Morphology Using a Novel Directed Search Algorithm: Taking the First Step |
d14131495 | This paper describes an unsupervised method for extracting product attributes and their values from an e-commerce product page. Previously, distant supervision has been applied for this task, but it is not applicable in domains where no reliable knowledge base (KB) is available. Instead, the proposed method automatically creates a KB from tables and itemizations embedded in the product's pages. This KB is applied to annotate the pages automatically and the annotated corpus is used to train a model for the extraction. Because of the incompleteness of the KB, the annotated corpus is not as accurate as a manually annotated one. Our method tries to filter out sentences that are likely to include problematic annotations based on statistical measures and morpheme patterns induced from the entries in the KB. The experimental results show that the performance of our method achieves an average F score of approximately 58.2 points and that filters can improve the performance. | Unsupervised Extraction of Attributes and Their Values from Product Description |
d14176162 | In this paper, we introduce SLQS, a new entropy-based measure for the unsupervised identification of hypernymy and its directionality in Distributional Semantic Models (DSMs). SLQS is assessed through two tasks: (i.) identifying the hypernym in hyponym-hypernym pairs, and (ii.) discriminating hypernymy among various semantic relations. In both tasks, SLQS outperforms other state-of-the-art measures. | Chasing Hypernyms in Vector Spaces with Entropy |
d13978200 | We present a domain-restricted rule based machine translation system based on dependency parsing. We replace the transfer phase of the classical analysis, transfer, and generation strategy with a syntax planning algorithm that directly linearizes the dependency parse of the source sentence as per the syntax of the target language. While we have built the system for English to Hindi translation, the approach can be generalized to other source languages too where a dependency parser is available. | A Domain-Restricted, Rule Based, English-Hindi Machine Translation System Based on Dependency Parsing |
d29767064 | This study explores the role of speech register and prosody for the task of word segmentation. Since these two factors are thought to play an important role in early language acquisition, we aim to quantify their contribution for this task. We study a Japanese corpus containing both infant-and adult-directed speech and we apply four different word segmentation models, with and without knowledge of prosodic boundaries. The results showed that the difference between registers is smaller than previously reported and that prosodic boundary information helps more adult-than infant-directed speech. | The Role of Prosody and Speech Register in Word Segmentation: A Computational Modelling Perspective |
d11547363 | This paper studies the derivation of multiple nominative constructions (MNC) in Japanese. First, discussing the MNC-sentences in which there is a relation of inalienable possession between nominative noun phrases, I will argue that the set of local economy principles that choose among potentially possible steps at a single stage of a derivation contains a principle that minimizes the size of moved elements. Second, considering the derivation of the MNC-sentences in which there is no relation of inalienable possession between nominative noun phrases, I will show a new piece of evidence for the Merge-over-Move principle. | Multiple Nominative Constructions in Japanese and Their Theoretical Implications |
d259370838 | Discovering new intents is of great significance for establishing the Task-Oriented Dialogue | A Probabilistic Framework for Discovering New Intents |
d15890923 | In this paper a social network is extracted from a literary text. The social network shows, how frequent the characters interact and how similar their social behavior is. Two types of similarity measures are used: the first applies co-occurrence statistics, while the second exploits cosine similarity on different types of word embedding vectors. The results are evaluated by a paid micro-task crowdsourcing survey. The experiments suggest that specific types of word embeddings like word2vec are well-suited for the task at hand and the specific circumstances of literary fiction text. | Extracting Social Networks from Literary Text with Word Embedding Tools |
d251404008 | This paper presents the NDC Treebank of spoken Norwegian dialects in the Bokmål variety of Norwegian. It consists of dialect recordings made between 2006 and 2012 which have been digitised, segmented, transcribed and subsequently annotated with morphological and syntactic analysis. The nature of the spoken data gives rise to various challenges both in segmentation and annotation. We follow earlier efforts for Norwegian, in particular the LIA Treebank of spoken dialects transcribed in the Nynorsk variety of Norwegian, in the annotation principles to ensure interusability of the resources. We have developed a spoken language parser on the basis of the annotated material and report on its accuracy both on a test set across the dialects and by holding out single dialects. | The Norwegian Dialect Corpus Treebank |
