Sentence Splitting Research Articles

Intelligent Technology Assessment of High-Speed Railway Based on Knowledge Graphs

This paper introduces an intelligent technology assessment framework for high-speed railways based on a knowledge graph approach. We employ rule-based knowledge extraction algorithms and a Bert-BiLSTM-CRF model for entity extraction from technical texts. Subsequently, we establish relationships among various entities, constructing a knowledge graph specific to high-peed railways. The knowledge graph is stored in a Neo4j graph database in triple format. Furthermore, we establish a comprehensive evaluation metric system, integrating knowledge graph insights to assess the utility of enabling technologies. We employ the CRITIC weighting method to calculate the value assessment results for target technologies. Simulation results indicate that the training results for word segmentation and sentence splitting are favorable, and the Bert-BiLSTM-CRF model achieves convergence in accuracy, recall, and F1 score after 30 iterations. The technical assessment results in this paper align closely with the actual technological value assessment.

Nepali ESL/EFL Student Translators' Manipulation of Sentences at the Textual Level

This paper analyzes English translations of Nepali short stories carried out by Nepali ESL/EFL student translators and examines the translators' manipulation of sentences across boundaries. The study adopted a product-oriented framework with a production task to elicit translation from 30 university translation students. The data were analyzed descriptively and discussed under three themes: sentence splitting, sentence joining and sentence-structure preserving. Findings show translators' tendency to preserve source-text sentence boundaries in target texts, with the minimum use of sentence-splitting and sentence-merging strategies to bring about shifts across sentence boundaries.

An improved Bulgarian natural language processing pipeline

In this paper, we present a language pipeline for processing Bulgarian language data. The pipeline consists of the following steps: tokenization, sentence splitting, part-of-speech tagging, dependency parsing, named entity recognition, lemmatization, and word sense disambiguation. The first two components are based on rules and lists of words specific to the Bulgarian language, while the rest of the components use machine learning algorithms trained on universal dependency data and pretrained word vectors. The pipeline is implemented in the Python library spaCy (https://spacy.io/) and achieves significant results on all the included subtasks. The pipeline is open source and is available on Github (https://github.com/melaniab/spacy-pipeline-bg/) for use by researchers and developers for a variety of natural language processing and text analysis tasks.

Does splitting make sentence easier?

In this study, we focus on sentence splitting, a subfield of text simplification, motivated largely by an unproven idea that if you divide a sentence in pieces, it should become easier to understand. Our primary goal in this study is to find out whether this is true. In particular, we ask, does it matter whether we break a sentence into two, three, or more? We report on our findings based on Amazon Mechanical Turk. More specifically, we introduce a Bayesian modeling framework to further investigate to what degree a particular way of splitting the complex sentence affects readability, along with a number of other parameters adopted from diverse perspectives, including clinical linguistics, and cognitive linguistics. The Bayesian modeling experiment provides clear evidence that bisecting the sentence leads to enhanced readability to a degree greater than when we create simplification with more splits.

Sentence splitting in Arabic to Spanish translation

Abstract Modern Standard Arabic makes extensive use of coordination particles whereas punctuation marks are scarce and erratic, leading to long clauses. This is generally assumed to hinder Sentence Boundary Detection and to promote sentence splitting when translating from Arabic into English. Previous literature on translation from Arabic to Spanish is practically inexistent. We have tested this hypothesis regarding translation from Arabic to Spanish on a sample of 282,714 graphic words extracted from a bilingual corpus of 8,681,110 graphic words and found that each Arabic sentence yielded an average of 1.5 Spanish sentences. Furthermore, our data shows the potential impact of directionality in that sentence splitting when translating from Arabic into Spanish is 50% more frequent than from English into Arabic. We also determined statistically that five elements (wa [و], ḥaythu [حيث], kamā [كما], wa-qad [وقد], and wa-dhalika [وذلك]) are the most salient potential markers for sentence splitting in the resulting Spanish translations. Our findings should be particularly interesting for Computational Linguistics and translator training.

Medication event extraction in clinical notes: Contribution of the WisPerMed team to the n2c2 2022 challenge

In this work, we describe the findings of the ‘WisPerMed’ team from their participation in Track 1 (Contextualized Medication Event Extraction) of the n2c2 2022 challenge. We tackle two tasks: (i) medication extraction, which involves extracting all mentions of medications from the clinical notes, and (ii) event classification, which involves classifying the medication mentions based on whether a change in the medication has been discussed.To address the long lengths of clinical texts, which often exceed the maximum token length that models based on the transformer-architecture can handle, various approaches, such as the use of ClinicalBERT with a sliding window approach and Longformer-based models, are employed. In addition, domain adaptation through masked language modeling and preprocessing steps such as sentence splitting are utilized to improve model performance. Since both tasks were treated as named entity recognition (NER) problems, a sanity check was performed in the second release to eliminate possible weaknesses in the medication detection itself. This check used the medication spans to remove false positive predictions and replace missed tokens with the highest softmax probability of the disposition types. The effectiveness of these approaches is evaluated through multiple submissions to the tasks, as well as with post-challenge results, with a focus on the DeBERTa v3 model and its disentangled attention mechanism. Results show that the DeBERTa v3 model performs well in both the NER task and the event classification task.

