Papers reviewed and determined not to be word norm studies. Use the flag icon to report errors or suggest re-inclusion.
18625 papers
Lexical units with reduplication of word-forming affixes in the texts of the 1114th centuries reflect the specificity of stylistic features of the Old Russian literature. They also serve as the means of promotion and fixation of important structural and semantic tendencies in the history of Russian. The reduplication of the affixes reflects peculiarity of literary speech, based on the intersection of the Church Slavonic and the original Russian linguistic traditions. It represents different genre-related norms of Old Russian, or norms with variation of expressive means as an inherent feature. Accordingly, the most active models of suffixal reduplication include genetically heterogeneous (Russian and Church Slavonic) synonymic affixes. The contamination of original Russian suffixes with their south Slavonic synonyms is considered to be an instrument of genre and stylistic adaptation of the derivatives to the specifics of a text. The semantic redundancy of words, being the result of the affixal reduplication, corresponds, first of all, to the general stylistic peculiarities of literary speech, and, secondly, is viewed as the basis of semantic development of some lexical and grammatical word classes: in the class of nomina abstracta the reduplication of suffixes contributs to concretization of the meanings of abstract names; in the sphere of verbal prefixation, the “threading” of synonymous affixes became the basis of development of new modifying meanings the modes of verbal action.
Summary of relatedness ratings and image features <br><br><br>
Abstract The paper describes standardized instruments that researchers and practitioners can use for assessing receptive vocabulary development and expressive vocabulary development in both Brazilian Portuguese and Brazilian Sign Language (Libras) in both hearing and deaf children from 2 to 14 years of age. It establishes comparisons among such instruments based on their characteristics and indications, and summarizes some data on their validity, reliability and norms. Norms provided herein permit distinguishing among five talker groups (very late talkers, late talkers, on-time talkers, early talkers, very early talkers) at each of five age levels (i.e., from 1 to 5 years of age), as well as five listener groups (very late listeners, late listeners, on-time listeners, early listeners, very early listeners) at each of nine school grade levels (i.e., from Kindergarten to 8th grade). Norms provided herein also permit distinguishing among five sign receiver deaf groups (very late receivers, late receivers, on-time receivers, early receivers, very early receivers) at each of seven school grade levels of deaf education (i.e., from 2nd to 8th grade). Keywords: vocabulary, speech comprehension, speech production, oral language, lexical development, PPVT
Lexical units with reduplication of word-forming affixes in the texts of the 11-14th centuries reflect the specificity of stylistic features of the Old Russian literature. They also serve as the means of promotion and fixation of important structural and semantic tendencies in the history of Russian. The reduplication of the affixes reflects peculiarity of literary speech, based on the intersection of the Church Slavonic and the original Russian linguistic traditions. It represents different genre-related norms of Old Russian, or norms with variation of expressive means as an inherent feature. Accordingly, the most active models of suffixal reduplication include genetically heterogeneous (Russian and Church Slavonic) synonymic affixes. The contamination of original Russian suffixes with their south Slavonic synonyms is considered to be an instrument of genre and stylistic adaptation of the derivatives to the specifics of a text. The semantic redundancy of words, being the result of the affixal reduplication, corresponds, first of all, to the general stylistic peculiarities of literary speech, and, secondly, is viewed as the basis of semantic development of some lexical and grammatical word classes: in the class of nomina abstracta the reduplication of suffixes contributs to concretization of the meanings of abstract names; in the sphere of verbal prefixation, the "threading" of synonymous affixes became the basis of development of new modifying meanings – the modes of verbal action.
The anterior medial temporal lobe (TL), including the amygdala, has been implicated in olfactory processing, e.g., coding for intensity and valence, and seems also involved in memory. With this background, the present study evaluated whether anterior medial TL-resections in TL epilepsy affected intensity and valence ratings, as well as free and cued identification of odors. These aspects of odor perception were assessed in 31 patients with unilateral anterior medial TL-resections (17 left, 14 right) and 16 healthy controls. Results suggest that the anterior medial TL is in particular necessary for free, but also cued, odor identification. TL resection was also found to impair odor valence, but not intensity ratings. Left resected patients rated nominally pleasant and unpleasant odors as more neutral suggesting a special role for the left anterior TL in coding for emotional saliency in response to odors.
Abstract This paper introduces a new bilingual Czech-English verbal valency lexicon (called CzEng-Vallex) representing a relatively large empirical database. It includes 20,835 aligned valency frame pairs (i.e., verb senses which are translations of each other) and their aligned arguments. This new lexicon uses data from the Prague Czech-English Dependency Treebank and also takes advantage of the existing valency lexicons for both languages: the PDT-Vallex for Czech and the EngVallex for English. The CzEngVallex is available for browsing as well as for download in the LINDAT/CLARIN repository. The CzEngVallex is meant to be used not only by traditional linguists, lexicographers, translators but also by computational linguists both for the purposes of enriching theoretical linguistic accounts of verbal valency from a cross-linguistic perspective and for an innovative use in various NLP tasks.
This paper presents the SyntaxTree Aligner, a web-based language-independent syntactic trees construction and alignment platform for monolingual and bilingual sentences. This platform has served as an assisting tool for language annotator to the construction of high-quality parallel treebank, in a visual way. It combines the strengths of some existing treebank construction tools, i.e. well-defined Tiger-XML representation for syntactic structures, Stockholm-XML format of tree alignments, and monolingual (bilingual) tree annotation and revision interface, while addressing some of their common shortcomings, i.e. tools have to be installed in the local computer. The platform supports various annotation schemes, such as the annotation scheme used in Penn Treebank, and the universal tag set. We report the ongoing construction work of two parallel treebanks, Chinese-English and Chinese-Portuguese, using SyntaxTree Aligner. The SyntaxTree Aligner is freely available at http://nlp2ct.cis.umac.mo.