d14444609 | Analogies are considered to be one of the core concepts of human cognition and communication, and are very efficient at encoding complex information in a natural fashion. However, computational approaches towards largescale analysis of the semantics of analogies are hampered by the lack of suitable corpora with real-life example of analogies. In this paper we therefore propose a workflow for discriminating and extracting natural-language analogy statements from the Web, focusing on analogies between locations mined from travel reports, blogs, and the Social Web. For realizing this goal, we employ feature-rich supervised learning models which we extensively evaluate. We also showcase a crowd-supported workflow for building a suitable Gold dataset used for this purpose. The resulting system is able to successfully learn to identify analogies to a high degree of accuracy (F-Score 0.9) by using a high-dimensional subsequence feature space. | Discriminating Rhetorical Analogies in Social Media |
d44158766 | Sentence compression methods based on LSTM can generate fluent compressed sentences. However, the performance of these methods is significantly degraded when compressing long sentences since it does not explicitly handle syntactic features. To solve this problem, we propose a higher-order syntactic attention network (HiSAN) that can handle higher-order dependency features as an attention distribution on LSTM hidden states. Furthermore, to avoid the influence of incorrect parse results, we train HiSAN by maximizing the probability of a correct output together with the attention distribution. Experiments on the Google sentence compression dataset show that our method achieved the best performance in terms of F 1 as well as ROUGE-1,2 and L scores, 83.2, 82.9, 75.8 and 82.7, respectively. In subjective evaluations, HiSAN outperformed baseline methods in both readability and informativeness. | Higher-order Syntactic Attention Network for Long Sentence Compression |
d219308529 | We present the Le Petit Prince Corpus (LPPC), a multi-lingual resource for research in (computational) psycho-and neurolinguistics. The corpus consists of the children's story The Little Prince in 26 languages. The dataset is in the process of being built using stateof-the-art methods for speech and language processing and electroencephalography (EEG). The planned release of LPPC dataset will include raw text annotated with dependency graphs in the Universal Dependencies standard, a near-natural-sounding synthetic spoken subset as well as EEG recordings. We will use this corpus for conducting neurolinguistic studies that generalize across a wide range of languages, overcoming typological constraints to traditional approaches. The planned release of the LPPC combines linguistic and EEG data for many languages using fully automatic methods, and thus constitutes a readily extendable resource that supports cross-linguistic and cross-disciplinary research. | c European Language Resources Association (ELRA), licensed under |
d12722998 | Relations among phenomena at different linguistic levels are at the essence of language properties but today we focus mostly on one specific linguistic layer at a time, without (having the possibility of) paying attention to the relations among the different layers. At the same time our efforts are too much scattered without much possibility of exploiting other people's achievements. To address the complexities hidden in multilayer interrelations even small amounts of processed data can be useful, improving the performance of complex systems. Exploiting the current trend towards sharing we want to initiate a collective movement that works towards creating synergies and harmonisation among different annotation efforts that are now dispersed. In this paper we present the general architecture of the Language Library, an initiative which is conceived as a facility for gathering and making available through simple functionalities the linguistic knowledge the field is able to produce, putting in place new ways of collaboration within the LRT community. In order to reach this goal, a first population round of the Language Library has started around a core of parallel/comparable texts that have been annotated by several contributors submitting a paper for LREC2012. The Language Library has also an ancillary aim related to language documentation and archiving and it is conceived as a theory-neutral space which allows for several language processing philosophies to coexist. | The Language Library: supporting community effort for collective resource production |
d9063641 | Recent years have witnessed a growing interest in analogical learning for NLP applications. If the principle of analogical learning is quite simple, it does involve complex steps that seriously limit its applicability, the most computationally demanding one being the identification of analogies in the input space. In this study, we investigate different strategies for efficiently solving this problem and study their scalability. | Coling 2008: Companion volume -Posters and Demonstrations |
