Sentence Segmentation Research Articles

Entity-fishing: A DARIAH Entity Recognition and Disambiguation Service

This paper presents an attempt to provide a generic named-entity recognition and disambiguation module (NERD) called entity-fishing as a stable online service that demonstrates the possible delivery of sustainable technical services within DARIAH, the European digital research infrastructure for the arts and humanities. Deployed as part of the national infrastructure Huma-Num in France, this service provides an efficient state-of-the-art implementation coupled with standardised interfaces allowing an easy deployment on a variety of potential digital humanities contexts. Initially developed in the context of the FP9 EU project CENDARI, the software was well received by the user community and continued to be further developed within the H2020 HIRMEOS project where several open access publishers have integrated the service to their collections of published monographs as a means to enhance retrieval and access. entity-fishing implements entity extraction as well as disambiguation against Wikipedia and Wikidata entries. The service is accessible through a REST API which allows easier and seamless integration, language independent and stable convention and a widely used service-oriented architecture (SOA) design. Input and output data are carried out over a query data model with a defined structure providing flexibility to support the processing of partially annotated text or the repartition of text over several queries. The interface implements a variety of functionalities, like language recognition, sentence segmentation and modules for accessing and looking up concepts in the knowledge base. The API itself integrates more advanced contextual parametrisation or ranked outputs, allowing for the resilient integration in various possible use cases. The entity-fishing API has been used as a concrete use case to draft the experimental stand-off proposal, which has been submitted for integration into the TEI guidelines. The representation is also compliant with the Web Annotation Data Model (WADM). In this paper we aim at describing the functionalities of the service as a reference contribution to the subject of web-based NERD services. In this paper, we detail the workflow from input to output and unpack each building box in the processing flow. Besides, with a more academic approach, we provide a transversal schema of the different components taking into account non-functional requirements in order to facilitate the discovery of bottlenecks, hotspots and weaknesses. We also describe the underlying knowledge base, which is set up on the basis of Wikipedia and Wikidata content. We conclude the paper by presenting our solution for the service deployment: how and which the resources where allocated. The service has been in production since Q3 of 2017, and extensively used by the H2020 HIRMEOS partners during the integration with the publishing platforms.

A new topic modeling based approach for aspect extraction in aspect based sentiment analysis: SS-LDA

With the widespread use of social networks, blogs, forums and e-commerce web sites, the volume of user generated textual data is growing exponentially. User opinions in product reviews or in other textual data are crucial for manufacturers, retailers and providers of the products and services. Therefore, sentiment analysis and opinion mining have become important research areas. In user reviews mining, topic modeling based approaches and Latent Dirichlet Allocation (LDA) are significant methods that are used in extracting product aspects in aspect based sentiment analysis. However, LDA cannot be directly applied on user reviews and on other short texts because of data sparsity problem and lack of co-occurrence patterns. Several studies have been published for the adaptation of LDA for short texts. In this study, a novel method for aspect based sentiment analysis, Sentence Segment LDA (SS-LDA) is proposed. SS-LDA is a novel adaptation of LDA algorithm for product aspect extraction. The experimental results reveal that SS-LDA is quite competitive in extracting products aspects.

The impact of differences in text segmentation on the automated quantitative evaluation of song-lyrics.

The text-evaluation application Coh-Metrix and natural language processing rely on the sentence for text segmentation and analysis and frequently detect sentence limits by means of punctuation. Problems arise when target texts such as pop song lyrics do not follow formal standards of written text composition and lack punctuation in the original. In such cases it is common for human transcribers to prepare texts for analysis, often following unspecified or at least unreported rules of text normalization and relying potentially on an assumed shared understanding of the sentence as a text-structural unit. This study investigated whether the use of different transcribers to insert typographical symbols into song lyrics during the pre-processing of textual data can result in significant differences in sentence delineation. Results indicate that different transcribers (following commonly agreed-upon rules of punctuation based on their extensive experience with language and writing as language professionals) can produce differences in sentence segmentation. This has implications for the analysis results for at least some Coh-Metrix measures and highlights the problem of transcription, with potential consequences for quantification at and above sentence level. It is argued that when analyzing non-traditional written texts or transcripts of spoken language it is not possible to assume uniform text interpretation and segmentation during pre-processing. It is advisable to provide clear rules for text normalization at the pre-processing stage, and to make these explicit in documentation and publication.