This article deals with the problem of development of the dialectological base of the Machine fund of the Bashkir language. Along with the written monuments, folklore material, dialect is one of the most important sources for the study of the historical development and formation of the literary language. The main task of dialectologists is not only the collection but also the storage of dialect materials of the Bashkir language. Taking into account the importance of translation of dialect materials in electronic format and opening of free access to a wide audience to the information provided, the staff of the Laboratory of Linguistics and Information Technologies of the Institute of History, Language and Literature of the Ufa Scientific Center established a dialectological base as a part of the Machine fund of the Bashkir language consisting of three separate databases — lexical database, the database of dialectological atlas and textological database. The lexical database includes information about the dialectal lexicon of the Bashkir language. The database was developed on the basis of dialectal dictionaries compiled and published by the staff of the Institute of History, Language and Literature. The volume of the database is more than 52 000 dialectal units. The database of dialectological atlas is developed on the basis of materials of Dialectological atlas of the Bashkir language. The database allows to select the types of linguistic phenomena (phonetic, morphological, syntactic, lexical), for each type specific isoglosses were identified. Isoglosses allocated 250 strong points of the republic and neighboring regions. The textological database represents illustrative materials collected during numerous expeditions by the staff of the Institute of History, Language and Literature. To date, the database contains more than 500 texts in all dialects of the Bashkir language. Introduced in dialectological base material forms the basis of synchronic and diachronic study of the language features of dialects and sub-dialects of the Bashkir language.
Parsing Arabic language is a difficult task given the specificities of the language and given the scarcity of linguistic resources. Linguistic resources such as grammars are very important to any natural language processing application. Unfortunately, the manual construction of these resources is laborious and time-consuming. The use of annotated corpora as a knowledge database might be a solution to a fast construction of a grammar for a given language. In this paper, we began by presenting an overview of our method to automatically induce a probabilistic context free grammar from an Arabic annotated corpus (The Penn Arabic TreeBank). Then we tested the obtained grammar in the parsing task and we expose the evaluation results. Finally we present our vision of a hybrid method for parsing Modern Standard Arabic (MSA) that we believe that it could enhance obtained results.
The Danish term Eventyr derives from the Latin adventura and the French aventure. The Latin verb is advenire, meaning “to happen” or “occur.” The lexical meaning in Danish refers to an event or incident which to a significant degree is accidental or remarkable, especially one that is strange, surprising, or exciting. When used in the plural, the Danish may refer to erotic experiences or tales of love. Eventyr may also suggest a daring or even dangerous undertaking. In a third meaning, Eventyr refers to a story or tale about strange or marvelous events. These are usually fantastical stories, something that lies out of the norm. Eventyr may also refer to an implausible or untruthful story.1
Kamusi has been developing a system to analyze texts on the source side and present users with sense-specified dictionary options. Similarly to spellcheck, the user selects the intended meaning. We then use a multilingual lexical database to bridge to matching vocabulary in other languages. When paired with Freeling, additional pre-processing is possible for several languages. Integration with MT via Moses and Apertium is planned, but not yet undertaken. MWEs treatment is important. An MWE is lexicalized in the Kamusi database and marked for separability, with a definition and translation equivalents (one or more words) in other languages. When the initial term of an MWE appears in the source text, Pre:D queries the database and scans the sentence for all MWEs that could follow. The user can select the relevant MWE rather than the component words. A user can submit a missing sense or MWE for inclusion in the lexicon. Named entities can also be identified from data sources or by users and rendered appropriately across languages. When users agree, we will also use sense-tagged sentences for machine learning. A prototype of the core system is already functional.
Multilingualism is common offline, but we have a more limited understanding of the ways multilingualism is displayed online and the roles that multilinguals play in the spread of content between speakers of different languages. We take a computational approach to studying multilingualism using one of the largest user-generated content platforms, Wikipedia. We study multilingualism by collecting and analyzing a large dataset of the content written by multilingual editors of the English, German, and Spanish editions of Wikipedia. This dataset contains over two million paragraphs edited by over 15,000 multilingual users from July 8 to August 9, 2013. We analyze these multilingual editors in terms of their engagement, interests, and language proficiency in their primary and non-primary (secondary) languages and find that the English edition of Wikipedia displays different dynamics from the Spanish and German editions. Users primarily editing the Spanish and German editions make more compl)
The project aims to provide a semi-supervised approach to identify Multiword Expressions in a multilingual context consisting of English and most of the major Indian languages. Multiword expressions are a group of words which refers to some conventional or regional way of saying things. If they are literally translated from one language to another the expression will lose its inherent meaning. To automatically extract multiword expressions from a corpus, an extraction pipeline have been constructed which consist of a combination of rule based and statistical approaches. There are several types of multiword expressions which differ from each other widely by construction. We employ different methods to detect different types of multiword expressions. Given a POS tagged corpus in English or any Indian language the system initially applies some regular expression filters to narrow down the search space to certain patterns (like, reduplication, partial reduplication, compound nouns, compound verbs, conjunct verbs etc.). The word sequences matching the required pattern are subjected to a series of linguistic tests which include verb filtering, named entity filtering and hyphenation filtering test to exclude false positives. The candidates are then checked for semantic relationships among themselves (using Wordnet). In order to detect partial reduplication we make use of Wordnet as a lexical database as well as a tool for lemmatising. We detect complex predicates by investigating the features of the constituent words. Statistical methods are applied to detect collocations. Finally, lexicographers examine the list of automatically extracted candidates to validate whether they are true multiword expressions or not and add them to the multiword dictionary accordingly.