d21708005 | The correct analysis of the output of a program based on supervised learning is inevitable in order to be able to identify the errors it produced and characterise its error types. This task is fairly difficult without a proper tool, especially if one works with complex data structures such as parse trees or sentence alignments. In this paper, we present a library that allows the user to interactively visualise and compare the output of any program that yields a well-known data format. Our goal is to create a tool granting the total control of the visualisation to the user, including extensions, but also have the common primitives and data-formats implemented for typical cases. We describe the common features of the common NLP tasks from the viewpoint of visualisation in order to specify the essential primitive functions. We enumerate many popular off-the-shelf NLP visualisation programs to compare with our implementation, which unifies all of the profitable features of the existing programs adding extendibility as a crucial feature to them. | What's Wrong, Python? -A Visual Differ and Graph Library for NLP in Python |
d3113004 | We describe the modification of a grammar to take advantage of prosodic information provided by a speech recognition system. This initial study is limited to the use of relative duration of phonetic segments in the assignment of syntactic structure, specifically in ruling out alternative parses in otherwise ambiguous sentences. Taking advantage of prosodic information in parsing can make a spoken language system more accurate and more efficient, if prosodicsyntactic mismatches, or unlikely matches, can be pruned. We know of no other work that has succeeded in automatically extracting speech information and using it in a parser to rule out extraneous parses. | PROSODY, SYNTAX AND PARSING |
d228076874 | ||
d9889920 | In this paper, we present a speech recording interface developed in the context of a project on automatic speech recognition for elderly native speakers of European Portuguese. In order to collect spontaneous speech in a situation of interaction with a machine, this interface was designed as a Wizard-of-Oz (WOZ) plateform. In this setup, users interact with a fake automated dialog system controled by a human wizard. It was implemented as a client-server application and the subjects interact with a talking head. The human wizard chooses pre-defined questions or sentences in a graphical user interface, which are then synthesized and spoken aloud by the avatar on the client side. A small spontaneous speech corpus was collected in a daily center. Eight speakers between 75 and 90 years old were recorded. They appreciated the interface and felt at ease with the avatar. Manual orthographic transcriptions were created for the total of about 45 minutes of speech. | El-WOZ: a client-server wizard-of-oz interface |
d250391076 | In recent years, NLP has advanced greatly along with the proliferation of pre-trained language models. The pre-trained language models are also properly adapted to downstream tasks when there is sufficient labeled data. However, in real-world applications, we often encounter the deficiency of labeled data. When only given a few instances for a new task, extracting task-aware features from a pre-trained language model regardless of the adaptation is a promising alternative. In the study, we propose a novel embedding transfer method, called LEA, for leveraging pre-trained language models with even only few-shot instances. LEA derives meta-level attention aspects using our new meta-learning framework. We evaluate our method on five text classification benchmark datasets. The results show that the novel method robustly provides the competitive performance compared to recent few-shot learning methods. | LEA: Meta Knowledge-Driven Self-Attentive Document Embedding for Few-Shot Text Classification |
d14064659 | Nous abordons la question du transfert d'annotations sémantiques, et plus spécifiquement d'étiquettes sur les prédicats, d'une langue à l'autre sur la base de corpus parallèles. Des travaux antérieurs ont transféré ces annotations directement au niveau des tokens, conduisant à un faible rappel. Nous présentons une approche globale de transfert qui agrège des informations repérées dans l'ensemble du corpus parallèle. Nous montrons que la performance de la méthode globale est supérieure aux résultats antérieurs en termes de rappel sans trop affecter la précision.Abstract. We address the problem of transferring semantic annotations, more specifically predicate labellings, from one language to another using parallel corpora. Previous work has transferred these annotations directly at the token level, leading to low recall. We present a global approach to annotation transfer that aggregates information across the whole parallel corpus. We show that this global method outperforms previous results in terms of recall without sacrificing precision too much.Mots-clés : transfert inter-langue, annotation sémantique automatique, prédicats, désambiguïsation lexicale, corpus parallèles. | 21 ème Traitement Automatique des Langues Naturelles |
d220047814 | Existing end-to-end dialog systems perform less effectively when data is scarce. To obtain an acceptable success in real-life online services with only a handful of training examples, both fast adaptability and reliable performance are highly desirable for dialog systems. In this paper, we propose the Meta-Dialog System (MDS), which combines the advantages of both meta-learning approaches and human-machine collaboration. We evaluate our methods on a new extended-bAbI dataset and a transformed MultiWOZ dataset for lowresource goal-oriented dialog learning. Experimental results show that MDS significantly outperforms non-meta-learning baselines and can achieve more than 90% per-turn accuracies with only 10 dialogs on the extended-bAbI dataset. | Learning Low-Resource End-To-End Goal-Oriented Dialog for Fast and Reliable System Deployment |