Translation Process Strategies: Psycholinguistic Aspects

The aim of the paper deals with the preliminary verification of the hypothesis concerning the impact of the source text structure on the choice of translation strategy (form-oriented or sense-oriented) in the process of rendering a text in a native (Ukrainian) into a foreign (English) language by university students majoring in Translation.
 The methods of the research included a comparative analysis of the target texts (English) translated (within a predetermined time limit) from the source text (Ukrainian) related to the domain of economics. The structure of the source text, while remaining grammatically acceptable in the Ukrainian language, had been deliberately made structurally non-congruent with that of the direct word order, which is most frequently used in English. The subjects, the fourth-year BA students majoring in Translation, whose command of English ranged between B2 and C1 levels within the CEFR classification, had been properly motivated to achieve the maximum possible result. The analysis of the target texts was based on a number of parameters, which included the preservation of the source text information and its structure in them.
 Results. It was established that in translating from a native into a foreign language, the subjects have a tendency to replicate the structure of the source text at the levels of clauses and sentences. However, it does not always result in the distortion of the source text sense or/and the violation of the target language norms as the subjects often managed to render the said sense and to keep to the said norms by means of changing the functions of the words in the sentence. The probability of the subjects’ abandoning the source text structure increases when the latter is evidently unsuitable for replication, in which case they switch over to the sense-oriented strategy. The correlation of the two strategies in translating sentence segments is generally identical to the one related to clauses and sentences, while the form-oriented strategy generally prevails in rendering phrases.
 Conclusions. The source text structure has a certain impact on the choice of translation strategy increasing the share of the form-oriented approach. However, this impact is not unequivocal and may depend on a number of factors, which require additional study. The paper outlines the prospects of further research.

A gated piecewise CNN with entity-aware enhancement for distantly supervised relation extraction

The piecewise convolutional neural network (PCNN) is an important method for distant supervision relation extraction. However, the existing methods based on the PCNN still have the following shortcomings: these methods lack the consideration of the impacts of entity pairs and the sentence context on word encoding and do not distinguish the different contributions of the three segments in PCNN to relation classification. To solve these problems, we propose a novel gated piecewise CNN with entity-aware enhancement for distantly supervised relation extraction. First, we use a multi-head self-attention mechanism to combine the word embedding with the head/tail entity embedding and relative position embedding to generate an entity-aware enhanced word representation, which is capable of capturing the semantic dependency between each word and entity pair. Then we introduce a global gate to combine each entity-aware enhanced word representation with their average in the input sentence to form the final word representation of the PCNN input. Moreover, to determine the key segments where the most important information for relation classification appears, we design another gate mechanism to assign a different weight to each sentence segment to highlight the effects of key segments on the PCNN. Experiments on New York Times dataset demonstrate that our model significantly outperforms most of the state-of-the-art models.

Argument corpus development and argument component classification: A study in academic Spanish

Abstract Academic writing is a complex task that requires an author to be skilled in argumentation. The goal of an academic author is to communicate ideas clearly and to convince the reader of the veracity of their claims. However, few students in lower level courses are good arguers, as this is a skill that takes time to master. With the aim of contributing to the development of this skill, we introduce a freely available annotated corpus to support research on argumentation in Spanish. To build the corpus, we developed an annotation guide to identify argumentation in paragraphs. The guide also outlines how to annotate segments of sentences as either claims or premises and how to indicate relations (support or attack) between such components. Then, an annotated corpus of 444 sections was created, and an almost perfect interannotator agreement rate was achieved for a number of its components. After its construction, the corpus was used to perform an exploratory analysis, which aimed to identify the amount of argumentation in each section as well as the patterns commonly used for argument identification. Thereafter, experiments were conducted to automatically classify argumentative components using lexical, semantic, discourse, structural, and sentiment features. A Document Occurrence Representation was also tested on the corpus constituting a novel approach in the representation of documents for automatic argument detection. The results of these experiments to automatically classify argumentative components in Spanish are very promising and seem to support the idea that linguistic feature engineering plays a crucial role in text classification.