In der vorliegenden Arbeit wurde die Wirksamkeit einer Expositionstherapie in virtueller Realitat bei Zahnbehandlungsphobikern untersucht. Uber eine Vorher- und Nachher-Analyse sollte herausgefunden werden, inwieweit die Angst vor phobischen Stimuli reduziert werden kann. Die Untersuchungen dieser Studie stutzten sich auf zwei empirische EEG-Studien von Kenntner-Mabiala & Pauli (2005, 2008), die evaluierten, dass Emotionen, die Schmerzwahrnehmung und die Toleranz der Schmerzschwelle modulieren konnen. Zudem konnte in einer EEG-Studie von Leutgeb et al. (2011) gezeigt werden, dass Zahnbehandlungsphobiker eine Erhohung der EKPs auf phobisches Stimulusmaterial aufwiesen. Die Frage nach dem Einfluss von emotionalen und phobischen Bildern auf die neuronale Verarbeitung sollte hier untersucht werden. Auserdem sollte herausgefunden werden welche Auswirkung emotionale und phobische Gerausche auf die Schmerzverarbeitung vor und nach der Therapie haben. Die Probanden wurden an drei aufeinanderfolgenden Terminen untersucht. Der erste Termin beinhaltete die Diagnostik zur Zahnbehandlungsphobie und den experimentellen Teil, der sich in drei Teile pro Termin gliederte. Der erste Teil enthielt die Aufzeichnung des EEG unter Schmerzreizapplikation im Kontext emotionaler Gerausche (neutral, negativ, positiv & zahn) und das Bewerten dieser Schmerzreize bezuglich der Intensitat und der Unangenehmheit des Schmerzes. Der zweite Teil enthielt Ratings zu Valenz und Arousal bezuglich dieser emotionalen Gerauschkategorien. Der dritte Teil enthielt die Aufzeichnung des EEG und das Rating zu Valenz und Arousal bezuglich emotionaler Bildkategorien (neutral, negativ, zahn). Am zweiten Termin folgte die Expositionstherapie unter psychologischer Betreuung. Der dritte Termin diente zur Erfolgsmessung und verlief wie Termin eins. Als Erfolgsmase der Therapie dienten Selbstbeurteilungsfragebogen, Valenz- und Arousal-Ratings des Stimulusmaterials, Schmerzratings und die durch das EEG aufgezeichneten visuell Ereigniskorrelierten- und Somatosensorisch-Evozierten-Potentialen. Die Ergebnisse zeigten, dass Gerausche mit unterschiedlichen emotionalen Kategorien zu eindeutig unterschiedlichen Valenz- und Arousalempfindungen bei Zahnbehandlungsphobikern fuhren. Die Studie konnte bestatigen, dass phobische Gerauschstimuli einen Einfluss auf die erhohte Erregung bei Zahnbehandlungsphobikern haben, die nach der Intervention als weniger furchterregend empfunden werden. Zudem konnte erwiesen werden, dass Personen mit Zahnbehandlungsphobie durch das Horen phobischer Zahnbehandlungsgerausche eine starkere Schmerzempfindung aufwiesen als durch positive, neutrale und negative Gerausche. Die Ergebnisse der Somatosensorisch-Evozierten-Potenziale (N150, P260) im Vergleich der Vorher und Nachher-Analyse zeigten tendenzielle Modulationen, die jedoch nicht signifikant waren. Im Vergleich zur Pra-Messung nahm die N150 Amplitude in der Post-Messung fur die schmerzhaften Stimuli wahrend der phobischen und negativen Gerausche ab. Auserdem wurden in dieser Studie parallel zum Gerauschparadigma weitere Sinnesmodalitaten mit phobie-relevanten Reizen anhand von Bildern getestet. Parallel zu den Ergebnissen der Studie von Leutgeb et al. (2011) fanden wir eine verstarkte elektrokortikale Verarbeitung im Late-Positive-Potential (LPP) auf phobische Bilder bei Zahnbehandlungsphobikern. Die Erwartung, dass die verstarkte elektrokortikale Verarbeitung des LPPs auf phobische Bilder bei Zahnbehandlungsphobikern durch Intervention reduziert werden kann, konnte nicht belegt werden. Rein deskriptiv gehen die Ergebnisse aber in diese Richtung. Auch das Verhalten anderte sich durch die Teilnahme an der Studie. Die Probanden gaben an, dass sich ihre Zahnbehandlungsangst nach der Expositionstherapie signifikant verringert hat. Das telefonische Follow-Up 6 Monate nach der Post-Messung zeigte, dass sich einige Probanden nach mehreren Jahren wieder in zahnarztliche Behandlung begeben haben. Insgesamt kann diese Studie zeigen, dass Zahnbehandlungsphobie durch psychologische Intervention reduziert werden kann und auch die Angst vor phobischem Stimulusmaterial durch eine wiederholte Reizkonfrontation abnimmt. Jedoch konnte auf elektrokortikaler Ebene keine Modulation der Schmerzempfindung uber emotionale Gerausche festgestellt werden.
Quantitative analysis of organismal form is an important component for almost every branch of biology. Although generally considered an easily-measurable structure, the quantification of gastropod shell form is still a challenge because many shells lack homologous structures and have a spiral form that is difficult to capture with linear measurements. In view of this, we adopt the idea of theoretical modelling of shell form, in which the shell form is the product of aperture ontogeny profiles in terms of aperture growth trajectory that is quantified as curvature and torsion, and of aperture form that is represented by size and shape. We develop a workflow for the analysis of shell forms based on the aperture ontogeny profile, starting from the procedure of data preparation (retopologising the shell model), via data acquisition (calculation of aperture growth trajectory, aperture form and ontogeny axis), and data presentation (qualitative comparison between shell forms) and ending with)
Статтю присвячено аналізові мовної компресії як чинника функціонування стислих текстів, а саме: дослідженню рівнів та засобів компресії повідомлення та їхнього впливу на видовий тип стислого тексту. У стислих текстах використовують компресію на всіх рівнях: графемному, морфологічному, синтаксичному, лексичному, синтактико-лексичному, семантичному, стилістичному. Багатство використаних рівнів і засобів конденсації змісту і форми та дотримання мовних норм або відхилення від них визначають тип стислого тексту за ступенем мовної компресії: нормовано стислий текст або форсовано стислий текст. Ключові слова: мовна компресія, рівні мовної компресії, засоби мовної компресії, стислий текст, види стислих текстів. Статья посвящена анализу языковой компрессии как фактора функционирования сжатых текстов, а именно: исследованию уровней и способов компрессии сообщения. В сжатых текстах используют компрессию на всех уровнях: графемном, морфологическом, синтаксическом, лексическом, синтактико-лексическом, семантическом, стилистическом. Богатство используемых уровней и способов конденсации содержания и формы, а также соблюдение языковых норм или отклонение от них определяют тип сжатого текста по степени языковой компрессии: нормированно сжатый или форсированно сжатый текст. Ключевые слова: языковая компрессия, уровни языковой компрессии, средства языковой компрессии, сжатый текст, виды сжатых текстов. The problems of compressed texts are very important for modern society due to their pragmatism. The issues of the rational representation and pragmatic approach to the information are relevant today because of the information space overload. The linguistic compression plays one of the key roles in the text compression process. The problems of the compressed texts haven’t been investigated completely in Ukrainian linguistics. This paper deals with linguistic compression as a factor of the compressed text functioning. Compressed texts function in the different fields of the human life. The aim of this article is to enlighten the role of the linguistic compression in creating and functioning of the compressed texts. Different levels of information condensation are used for shorting the size and condensing the text content: grapheme, morphologic, syntactic, lexical, semantic, stylistic levels. Every level has the wide spectrum of its means for the text condensation. The quantity of the levels and the means of the text condensation as well as following the linguistic norms or breaking them determine the type of the compressed texts: standardized compressed text or forced compressed text. They have the wider information content and semantic meaning under their formal structure. The means of information condensation assist in reducing efforts, time, space for making up a text. Further investigations should deal with analyses of the contemporary aproach to the compressed texts creating and their specific characteristics. Key words: linguistic compression, linguistic compression levels, linguistic compression means, compressed text, compressed texts types.