d13106921 | We present an approach which uses the similarity in semantic structure of bilingual parallel sentences to bootstrap a pair of semantic role labeling (SRL) models. The setting is similar to co-training, except for the intermediate model required to convert the SRL structure between the two annotation schemes used for different languages. Our approach can facilitate the construction of SRL models for resource-poor languages, while preserving the annotation schemes designed for the target language and making use of the limited resources available for it. We evaluate the model on four language pairs, English vs German, Spanish, Czech and Chinese. Consistent improvements are observed over the self-training baseline. | Bootstrapping Semantic Role Labelers from Parallel Data |
d6320666 | This workshopoutlines the progress that has been made on the TAUS Dynamic Quality Framework (DQF) in the past year and introduces the TAUS Quality Dashboard where all stakeholders in the global translation services can monitor their performance using industry-shared metrics and benchmark themselves against industry average productivity and quality. The TAUS DQF integration with translation tools via an open API will also be demonstrated.61Address format | The TAUS Quality Dashboard |
d17871858 | Named entity recognition (NER) systems trained on newswire perform very badly when tested on Twitter. Signals that were reliable in copy-edited text disappear almost entirely in Twitter's informal chatter, requiring the construction of specialized models. Using wellunderstood techniques, we set out to improve Twitter NER performance when given a small set of annotated training tweets. To leverage unlabeled tweets, we build Brown clusters and word vectors, enabling generalizations across distributionally similar words. To leverage annotated newswire data, we employ an importance weighting scheme. Taken all together, we establish a new state-of-the-art on two common test sets. Though it is wellknown that word representations are useful for NER, supporting experiments have thus far focused on newswire data. We emphasize the effectiveness of representations on Twitter NER, and demonstrate that their inclusion can improve performance by up to 20 F1. | The Unreasonable Effectiveness of Word Representations for Twitter Named Entity Recognition |
d5450664 | This paper presents our work on Semantic Role Labeling using a Transformation-Based Error-Driven approach in the style of Eric Brill(Brill, 1995). Our approach achieved an overall F 1 score of 43.48 on non-verb annotations. We believe our approach is noteworthy because of its novelty in this area and because it produces short lists of human-understandable transformation rules as its output. | Learning Transformation Rules for Semantic Role Labeling |
d226283536 | ||
d3203884 | We present an authoring system for logical forms encoded as conceptual graphs (CG). The system belongs to the family of WYSIWYM (What You See Is What You Mean) text generation systems: logical forms are entered interactively and the corresponding linguistic realization of the expressions is generated in several languages. The system maintains a model of the discourse context corresponding to the authored documents. The system helps users author documents formulated in the CG format. In a first stage, a domainspecific ontology is acquired by learning from example texts in the domain. The ontology acquisition module builds a typed hierarchy of concepts and relations derived from the WordNet and Verbnet. The user can then edit a specific document, by entering utterances in sequence, and maintaining a representation of the context. While the user enters data, the system performs the standard steps of text generation on the basis of the authored logical forms: reference planning, aggregation, lexical choice and syntactic realization -in several languages (we have implemented English and Hebrew -and are exploring an implementation using the Bliss graphical language). The feedback in natural language is produced in real-time for every single modification performed by the author. We perform a cost-benefit analysis of the application of NLG techniques in the context of authoring cooking recipes in English and Hebrew. By combining existing large-scale knowledge resources (WordNet, Verbnet, the SURGE and HUGG realization grammars) and techniques from modern integrated software development environment (such as the Eclipse IDE), we obtain an efficient tool for the generation of logical forms, in domains where content is not available in the form of databases. | Interactive Authoring of Logical Forms for Multilingual Generation * |