Semi-supervised Thai Sentence Segmentation Using Local and Distant Word Representations

A sentence is typically treated as the minimal syntactic unit used for extracting valuable information from a longer piece of text. However, in written Thai, there are no explicit sentence markers. We proposed a deep learning model for the task of sentence segmentation that includes three main contributions. First, we integrate n-gram embedding as a local representation to capture word groups near sentence boundaries. Second, to focus on the keywords of dependent clauses, we combine the model with a distant representation obtained from self-attention modules. Finally, due to the scarcity of labeled data, for which annotation is difficult and time-consuming, we also investigate and adapt Cross-View Training (CVT) as a semi-supervised learning technique, allowing us to utilize unlabeled data to improve the model representations. In the Thai sentence segmentation experiments, our model reduced the relative error by 7.4% and 10.5% compared with the baseline models on the Orchid and UGWC datasets, respectively. We also applied our model to the task of pronunciation recovery on the IWSLT English dataset. Our model outperformed the prior sequence tagging models, achieving a relative error reduction of 2.5%. Ablation studies revealed that utilizing n-gram presentations was the main contributing factor for Thai, while the semi-supervised training helped the most for English.

Fine-Grained Argument Unit Recognition and Classification

Prior work has commonly defined argument retrieval from heterogeneous document collections as a sentence-level classification task. Consequently, argument retrieval suffers both from low recall and from sentence segmentation errors making it difficult for humans and machines to consume the arguments. In this work, we argue that the task should be performed on a more fine-grained level of sequence labeling. For this, we define the task as Argument Unit Recognition and Classification (AURC). We present a dataset of arguments from heterogeneous sources annotated as spans of tokens within a sentence, as well as with a corresponding stance. We show that and how such difficult argument annotations can be effectively collected through crowdsourcing with high inter-annotator agreement. The new benchmark, AURC-8, contains up to 15% more arguments per topic as compared to annotations on the sentence level. We identify a number of methods targeted at AURC sequence labeling, achieving close to human performance on known domains. Further analysis also reveals that, contrary to previous approaches, our methods are more robust against sentence segmentation errors. We publicly release our code and the AURC-8 dataset.1

Aspect-based sentiment analysis of reviews in the domain of higher education

PurposeStudent recruitment and retention are important issues for all higher education institutions. Constant monitoring of student satisfaction levels is therefore crucial. Traditionally, students voice their opinions through official surveys organized by the universities. In addition to that, nowadays, social media and review websites such as “Rate my professors” are rich sources of opinions that should not be ignored. Automated mining of students’ opinions can be realized via aspect-based sentiment analysis (ABSA). ABSA s is a sub-discipline of natural language processing (NLP) that focusses on the identification of sentiments (negative, neutral, positive) and aspects (sentiment targets) in a sentence. The purpose of this paper is to introduce a system for ABSA of free text reviews expressed in student opinion surveys in the Serbian language. Sentiment analysis was carried out at the finest level of text granularity – the level of sentence segment (phrase and clause).Design/methodology/approachThe presented system relies on NLP techniques, machine learning models, rules and dictionaries. The corpora collected and annotated for system development and evaluation comprise students’ reviews of teaching staff at the Faculty of Technical Sciences, University of Novi Sad, Serbia, and a corpus of publicly available reviews from the Serbian equivalent of the “Rate my professors” website.FindingsThe research results indicate that positive sentiment can successfully be identified with the F-measure of 0.83, while negative sentiment can be detected with the F-measure of 0.94. While the F-measure for the aspect’s range is between 0.49 and 0.89, depending on their frequency in the corpus. Furthermore, the authors have concluded that the quality of ABSA depends on the source of the reviews (official students’ surveys vs review websites).Practical implicationsThe system for ABSA presented in this paper could improve the quality of service provided by the Serbian higher education institutions through a more effective search and summary of students’ opinions. For example, a particular educational institution could very easily find out which aspects of their service the students are not satisfied with and to which aspects of their service more attention should be directed.Originality/valueTo the best of the authors’ knowledge, this is the first study of ABSA carried out at the level of sentence segment for the Serbian language. The methodology and findings presented in this paper provide a much-needed bases for further work on sentiment analysis for the Serbian language that is well under-resourced and under-researched in this area.