As visual media spread to all domains of public and scientific life, nonverbal behavior is taking its place as an important form of communication alongside the written and spoken word. An objective and reliable method of analysis for hand movement behavior and gesture is therefore currently required in various scientific disciplines, including psychology, medicine, linguistics, anthropology, sociology, and computer science. However, no adequate common methodological standards have been developed thus far. Many behavioral gesture-coding systems lack objectivity and reliability, and automated methods that register specific movement parameters often fail to show validity with regard to psychological and social functions. To address these deficits, we have combined two methods, an elaborated behavioral coding system and an annotation tool for video and audio data. The NEUROGES–ELAN system is an effective and user-friendly research tool for the analysis of hand movement behavior, including gesture, self-touch, shifts, and actions. Since its first publication in 2009 in Behavior Research Methods, the tool has been used in interdisciplinary research projects to analyze a total of 467 individuals from different cultures, including subjects with mental disease and brain damage. Partly on the basis of new insights from these studies, the system has been revised methodologically and conceptually. The article presents the revised version of the system, including a detailed study of reliability. The improved reproducibility of the revised version makes NEUROGES–ELAN a suitable system for basic empirical research into the relation between hand movement behavior and gesture and cognitive, emotional, and interactive processes and for the development of automated movement behavior recognition methods.
In this study we propose a novel, unsupervised clustering methodology for analyzing large datasets. This new, efficient methodology converts the general clustering problem into the community detection problem in graph by using the Jensen-Shannon distance, a dissimilarity measure originating in Information Theory. Moreover, we use graph theoretic concepts for the generation and analysis of proximity graphs. Our methodology is based on a newly proposed memetic algorithm (iMA-Net) for discovering clusters of data elements by maximizing the modularity function in proximity graphs of literary works. To test the effectiveness of this general methodology, we apply it to a text corpus dataset, which contains frequencies of approximately 55,114 unique words across all 168 written in the Shakespearean era (16th and 17th centuries), to analyze and detect clusters of similar plays. Experimental results and comparison with state-of-the-art clustering methods demonstrate the remarkable performance )
Objective: This study examines reading aloud in patients with amyotrophic lateral sclerosis (ALS) and those with frontotemporal dementia (FTD) in order to determine whether differences in patterns of speaking and pausing exist between patients with primary motor vs. primary cognitive-linguistic deficits, and in contrast to healthy controls. Design: 136 participants were included in the study: 33 controls, 85 patients with ALS, and 18 patients with either the behavioural variant of FTD (FTD-BV) or progressive nonfluent aphasia (FTD-PNFA). Participants with ALS were further divided into 4 non-overlapping subgroups—mild, respiratory, bulbar (with oral-motor deficit) and bulbar-respiratory—based on the presence and severity of motor bulbar or respiratory signs. All participants read a passage aloud. Custom-made software was used to perform speech and pause analyses, and this provided measures of speaking and articulatory rates, duration of speech, and number and duration of pauses. Thes)
The historical development and the linguistic triggering environments for Oberfeld formation in German subordinate clauses represent long-standing research questions in Germanic Linguistics, dating at least as far back as Jacob Grimm's famous Deutsche Grammatik. The present corpus study traces this historical development back to the 17th century. The study is based on three text corpora. For contemporary German, two syntactically annotated newspaper corpora were consulted: the TüBa-D/Z and TüPP-D/Z treebanks,1 which provide linguistic annotations for articles published in the daily newspaper die tageszeitung (taz). For diachronic data, the corpus collection Deutsches Textarchiv (DTA)2 was utilized. The DTA contains texts ranging from 1610 to 1900. All three corpus collections are part of the Common Language Resources and Technology Infrastructure (CLARIN) initiative3 and will be developed further as part of the CLARIN research infrastructure. The study demonstrates the added value that annotated corpora can provide for in-depth studies in historical syntax. At the same time it showcases the added value of interoperable language resources for linguistic investigations that require access to and analysis of multiple linguistic resources.
Abstract In this article we report the results for five POS taggers, i.e., the Mate tagger, the Hunpos tagger, RFTagger, theOpenNLP tagger, andNLTKUnigramtagger, tested on the data of the Ancient Greek Dependency Treebank. This is done in order to find the most efficient POS tagger to use for pre-annotation of new treebank data. A corrected 1-run 10-fold cross validation t test shows that the Mate tagger outperforms all the other taggers, with an accuracy score of 88%.
This paper builds off recent work from Kiperwasser & Goldberg (2016) using neural attention in a simple graph-based dependency parser. We use a larger but more thoroughly regularized parser than other recent BiLSTM-based approaches, with biaffine classifiers to predict arcs and labels. Our parser gets state of the art or near state of the art performance on standard treebanks for six different languages, achieving 95.7% UAS and 94.1% LAS on the most popular English PTB dataset. This makes it the highest-performing graph-based parser on this benchmark---outperforming Kiperwasser Goldberg (2016) by 1.8% and 2.2%---and comparable to the highest performing transition-based parser (Kuncoro et al., 2016), which achieves 95.8% UAS and 94.6% LAS. We also show which hyperparameter choices had a significant effect on parsing accuracy, allowing us to achieve large gains over other graph-based approaches.
Research finds we make spontaneous trait inferences from facial appearance, even after brief exposures to a face (i.e., ≤ 100 ms). We examined spontaneous impressions of criminality from facial appearance, testing whether these impressions persist after repeated presentation (i.e., one to three exposures) and increased exposure duration (100, 500, or 1000 ms) to the face. Judgement confidence and response times were recorded. Other participants viewed the faces for an unlimited period of time, rating trustworthiness dominance, and criminal appearance. We found evidence that participants spontaneously make criminal appearance attributions. These inferences persisted with repeated presentation and increased exposure duration, were related to trustworthiness and dominance ratings, and were made with high confidence. Implications are discussed.