d917915 | We propose a novel probabilistic synchoronous context-free grammar formalism for statistical machine translation, in which syntactic nonterminal labels are represented as "soft" preferences rather than as "hard" matching constraints. This formalism allows us to efficiently score unlabeled synchronous derivations without forgoing traditional syntactic constraints. Using this score as a feature in a log-linear model, we are able to approximate the selection of the most likely unlabeled derivation. This helps reduce fragmentation of probability across differently labeled derivations of the same translation. It also allows the importance of syntactic preferences to be learned alongside other features (e.g., the language model) and for particular labeling procedures. We show improvements in translation quality on small and medium sized Chinese-to-English translation tasks. | Preference Grammars: Softening Syntactic Constraints to Improve Statistical Machine Translation |
d34363067 | Does sharing moral values encourage people to connect and form communities? The importance of moral homophily (love of same) has been recognized by social scientists, but the types of moral similarities that drive this phenomenon are still unknown. In this talk, I will present a series of experiments (both large-scale, observational social-media analyses and behavioral lab experiments) that investigate which types of moral similarities influence tie formations. Our results indicate that social network processes reflect moral selection, and both online and offline differences in moral purity concerns are particularly predictive of social distance.Dr. Morteza Dehghani is an Assistant Professor of psychology, computer science and the Brain and Creativity Institute at University of Southern California. His research spans the boundary between psychology and artificial intelligence, as does his education. His work investigates properties of cognition by using documents of the social discourse, such as narratives, social media, transcriptions of speeches and news articles, in conjunction to behavioral studies.16 | Purity Homophily in Social Networks -Invited Talk |
d22901863 | Au cours des deux dernières décennies, de nombreux algorithmes ont été développés pour capturer la sémantique des mots simples en regardant leur répartition dans un grand corpus, et en comparant ces distributions dans un modèle d'espace vectoriel. En revanche, il n'est pas trivial de combiner les objets algébriques de la sémantique distributionnelle pour arriver à une dérivation d'un contenu pour des expressions complexes, composées de plusieurs mots. Notre contribution a deux buts. Le premier est d'établir une large base de comparaison pour les méthodes de composition pour le cas adjectif_nom. Cette base nous permet d'évaluer en profondeur la performance des différentes méthodes de composition. Notre second but est la proposition d'une nouvelle méthode de composition, qui est une généralisation de la méthode deBaroni & Zamparelli (2010). La performance de notre nouvelle méthode est également évaluée sur notre nouveau ensemble de test.Abstract. In the course of the last two decades, numerous algorithms have sprouted up that successfully capture the semantics of single words by looking at their distribution in text, and comparing these distributions in a vector space model. However, it is not straightforward to construct meaning representations beyond the level of individual words -i.e. the combination of words into larger units -using distributional methods. Our contribution is twofold. First of all, we carry out a large scale evaluation, comparing different composition methods within the distributional framework for the case of adjective-noun composition, making use of a newly developed dataset. Secondly, we propose a novel method for adjective-noun composition, which is a generalization of the approach byBaroni & Zamparelli (2010). The performance of our novel method is equally evaluated on our new dataset.Mots-clés : sémantique lexicale, sémantique distributionnelle, compositionalité. | Une évaluation approfondie de différentes méthodes de compositionalité sémantique |
d237366368 | The neural network model has shown good results in relation extraction tasks in recent years. However, we know very little about the process of feature capture, which limits the further development of deep neural network models in relation extraction tasks. Current research work has explored the linguistic features of English relation extraction, and some rules have been obtained. However, due to the obvious differences between Chinese and Western languages, the laws and explanatory nature explored are not suitable for Chinese relationship extraction. This paper explores the Chinese relation extraction neural network for the first time, using a total of 13 types of exploration tasks from four perspectives, including Chinese-specific word segmentation exploration tasks. Experiments were carried out on two relation extraction data sets to explore the law of feature extraction in Chinese relation extraction model. | A Probe into the Sentence-level Linguistic Features of Chinese Relation Extraction |