Recognition and Segmentation of English Long and Short Sentences Based on Machine Translation

With the advent of the information age, long sentences which include many words and have more complex structures.. The translation of long sentences in English-Chinese machine translation has always been the focus of research. In this study, 400 long sentences were randomly selected from NTCIR-9 patent corpus for testing the recognition and segmentation effects of regular match method and error-driven method, and the accuracy rate of the translation was compared on Baidu Online Translation Platform. The results demonstrated that the regular matching method was effective in recognizing and segmenting long sentences, nevertheless there were many defects; the error-driven method was more effective in recognizing and segmenting long sentences; the former increased by 4.8% of the BLEU value of the translated text on Baidu Online Translation Platform and the latter increased by 12.1%, which showed that the error-driven method was more effective in machine translation.

Multi-Information Spatial–Temporal LSTM Fusion Continuous Sign Language Neural Machine Translation

There are two basic problems in sign language recognition (SLR): (a) isolated word SLR and (b) continuous SLR. Most of the existing continuous SLR methods are extensions of the isolated word SLR methods. These methods use the isolated word SLR results as the basic module and obtain the sentence recognition results through sentence segmentation and word alignment. However, sentence segmentation and word alignment are often not accurate, resulting in a low sentence recognition accuracy. At the same time, continuous SLR usually requires strict sample labels, leading to the difficult task of manual labeling and limited training data availability. To address these challenges, this paper proposes a bidirectional spatial–temporal LSTM fusion attention network (Bi-ST-LSTM-A) for continuous SLR. This approach avoids problems such as sentence segmentation, word alignment, and tedious manual labeling. Our contributions are summarized as follows: (1) we proposed a sign language video feature representation method using a convolutional neural network (CNN) and spatial–temporal LSTM (ST-LSTM) information fusion technology; and (2) we constructed a uniform neural machine translation framework that can be used for complex continuous SLR and gesture recognition of nonspecific people in nonspecific environments. Experiments were carried out on some large continuous sign language datasets. The sign language recognition accuracy reached 81.22% on the 500 CSL dataset, 76.12% on the RWTH-PHOENIX-Weather dataset and 75.32% on the RWTH-PHOENIX-Weather-2014T dataset, thereby illustrating the effectiveness of the proposed framework.

Middlebrow Matters: Women’s Reading and the Literary Canon in France since the Belle Époque by Diana Holmes

MLR, .,   (Paris: P.O.L., ), noting how creation is linked to destruction. e second section is entitled ‘La quête de l’autre’. In fact, Liman-Tnani shows how in Le Marin, Duras continues the denunciation of colonialism undertaken in Un barrage contre le Pacifique (Paris: Gallimard, ), resolutely turning her back on the pro-colonialist, propagandist L’Empire français that she had co-written with Philippe Roques in  (Paris: Gallimard). Covering known ground, André Not shows the various forms, including different literary models, that the quest for the ‘sailor’ takes. e next section, ‘La langue du Marin et l’idiolecte durassien’, is the most interesting and original. Sandrine Vaudrey-Luigi demonstrates how Le Marin constituted a space for experiments. In particular, she juxtaposes in an illuminating way two passages, one from Le Marin and one from L’Amant (Paris: Minuit, ), to show that the ‘débordements phrastiques surlignés syntaxiquement’ (p. ), normally thought to have been discovered in the later text, actually date back to the first. She also studies the length and segmentation of sentences, corroborating Duras’s view that for her the choice of words themselves was paramount. In another similarly fascinating article, July examines the locution ‘quand même’ in several of Duras’s novels and in Le Marin, where it is abundant and has replaced the more literary ‘tout de même’, thus underlining the orality of Duras’s style, but also the ambiguity and self-questioning that are at the heart of this novel. e last section, ‘Lecture de l’image et en images’, contains an article by Jean Arrouye which analyses how e Annunciation by Fra Angelico is a generating element of Le Marin. is section also contains short readings by Arrouye of six photographs by Françoise Nuñez, reproduced in the volume and formerly part of an art exhibition inspired by Le Marin. Arrouye establishes connections between the photographs and Le Marin, which he justifies by affirming that Nuñez’s photographs are ‘des matrices narratives ou allégoriques’ (p. ). While some articles go over ideas pointed out by Alazet or other critics such as Cécile Hanania or Julia Waters, this volume is still a worthwhile addition to the small body of research on this intriguing novel. S U C R Middlebrow Matters: Women’s Reading and the Literary Canon in France since the Belle Époque. By D H. Liverpool: Liverpool University Press. .£.  pp. ISBN ––––. Middlebrow Matters is a timely exploration of the intertwining of canonization, social class, and gender in recent French literary-cultural history. Diana Holmes’s book takes as its focus middlebrow literature—literature that sits somewhere between popular and ‘high’ culture—and her approach combines literary-cultural history with detailed literary analysis. Holmes observes twentieth-century trends in France—particularly modernist suspicion of mimesis and immersive narratives— through the lens of gender and, to a lesser extent, that of social class, drawing notably upon Pierre Bourdieu’s theories on culture and social structures (La Distinction , Paris: Minuit, ). She demonstrates how the valuing of complexity and  Reviews (assumed) universality and the concurrent devaluing of popular, easily interpreted narratives that engaged with, for example, the domestic lives of protagonists set in recognizable milieux, underpinned a hierarchization of literary texts that devalorized novels most popular with the (majority female) reading public and acclaimed those that, as Holmes suggests, required a certain type or degree of education to interpret. As Holmes notes early in Middlebrow Matters, there is no direct equivalent of the English term ‘middlebrow’ in French, and recent efforts to revive and reprint middlebrow women authors in Britain have not been matched in France. As well as exploring the possible cultural and structural reasons behind this discrepancy, Middlebrow Matters also performs such revivals through the evocation and close reading of significant—though in many cases underexplored—French middlebrow authors throughout the book. In Chapter  Holmes situates Daniel Lesueur (Jeanne Loiseau) and Marcelle Tinayre as important middlebrow novelists, offering detailed readings of their novels and demonstrating their significance in the literary fabric of the time. Other more critically recognized French women writers who successfully bridged the gap between literary recognition and...