This paper reports on the development of a French FrameNet, within the ASFALDA project. While the first phase of the project focused on the development of a French set of frames and corresponding lexicon (Candito et al., 2014), this paper concentrates on the subsequent corpus annotation phase, which focused on four notional domains (commercial transactions, cognitive stances, causality and verbal communication). Given full coverage is not reachable for a relatively " new " FrameNet project, we advocate that focusing on specific notional domains allowed us to obtain full lexical coverage for the frames of these domains, while partially reflecting word sense ambiguities. Furthermore, as frames and roles were annotated on two French Treebanks (the French Treebank (Abeillé and Barrier, 2004) and the Sequoia Treebank (Candito and Seddah, 2012), we were able to extract a syntactico-semantic lexicon from the annotated frames. In the resource's current status, there are 98 frames, 662 frame-evoking words, 872 senses, and about 13000 annotated frames, with their semantic roles assigned to portions of text. The French FrameNet is freely available at alpage.inria.fr/asfalda.
Affective computing is a very important issue. An increasing amount of research focused on representing affective states as continuous numerical values on multiple dimensions. Such as the emotional space, which is about the valence and arousal. Due to the affective dimension representation can be useful to sentiment analysis, building dimensional sentiment resources with valence-arousal ratings are very important. Therefore, this study proposes a method to automatically obtain the valence-arousal ratings of affective words. Experiment results using the evaluation metrics to get the error rates about the mean absolute error and pearson correlation coefficient.
This work elaborates the semi-semantic part of speech annotation guidelines for the URDU.KON-TB treebank: an annotated corpus. A hierarchical annotation scheme was designed to label the part of speech and then applied on the corpus. This raw corpus was collected from the Urdu Wikipedia and the Jang newspaper and then annotated with the proposed semi-semantic part of speech labels. The corpus contains text of local & international news, social stories, sports, culture, finance, religion, traveling, etc. This exercise finally contributed a part of speech annotation to the URDU.KON-TB treebank. Twenty-two main part of speech categories are divided into subcategories, which conclude the morphological, and semantical information encoded in it. This article reports the annotation guidelines in major; however, it also briefs the development of the URDU.KON-TB treebank, which includes the raw corpus collection, designing & employment of annotation scheme and finally, its statistical evaluation and results. The guidelines presented as follows, will be useful for linguistic community to annotate the sentences not only for the national language Urdu but for the other indigenous languages like Punjab, Sindhi, Pashto, etc., as well.
Attention-Deficit/Hyperactivity Disorder (ADHD) is characterized by symptoms of inattentiveness and hyperactivity/impulsivity. Besides, increasing evidence points to ADHD patients showing emotional dysfunctions and concomitant problems in social life. However, systematic research on emotional dysfunctions in ADHD is still rare, and to date most studies lack conceptual differentiation between emotion processing and emotion regulation. The aim of this thesis was to systematically investigate emotion processing and emotion regulation in adult ADHD in a virtual reality paradigm implementing social interaction. Emotional reactions were assessed on experiential, physiological, and behavioral levels. Experiment 1 was conducted to develop a virtual penalty kicking paradigm implying social feedback and to test it in a healthy sample. This paradigm should then be applied in ADHD patients later on. Pleasant and unpleasant trials in this paradigm consisted of hits respectively misses and subsequent feedback from a virtual coach. In neutral trials, participants were teleported to different spots of the virtual stadium. Results indicated increased positive affectivity (higher valence and arousal ratings, higher zygomaticus activations, and higher expression rates of positive emotional behavior) in response to pleasant compared to neutral trials. Reactions to unpleasant trials were contradictory, indicating increased levels of both positive and negative affectivity, compared to neutral trials. Unpleasant vs. neutral trials revealed lower valence ratings, higher arousal ratings, higher zygomaticus activations, slightly lower corrugator activations, and higher expression rates of both positive and negative emotional behavior. The intensity of emotional reactions correlated with experienced presence in the virtual reality. To better understand the impact of hits or misses per se vs. hits or misses with coach feedback healthy participants’ emotional reactions, only 50% of all shots were followed by coach feedback in experiment 2. Neutral trials consisted of shots over the free soccer field which were followed by coach feedback in 50 % of all trials. Shots and feedback evoked more extreme valence and arousal ratings, higher zygomaticus activations, lower corrugator activations, and higher skin conductance responses than shots alone across emotional conditions. Again, results speak for the induction of positive emotions in pleasant trials whereas the induction of negative emotions in unpleasant trials seems ambiguous. Technical improvements of the virtual reality were reflected in higher presence ratings than in experiment 1. Experiment 3 investigated emotional reactions of adult ADHD patients and healthy controls after emotion processing and response-focused emotion regulation. Participants successively went through an ostensible online ball-tossing game (cyber ball) inducing negative emotions, and an adapted version of the virtual penalty kicking game. Throughout cyber ball, participants were included or ostracized by two other players in different experimental blocks. Participants were instructed to explicitly show, not regulate, or hide their emotions in different experimental blocks. Results provided some evidence for deficient processing of positive emotions in ADHD. Patients reported slightly lower positive affect than controls during cyber ball, gave lower valence ratings than controls in response to pleasant penalty kicking trials, and showed lower zygomaticus activations than controls especially during penalty kicking. Patients in comparison with controls showed slightly increased processing of unpleasant events during cyber ball (higher ratings of negative affect, especially in response to ostracism), but not during penalty kicking. Patients showed lower baseline skin conductance levels than controls, and impaired skin conductance modulations. Compared to controls, patients showed slight over-expression of positive as well as negative emotional behavior. Emotion regulation analyses revealed no major difficulties of ADHD vs. controls in altering their emotional reactions through deliberate response modulation. Moreover, patients reported to habitually apply adaptive emotion regulation strategies even more frequently than controls. The analyses of genetic high-risk vs. low-risk groups for ADHD across the whole sample revealed similar results as analyses for patients vs. controls for zygomaticus modulations during emotion processing, and for modulations of emotional reactions due to emotion regulation. To sum up, the virtual penalty kicking paradigm proved to be successful for the induction of positive, but not negative emotions. The importance of presence in virtual reality for the intensity of induced emotions could be replicated. ADHD patients showed impaired processing of primarily positive emotions. Aberrations in negative emotional responding were less clear and need further investigation. Results point to adult ADHD in comparison to healthy controls suffering from baseline deficits in autonomic arousal and deficits in arousal modulation. Deficits of ADHD in the deliberate application of response-focused emotion regulation could not be found.
A description and annotation guidelines for the Yahoo Webscope release of Query Treebank, Version 1.0, May 2016.