d21677633 | To attract many foreign students is among the goals of the Karlsruhe Institute of Technology (KIT). One obstacle to achieving this goal is the fact that lectures at KIT are usually held in German which many foreign students are not sufficiently proficient in, as, e.g., opposed to English. While the students from abroad are learning German during their stay at KIT, it is very challenging to become proficient enough in it in order to follow such a complex communication situation as a lecture. As a solution to this problem we offer our automatic simultaneous lecture translation at KIT's lecture halls which automatically translates the German lectures into English in real time for the students. While not as good as human interpreters, the system is available at a price that KIT can afford in order to offer it in potentially all lectures. In order to assess whether the quality of the system is high enough in order to be of use to our foreign students at KIT we have conducted a user study on the benefit of the system to its users over the course of two terms. In this paper we present this study, the way it was conducted and its results. As it turns out the results indicate that the quality of the system has passed a threshold as to be able to support students in their studies. The detailed feedback participants to the study have helped to identify the most crucial weaknesses of the systems and has guided which development steps to take next. | Evaluation of the KIT Lecture Translation System |
d232021890 | ||
d250390593 | Style transfer is the task of transferring a sentence into the target style while keeping its content. The major challenge is that parallel corpora are not available for various domains. In this paper, we propose a Mask-And-Regenerate approach (MAR). It learns from unpaired sentences by modifying the word-level style attributes. We cautiously integrate the deletion, insertion and substitution operations into our model. This enables our model to automatically apply different edit operations for different sentences. Specifically, we train a multilayer perceptron (MLP) as a style classifier to find out and mask style-characteristic words in the source inputs. Then we learn a language model on non-parallel data sets to score sentences and remove unnecessary masks. Finally, the masked source sentences are input to a Transformer to perform style transfer. The final results show that our proposed model exceeds baselines by about 2 per cent of accuracy for both sentiment and style transfer tasks with comparable or better content retention. | Mask and Regenerate: A Classifier-based Approach for Unpaired Sentiment Transformation of Reviews for Electronic Commerce Websites |
d2432354 | Although research in other languages is increasing, much of the work in subjectivity analysis has been applied to English data, mainly due to the large body of electronic resources and tools that are available for this language. In this paper, we propose and evaluate methods that can be employed to transfer a repository of subjectivity resources across languages. Specifically, we attempt to leverage on the resources available for English and, by employing machine translation, generate resources for subjectivity analysis in other languages. Through comparative evaluations on two different languages (Romanian and Spanish), we show that automatic translation is a viable alternative for the construction of resources and tools for subjectivity analysis in a new target language. | Multilingual Subjectivity Analysis Using Machine Translation |
d9403299 | The problem of style is highly relevant to computational linguistics, but current systems deal only superficially, if at all, with subtle but significant nuances of language. Expressive effects, together with their associated meaning, contained in the style of a text are lost to analysis and absent from generation.We have developed an approach to the computational treatment of style that is intended to eventually incorporate three selected components--lexical, syntactic, and semantic. In this paper, we concentrate on certain aspects of syntactic style. We have designed and implemented a computational theory of goal-directed stylistics that can be used in various applications, including machine translation, second-language instruction, and natural language generation.We have constructed a vocabulary of style that contains both primitive and abstract elements of style. The primitive elements describe the stylistic effects of individual sentence components. These elements are combined into patterns that are described by a stylistic meta-language, the abstract elements, that define the concordant and discordant stylistic effects common to a group of sentences. Higher-level patterns are built from the abstract elements and associated with specific stylistic goals, such as clarity or concreteness. Thus, we have defined rules for a syntactic stylistic grammar at three interrelated levels of description: primitive elements, abstract elements, and stylistic goals. Grammars for both English and French have been constructed, using the same vocabulary and the same development methodology. Parsers that implement these grammars have also been built.The stylistic grammars codify aspects of language that were previously defined only descriptively. The theory is being applied to various problems in which the form of an utterance conveys an essential part of meaning and so must be precisely represented and understood. | A Computational Theory of Goal-Directed Style in Syntax |
d26412 | Document indexing and representation of term-document relations are very important for document clustering and retrieval. In this paper, we combine a graph-based dimensionality reduction method with a corpus-based association measure within the Generalized Latent Semantic Analysis framework. We evaluate the graph-based GLSA on the document clustering task. | Graph-based Generalized Latent Semantic Analysis for Document Representation |