Semantic Conditioned Dynamic Modulation for Temporal Sentence Grounding in Videos.

Temporal sentence grounding in videos aims to localize one target video segment, which semantically corresponds to a given sentence. Unlike previous methods mainly focusing on matching semantics between the sentence and different video segments, in this paper, we propose a novel semantic conditioned dynamic modulation (SCDM) mechanism, which leverages the sentence semantics to modulate the temporal convolution operations for better correlating and composing the sentence-relevant video contents over time. The proposed SCDM also performs dynamically with respect to the diverse video contents so as to establish a precise semantic alignment between sentence and video. By coupling the proposed SCDM with a hierarchical temporal convolutional architecture, video segments with various temporal scales are composed and localized. Besides, more fine-grained clip-level actionness scores are also predicted with the SCDM-coupled temporal convolution on the bottom layer of the overall architecture, which are further used to adjust the temporal boundaries of the localized segments and thereby lead to more accurate grounding results. Experimental results on benchmark datasets demonstrate that the proposed model can improve the temporal grounding accuracy consistently, and further investigation experiments also illustrate the advantages of SCDM on stabilizing the model training and associating relevant video contents for temporal sentence grounding. Our code for this paper is available at https://github.com/yytzsy/SCDM-TPAMI.

Research on partitioning algorithm based on RDF graph

SummaryWith the increasing size of the RDF dataset, the application of parallel computing frameworks is applying more and more widely in big data. In order to realize the parallel query of data, an edge partitioning algorithm is proposed based on RDF sentence graph. Firstly, the algorithm transforms the RDF graph into an RDF sentence graph using coarsening and partitioning. Secondly, the minimum degree vertex partitioning algorithm is put forward to partition the RDF sentence graph. Finally, according to the equivalence thought between edge segmentation of RDF sentence and vertex segmentation of RDF graph, the intersection between RDF subgraphs is the vertex cut set of RDF graph to achieve parallel query of RDF data. The experimental results show that the algorithm's partitioning time and efficiency are better than the traditional algorithms.

Levels of Annotation in the Slovene Training Corpus ssj500k 2.2

Abstract This paper presents the Slovene Training Corpus ssj500k 2.2, which has been annotated on the levels of tokenization, sentence segmentation, part-of-speech tagging, lemmatization, syntactic dependencies, named entities, verbal multi-word expressions, and semantic role labeling. It describes the individual layers of annotation and shows the scope of using the training corpus in the production of various lexicons, such as the lexicon of multi-word units and the valency lexicon of modern Slovene. It concludes by presenting our future work, i.e. the annotation of multi-word expressions based on the Slovene Lexical Database.