In this paper, we present an overview of some issues related to the use of Big Data in the area of Linguistics that have been debated in workshops and conferences in the last two years. We also consider some requirements that "big" linguistic databases should have in order to tackle some of these issues; finally, we discuss a set of possible interactive visualization approaches of large datasets that may have an impact in this research field.
Transient global amnesia (TGA) is a disorder with reversible anterograde disturbance of explicit memory, frequently preceded by an emotionally or physically stressful event. By using magnetic resonance imaging (MRI) following an episode of TGA, small hippocampal lesions have been observed. Hence it has been postulated that the disorder is caused by the stress-related transient inhibition of memory formation in the hippocampus. In experimental studies, stress has been shown to affect both explicit and implicit learning-the latter defined as learning and memory processes that lack conscious awareness of the information acquired. To test the hypothesis that impairment of implicit learning in TGA is present and related to stress, we determined the effect of experimental exposure to stress on hippocampal activation patterns during an implicit learning paradigm in patients who suffered a recent TGA and healthy matched control subjects. We used a hippocampus-dependent aversive learning procedure (context conditioning with the phases habituation, acquisition, and extinction) during functional MRI following experimental stress exposure (socially evaluated cold pressor test). After a control procedure, controls showed successful learning during the acquisition phase, indicated by increased valence, arousal and contingency ratings to the paired (CON+) vs. the non-paired (CON-) conditioned stimulus, and successful extinction of the conditioned responses. Following stress, acquisition was still successful, however extinction was impaired with persistently increased contingency ratings. In contrast, TGA patients showed impairment of conditioned responses and insufficient extinction after the control procedure, indicated by a lack of significant differences between CON+ and CON- for valence and arousal ratings after the acquisition phase and by significantly increased contingency ratings after the extinction. After stress, aversive learning was not successful with non-significant ratings of all parameters. Concerning brain activation patterns after the control procedure, controls showed increased hippocampal response during acquisition after the control procedure. This was not seen after stress exposure. In TGA patients, we observed an increased response in the right ventral striatum in the acquisition phase following stress. These findings suggest that alterations in implicit learning processes, including impaired hippocampal and increased striatal responses, might play a role in TGA pathophysiology, partly related to acute stress.
This article addresses the question of the types of morphological relatedness that the lexical class of the adjective presents in Old English. After an exhaustive analysis of the derivational paradigms of the language based on data retrieved from the lexical database of Old English Nerthus, the following conclusions are reached. Two types of morphological relatedness are identifiable, namely explicit and implicit. Short distance and long distance relations overlap with explicit and implicit morphological relatedness. These relations involve four types of units, to wit lexical primes (the bases of lexical paradigms), derived adjectives (the input to recursive processes of word-formation), target adjectives (the output of processes that cannot be inputted to a recursive process) and morphologically unrelated adjectives (which are neither the input nor the output of a process of word-formation). © 2016 Society for Studia Neophilologica.
Anxiety disorders may not only be characterized by specific symptomatology (e.g., tachycardia) in response to the fearful stimulus (primary problem or first-level emotion) but also by the tendency to negatively evaluate oneself for having those symptoms (secondary problem or negative meta-emotion). An exploratory study was conducted driven by the hypothesis that reducing the secondary or meta-emotional problem would also diminish the fear response to the phobic stimulus. Thirty-three phobic participants were exposed to the phobic target before and after undergoing a psychotherapeutic intervention addressed to reduce the meta-emotional problem or a control condition. The electrocardiogram was continuously recorded to derive heart rate (HR) and heart rate variability (HRV) and affect ratings were obtained. Addressing the meta-emotional problem had the effect of reducing the physiological but not the subjective symptoms of anxiety after phobic exposure. Preliminary findings support the role of the meta-emotional problem in the maintenance of response to the fearful stimulus (primary problem).
Drawing on the work of Lawrence Abu Hamdan, a British-Lebanese artist and researcher currently based in Beirut, this essay examines the juridical and conceptual field of critical forensis which is situated at the juncture of security studies, art, and architecture. Abu Hamdan extends forensics to the area of “new audibilities,” with a focus on the politics of juridical hearing in situations of legal-identity profiling and voice authentication (the “shibboleth test”). Abu Hamdan's projects investigate how accent monitoring and audio surveillance, voice recognition, translation technologies, sovereign acts of listening, and court determinations of linguistic norms emerge as so many technical constraints on “freedom of speech,” itself a malleable term ascribed to discrepant claims and principles, yet taking on performative force in site-specific situations.
PURPOSE: The focus of this study was to examine the influence of fundamental frequency (F0) and vocal tract length (VTL) modifications on speaker gender recognition in cochlear implant (CI) recipients for different stimulus types. METHOD: Single words and sentences were manipulated using isolated or combined F0 and VTL cues. Using an 11-point rating scale, CI recipients and listeners with normal hearing rated the maleness/femaleness of the corresponding voice. RESULTS: Speaker gender ratings for combined F0 and VTL modifications were similar across all stimulus types in both CI recipients and listeners with normal hearing, although the CI recipients showed a somewhat larger ambiguity. In contrast to listeners with normal hearing, F0-VTL and F0-only modifications revealed similar ratings in the CI recipients when using words as stimuli. However, when sentences were used, a difference was found between F0-VTL-based and F0-based ratings. Modifying VTL cues alone did not affect ratings in the CI group. CONCLUSIONS: Whereas speaker gender ratings by listeners with normal hearing relied on combined VTL and F0 cues, CI recipients made only limited use of VTL cues, which might be one reason behind problems with identifying the speaker on the basis of voice. However, use of the voice cues depended on stimulus type, with the greater information in sentences allowing a more detailed analysis than single words in both listener groups.
In recent years, dependency parsing is a fascinating research topic and has a lot of applications in natural language processing. In this paper, we present an effective approach to improve dependency parsing by utilizing supertag features. We performed experiments with the transition-based dependency parsing approach because it can take advantage of rich features. Empirical evaluation on Vietnamese Dependency Treebank showed that, we achieved an improvement of 18.92% in labeled attachment score with gold supertags and an improvement of 3.57% with automatic supertags.
International audience
In this paper, a German verb resource for verb-centered sentiment inference is introduced and evaluated. Our model specifies verb polarity frames that capture the polarity effects on the fillers of the verb's arguments given a sentence with that verb frame. Verb signatures and selectional restrictions are also part of the model. An algorithm to apply the verb resource to treebank sentences and the results of our first evaluation are discussed.