d252212941 | In this paper, we present an approach toward grounding linguistic positional and directional labels directly to human motions in a disoriented balancing task in a multi-axis rotational device. We use deep neural models to predict human subjects' joystick motions and proficiency in the task. We combine these with BERT embeddings for annotated positional and directional labels into an embodied direction classifier. Combining contextualized BERT embeddings with embeddings representing human motion and proficiency can successfully predict the direction a hypothetical human participant should move to achieve better balance. Our accuracy is comparable to a moderatelyproficient human subject, and we find that our combined embodied model may actually make objectively better decisions than some humans. | Where am I and where should I go? Grounding positional and directional labels in a disoriented human balancing task |
d2691618 | We present a neural network method for inducing representations of parse histories and using these history representations to estimate the probabilities needed by a statistical left-corner parser. The resulting statistical parser achieves performance (89.1% F-measure) on the Penn Treebank which is only 0.6% below the best current parser for this task, despite using a smaller vocabulary size and less prior linguistic knowledge. Crucial to this success is the use of structurally determined soft biases in inducing the representation of the parse history, and no use of hard independence assumptions.Estimating the Parameters of the Probability ModelThe parsing system we propose consists of two components, one which estimates the parameters of a probability model for phrase structure trees, and one which searches for the most probable phrase structure tree given these parameters. The probability model we use is generative and history-based. At each step, the model's stochastic process generates a characteristic of the tree Edmonton, | Inducing History Representations for Broad Coverage Statistical Parsing |
d16533276 | The work of the research project "Variance of Njáls saga" at the Árni Magnússon Institute for Icelandic Studies in Reykjavík relies mainly on an annotated XML-corpus of manuscripts of Brennu-Njáls saga or 'The Story of Burnt Njál', an Icelandic prose narrative from the end of the 13th century. One part of the project is devoted to linguistic variation in the earliest transmission of the text in parchment manuscripts and fragments from the 14th century. The following article gives a short overview over the design of the corpus that has to serve quite different purposes from palaeographic over stemmatological to literary research. I will focus on features important for the analysis of certain linguistic variables and the challenge lying in their implementation in a corpus consisting of close transcriptions of medieval manuscripts and give examples for the use of the corpus for linguistic research in the frame of the project that mainly consists of the analysis of different grammatical/syntactic constructions that are often referred to in connection with stylistic research (narrative inversion, historical present tense, indirect-speech constructions). | Mörkum Njálu! An Annotated Corpus to Analyse and Explain Grammatical Divergences between 14th-Century Manuscripts of Njáls saga |
d219302630 | ||
d13488809 | We present a discussion forum assistant based on deep recurrent neural networks (RNNs). The assistant is trained to perform three different tasks when faced with a question from a user. Firstly, to recommend related posts. Secondly, to recommend other users that might be able to help. Thirdly, it recommends other channels in the forum where people may discuss related topics. Our recurrent forum assistant is evaluated experimentally by prediction accuracy for the end-to-end trainable parts, as well as by performing an end-user study. We conclude that the model generalizes well, and is helpful for the users. | Assisting Discussion Forum Users using Deep Recurrent Neural Networks |
d960007 | We describe the Nara Institute of Science and Technology (NAIST) spelling check system in the shared task. Our system contains three components: a word segmentation based language model to generate correction candidates; a statistical machine translation model to provide correction candidates and a Support Vector Machine (SVM) classifier to rerank the candidates provided by the previous two components. The experimental results show that the kbest language model and the statistical machine translation model could generate almost all the correction candidates, while the precision is very low. However, using the SVM classifier to rerank the candidates, we could obtain higher precision with a little recall dropping. To address the low resource problem of the Chinese spelling check, we generate 2 million artificial training data by simply replacing the character in the provided training sentence with the character in the confusion set. | A Hybrid Chinese Spelling Correction Using Language Model and Statistical Machine Translation with Reranking |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.