Extractive Text Summarization using Deep Natural Language Fuzzy Processing

Text summarization is most trending research areas in a modern context. The main aim of this project is to reduce text size while preserving the information underlying into it. In summary construction level, in general, given complex task which are basically will involve with deep natural language fuzzy processing methodologies. In general, an extractive based summary method is the very simple original text of subset of which will not guarantee as best narrative coherence output, because they are most conveniently representing an approximate summarized content from given text-based only on relevance judgment. In an automatic process of fuzzy summarization which is divided into the following steps: Pre-processing (sentence segmentation, tokenization, stop words removal), Feature Extraction, Sentence Scoring, Sentence Ranking and Summary Extraction.

Ancient Chinese Sentence Segmentation Based on Bidirectional LSTM+CRF Model

This study proposes a novel method for the segmentation of Archaic Chinese sentences based on a bidirectional long short-term memory (LSTM) + conditional random field (CRF) model. The method added a layer of linear statistical model to the traditional bidirectional LSTM neural network; it can be used for sequence annotation from the sentence level. In addition, this model introduced the stochastic gradient descent (SGD) to prevent excessive fitting, and the viterbi algorithm was used to calculate the optimal sequence of the sentences. In the experiment, this study tests the performance of the proposed method using the History of the Han Dynasty, the History of the later Han Dynasty, Three Kingdoms, and the Book of Jin, amongst others. The results show that the precision value, recall value, and F1 value are 0.77, 0.75, and 0.76, respectively, in the open test, and 0.90, 0.88, and 0.76, respectively, in the closed test.

R-Trans: RNN Transformer Network for Chinese Machine Reading Comprehension

Machine reading comprehension (MRC) has gained increasingly wide attention over the past few years. A variety of benchmark datasets have been released, which triggers the development of quite a few MRC approaches based on deep learning techniques. However, most existing models are designed to address English MRC. When applying them directly to Chinese documents, the performance often degrades considerably because of some special characteristics of Chinese, the inevitable segmentation errors in particular. In this paper, we present the RNN Transformer network to tackle the Chinese MRC task. To mitigate the influence of incorrect word segmentation and mine sequential information of whole sentences, deep contextualized word representations and bidirectional gated recurrent units networks are adopted in our model. The extensive experiments have been conducted on a very large scale Chinese MRC corpus, viz., the Les MMRC dataset. The results show that the proposed model outperforms the baseline and other prevalent MRC models notably, and established a new state-of-the-art record on the Les MMRC dataset.

Chunk Meets Image

This study investigated the optimal conditions under which foreign language learning takes place using a smartphone. The authors proposed that two factors, chunking and imagery, would increase language learning. On the basis of previous findings, they formulated three hypotheses: (a) the use of images will have a positive effect on learning English sentences; (b) the use of chunks will have a positive effect on learning English sentences; and (c) the combined use of images and chunks will have a greater positive effect on learning English sentences than either feature alone. A total of 92 Korean seventh graders participated in this study. To examine the learning effect of chunking (i.e., sentence segmentation unit) and imagery (i.e., visual aid) in an experimental setting, they produced a smartphone learning application that incorporated the two methods. The authors measured learning effect with respect to lexical memory retention (i.e., word retrieval ability) and word order composition (i.e., ability to arrange words according to standard English syntax). The results show that the main effects of both chunking and imagery were significant and that the interaction effect between the two on lexical memory retention was also significant. The interaction effect was greater in the delayed effect measurement than in the immediate effect measurement. These findings suggest optimal conditions for designing a smartphone-based, self-learning application.

민주당 대의원제도 개혁사에서 본 2018년 수퍼대의원 제도의 개혁 - 2016년∼2018년 개혁의 내용, 원인, 그리고 함의

Spoken language translation (SLT) has become very important in an increasingly globalized world. Machine translation (MT) for automatic speech recognition (ASR) systems is a major challenge of great interest. This research investigates that automatic sentence segmentation of speech that is important for enriching speech recognition output and for aiding downstream language processing. This article focuses on the automatic sentence segmentation of speech and improving MT results. We explore the problem of identifying sentence boundaries in the transcriptions produced by automatic speech recognition systems in the Polish language. We also experiment with reverse normalization of the recognized speech samples.