This article attempts to place dependency annotation options on a solid theoretical and applied footing. By verifying the validity of some basic choices of the current dependency reference framework, Universal Dependencies (UD), in a perspective of general annotation principles, we show how some choices can lead to inconsistencies and discontinuities, partly due to UD's alternation between syntax and semantics. For some constructions, we propose better suited alternative structures with a clear-cut distinction of syntax and semantics. We propose a classification of conception-oriented, annotatororiented, and finally, treebank end-useroriented considerations to be used in the creation of new annotation schemes.
One of the characteristics of writing in Modern Standard Arabic (MSA) is that the commonly used orthography is mostly consonantal and does not provide full vocalization of the text. It sometimes includes optional diacritical marks (henceforth, diacritics or vowels). Arabic script consists of two classes of symbols: letters and diacritics. Letters comprise long vowels such as A, y, w as well as consonants. Diacritics on the other hand comprise short vowels, gemination markers, nunation markers, as well as other markers (such as hamza, the glottal stop which appears in conjunction with a small number of letters, dots on letters, elongation and emphatic markers) which in all, if present, render a more or less exact precise reading of a word. In this study, we are mostly addressing three types of diacritical marks: short vowels, nunation, and shadda (gemination). Diacritics are extremely useful for text readability and understanding. Their absence in Arabic text adds another layer of lexical and morphological ambiguity. Naturally occurring Arabic text has some percentage of these diacritics present depending on genre and domain. For instance, religious text such as the Quran is fully diacritized to minimize chances of reciting it incorrectly. So are children's educational texts. Classical poetry tends to be diacritized as well. However, news text and other genre are sparsely diacritized (e.g., around 1.5% of tokens in the United Nations Arabic corpus bear at least one diacritic (Diab et al., 2007)). In general, building models to assign diacritics to each letter in a word requires a large amount of annotated training corpora covering different topics and domains to overcome the sparseness problem. The currently available diacritized MSA corpora are generally limited to the newswire genres (those distributed by the LDC) or religion related texts such as Quran or the Tashkeela corpus. In this paper we present a pilot study where we annotate a sample of non-diacritized text extracted from five different text genres. We explore different annotation strategies where we present the data to the annotator in three modes: basic (only forms with no diacritics), intermediate (basic forms–POS tags), and advanced (a list of forms that is automatically diacritized). We show the impact of the annotation strategy on the annotation quality. It has been noted in the literature that complete diacritization is not necessary for readability Hermena et al. (2015) as well as for NLP applications, in fact, (Diab et al., 2007) show that full diacritization has a detrimental effect on SMT. Hence, we are interested in discovering the optimal level of diacritization. Accordingly, we explore different levels of diacritization. In this work, we limit our study to two diacritization schemes: FULL and MIN. For FULL, all diacritics are explicitly specified for every word. For MIN, we explore what is a minimum and optimal number of diacritics that needs to be added in order to disambiguate a given word in context and make a sentence easily readable and unambiguous for any NLP application. We conducted several experiments on a set of sentences that we extracted from five corpora covering different genres. We selected three corpora from the currently available Arabic Treebanks from the Linguistic Data Consortium (LDC). These corpora were chosen because they are fully diacritized and had undergone significant quality control, which will allow us to evaluate the anno tation accuracy as well as our annotators understanding of the task. We select a total of 16,770 words from these corpora for annotation. Three native Arabic annotators with good linguistic background annotated the corpora samples. Diab et al. (2007), define six different diacritization schemes that are inspired by the observation of the relevant naturally occurring diacritics in different texts. We adopt the FULL diacritization scheme, in which all the diacritics should be specified in a word. Annotators were asked to fully diacritize each word. The text genres were annotated following the different strategies: - Basic: In this mode, we ask for annotation of words where all diacritics are absent, including the naturally occurring ones. The words are presented in a raw tokenized format to the annotators in context. - Intermediate: In this mode, we provide the annotator with words along with their POS information. The intuition behind adding POS is to help the annotator disambiguate a word by narrowing down on the diacritization possibilities. - Advanced: In this mode, the annotation task is formulated as a selection task instead of an editng task. Annotators are provided with a list of automatically diacritized candidates and are asked to choose the correct one, if it appears in the list. Otherwise, if they are not satisfied with the given candidates, they can manually edit the word and add the correct diacritics. This technique is designed in order to reduce annotation time and especially reduce annotator workload. For each word, we generate a list of vowelized candidates using MADAMIRA (Pasha et al., 2014). MADAMIRA is able to achieve a lemmatization accuracy 99.2% and a diacritization accuracy of 86.3%. We present the annotator with the top three candidates suggested by MADAMIRA, when possible. Otherwise, only the available candidates are provided. We also provided annotators with detailed guidelines, describing our diacritization scheme and specifying how to add diacritics for each annotation strategy. We described the annotation procedure and specified how to deal with borderline cases. We also provided in the guidelines many annotated examples to illustrate the various rules and exceptions. In order to determine the most optimized annotation setup for the annotators, in terms of speed and efficiency, we test the results obtained following the three annotation strategies. These annotations are all conducted for the FULL scheme. We first calculated the number of words annotated per hour, for each annotator and in each mode. As expected, following the Advanced mode, our three annotators could annotate an average of 618.93 words per hour which is double those annotated in the Basic mode (only 302.14 words). Adding POS tags to the Basic forms, as in the Intermediate mode, does not accelerate the process much. Only − 90 more words are diacritized per hour compared to the basic mode. Then, we evaluated the Inter-Annotator Agreement (IAA) to quantify the extent to which independent annotators agree on the diacritics chosen for each word. For every text genre, two annotators were asked to annotate independently a sample of 100 words. We measured the IAA between two annotators by averaging WER (Word Error Rate) over all pairs of words. The higher the WER between two annotations, the lower their agreement. The results obtained show clearly that the Advanced mode is the best strategy to adopt for this diacritization task. It is the less confusing method on all text genres (with WER between 1.56 and 5.58). We also conducted a preliminary study for a minimum diacritization scheme. This is a diacritization scheme that encodes the most relevant differentiating diacritics to reduce confusability among words that look the same (homographs) when undiacritized but have different readings. Our hypothesis in MIN is that there is an optimal level of diacritization to render a text unambiguous for processing and enhance its readability. We showed the difficulty in defining such a scheme and how subjective this task can be. Acknowledgement This publication was made possible by grant NPRP-6-1020-1-199 from the Qatar National Research Fund (a member of the Qatar Foundation).
My research focuses on the study of grammatical change in the recent history of the English language; in particular, I am currently writing my PhD dissertation on from Early Modern English to Present-Day English. In this PhD project, I analyse and compare the different factors that appear to influence in Present-Day English with earlier stages of the language (Late Modern English).The concept of ellipsis refers to a syntactic strategy in which expected elements have been left unpronounced in certain constructions. This omission triggers a mismatch between meaning (the intended message) and sound (what is in fact uttered). In particular, my research focuses on those examples of (Miller 2011, Miller and Pullum 2013), i.e. ellipsis types that occur after the following licensors (that is, those elements that license ellipsis): modal verbs, auxiliaries be, have and do, infinitival marker to and negator not. The main aim is to carry out an empirical analysis of from Late Modern English to Present-Day English (1700-1914), both quantitatively and qualitatively, by means of data retrieved from the Penn Corpora of Historical English. This project pays attention to syntactic variation, genre distribution and discourse variables (type of anaphora, mismatches in polarity, aspect, voice, modality, tense; comparison of clause types; distance, linking, type of focus).Esta tesis doctoral versa sobre la variacion diacronica de la elipsis sintactica desde ingles moderno temprano hasta la actualidad. La elipsis representa un desajuste entre el significado de lo que se dice (la intencion de un mensaje) y lo que se pronuncia en realidad. En el ambito de la linguistica moderna, la elipsis se estudia en el campo la semantica, la sintaxis, la pragmatica, la psicolinguistica, la linguistica de corpus, etc. y constituye la novedad de esta investigacion el tratamiento empirico de este fenomeno linguistico tratando de juntar variables procedentes de las distintas teorias consultadas. Cabe destacar que no existe un estudio multidisciplinar sobre este concepto sintactico en ingles moderno. Existen solo unos pocos estudios relativamente recientes pero se centran unicamente en el estudio del ingles actual. Por esta razon, se decidio llevar a cabo la investigacion tanto de los aspectos formales como de los funcionales de la elipsis y su evolucion diacronica teniendo en cuenta distintas variables discursivas. Para eso, se creo una base de datos en la que apareceria el ejemplo de Post-Auxiliary Ellipsis (elipsis despues de un auxiliar) con su numero identificador; el genero al que pertenece (de los dieciocho diferentes que existen en el corpus utilizado, el Penn Treebank); el licensor (aquel elemento que posibilita la elipsis); el tipo de union entre el antecedente de la elipse y la clausula eliptica (coordinacion, subordinacion, parataxis, etc.); el tipo de conector entre el antecedente y la parte la distancia existente entre el antecedente y la clausula eliptica (numero de clausulas); contexto sintactico en el que aparece la elipsis (clausulas principales, subordinadas, question-tags, etc); tipo de anafora (anaforico, cataforico, exoforico); categoria del antecedente (sintagma verbal, nominal, adjetival o no constituyente); categoria del material elidido (sintagma verbal, nominal, adjetival o no constituyente); presencia o ausencia de cambio de referente; comparacion del aspecto, la voz, la modalidad y el tiempo del antecedente con respecto a la parte presencia o ausencia de question-tag; comparacion entre el tipo de clausula del antecedente (declarativa, interrogativa o imperativa) y el de la clausula elidida; y tipo de foco de la parte elidida (auxiliary-choice (eleccion de auxiliar), subject-choice (eleccion de sujeto) o ambos).Esta tese de doutoramento versa sobre a variacion diacronica da elipse sintactica dende ingles moderno temperan ata a actualidade. A elipse representa un desaxuste entre o significado do que se di (a intencion dunha mensaxe) e o que se pronuncia en realidade. No ambito da linguistica moderna, a elipse estudase no campo a semantica, a sintaxe, a pragmatica, a psicolinguistica, a linguistica de corpus, etc. e constitue a novidade desta investigacion o tratamento empirico deste fenomeno linguistico tratando de xuntar variables procedentes das distintas teorias consultadas. Cabe destacar que non existe un estudo multidisciplinar sobre este concepto sintactico en ingles moderno. Existen so uns poucos estudos relativamente recentes pero centranse unicamente no estudo do ingles actual. Por esta razon, decidiuse levar a cabo a investigacion tanto dos aspectos formais coma dos funcionais da elipse e a sua evolucion diacronica tendo en conta distintas variables discursivas. Para iso, creouse unha base de datos na que apareceria o exemplo de Post-Auxiliary Ellipsis (elipse despois dun auxiliar) co seu numero identificador; o xenero ao que pertence (dos dezaoito diferentes que existen no corpus utilizado, o Penn Treebank); o licensor (aquel elemento que posibilita a elipse); o tipo de union entre o antecedente da elipse e a clausula eliptica (coordinacion, subordinacion, parataxe, etc.); o tipo de conector entre o antecedente e a parte a distancia existente entre o antecedente e a clausula eliptica (numero de clausulas); contexto sintactico no que aparece a elipse (clausulas principais, subordinadas, question-tags, etc); tipo de anafora (anaforico, cataforico, exoforico); categoria do antecedente (sintagma verbal, nominal, adxectival ou non constituinte); categoria do material elidido (sintagma verbal, nominal, adxectival ou non constituinte); presenza ou ausencia de cambio de referente; comparacion do aspecto, a voz, a modalidade e o tempo do antecedente con respecto a parte presenza ou ausencia de question-tag; comparacion entre o tipo de clausula do antecedente (declarativa, interrogativa ou imperativa) e o da clausula elidida; e tipo de foco da parte elidida (auxiliary-choice (eleccion de auxiliar), subject-choice (eleccion de suxeito) ou ambos).
This paper presents a novel latent variable recurrent neural network architecture for jointly modeling sequences of words and (possibly latent) discourse relations between adjacent sentences. A recurrent neural network generates individual words, thus reaping the benefits of discriminatively-trained vector representations. The discourse relations are represented with a latent variable, which can be predicted or marginalized, depending on the task. The resulting model can therefore employ a training objective that includes not only discourse relation classification, but also word prediction. As a result, it outperforms state-of-the-art alternatives for two tasks: implicit discourse relation classification in the Penn Discourse Treebank, and dialog act classification in the Switchboard corpus. Furthermore, by marginalizing over latent discourse relations at test time, we obtain a discourse informed language model, which improves over a strong LSTM baseline.