Radiology Text Analysis System (RadText): Architecture and Evaluation.

Analyzing radiology reports is a time-consuming and error-prone task, which raises the need for an efficient automated radiology report analysis system to alleviate the workloads of radiologists and encourage precise diagnosis. In this work, we present RadText, a high-performance open-source Python radiology text analysis system. RadText offers an easy-to-use text analysis pipeline, including de-identification, section segmentation, sentence split and word tokenization, named entity recognition, parsing, and negation detection. Superior to existing widely used toolkits, RadText features a hybrid text processing schema, supports raw text processing and local processing, which enables higher accuracy, better usability and improved data privacy. RadText adopts BioC as the unified interface, and also standardizes the output into a structured representation that is compatible with Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM), which allows for a more systematic approach to observational research across multiple, disparate data sources. We evaluated RadText on the MIMIC-CXR dataset, with five new disease labels that we annotated for this work. RadText demonstrates highly accurate classification performances, with a 0.91 average precision, 0.94 average recall and 0.92 average F-1 score. We also annotated a test set for the five new disease labels to facilitate future research or applications. We have made our code, documentations, examples and the test set available at https://github.com/bionlplab/radtext.

Contexts and Consequences of Sentence Splitting in Translation (English-French-Czech)

The present paper examines the contexts and consequences of sentence splitting in English, Czech and French translated fiction. In the data extracted from a parallel (multilingual) corpus, we analyze first a language-specific context of sentence splitting (sententialization of non-finite verb forms in translations from English and French into Czech), and second, contexts of splitting occurring in all directions of translation. We conclude that sentence boundaries are usually introduced at the point of a sentence entailing the fewest modifications in the target sentence, especially between two coordinate clauses; and that a systematic sentence splitting, deeply modifying the style of the source text, involves the effect of simplification and normalization.

Evaluation of split-and-rephrase output of the knowledge extraction tool in the intelligent tutoring system

Various approaches to text simplification have been proposed in an attempt to increase text readability. The rephrasing of syntactically and semantically complex structures is still challenging. A pedagogically motivated simplified version of the same text can have both positive and negative side effects. On the one hand, it can facilitate reading comprehension because of much shorter sentences and a limited vocabulary, but on the other hand, the simplified text often lacks coherence, unity and style. Therefore, reasonable trade-offs among linguistic simplicity, naturalness and informativeness are highly needed. This is a survey paper that discusses state-of-the-art approaches to sentence/text simplification and evaluation methods, along with an empirical evaluation of our approach. The quality of sentence splitting, using the knowledge extraction tool SAAT was compared to state-of-the-art syntactic simplification systems. The research was carried out on the WikiSplit, the HSplit and MinWikiSplit simplification corpora. Automatic metrics for the HSplit showed that the SAAT outperformed other TS systems in all categories. For the WikiSplit dataset, automatic metrics scores were slightly lower than that of the baseline system DisSim. However, the human evaluation showed that DisSim outperformed the SAAT in terms of simplicity and grammar. The quality of AG18copy output corresponded to that of the SAAT. The inter-annotator agreement was calculated. Research limitations as well as suggestions for future research were also provided.

Algorithmically Exploiting the Knowledge Accumulated in Textual Domains for Technical Support

Processing natural language and extract relevant information in deep technical engineering domain remains an open challenge. 
 On the other side, manufacturers of high-value assets which often deliver product services through the equipment life, supporting maintenance, spare parts management and remote monitoring and diagnostics for issues resolution, have availability of a good amount of textual data containing technical cases with a certain engineering depth.
 This paper presents a case study in which various Artificial Intelligence algorithms were applied to historical technical cases to extract know-how useful to help technicians in approaching new cases.
 Initially the work process and available data are presented; the focus is on the outbound communication delivered from the technical team to the site operators, that is structured in 3 main paragraphs: event description, technical assessment, recommended actions. 
 The work proceeded in two parallel streams: the first concerned the analysis of event descriptions and technical assessments, aiming to detect recurring topics; the second concerned the analysis of recommended actions that technical support delivered trough years to site operators in order to create a library, which can help for enabling statistical data analysis, quality check review and being the starting point for further AI/NLP developments.
 A text preprocessing was applied to both streams, consisted in defining standard and domain entities / stopwords and identifying / removing them, creating acronyms and synonyms maps in order to make context disambiguation, sentence splitting for the recommended actions, and finally text lemmatization. For every text the output of the preprocess was a series of keywords. 
 Then, unsupervised learning algorithms were applied. For this purpose, firstly, we applied feature extraction, bag of words (TF-IDF) and word embeddings (W2V, D2V, BERT), in order to transform our data from language domain into points in a n-features domain. Afterwards, different combinations of unsupervised algorithms were applied to split data into homogeneous groups, such as: LDA, K-means, Spectral, Affinity Propagation and HDBSCAN. 
 The combinations between language modeling and clustering were evaluated using the Silhouette score and visual analysis. 
 To validate the effectiveness, the developed NLP algorithms have been implemented into the current SW application used by technical support to perform the service. Moreover, a dedicated app to show trending topics and retrieve insightful information has been developed. 
 An outlook of the open technical challenges and on the future perspective of NLP applications in the work process are finally delivered.

Effect of in-app components, medium, and screen size of electronic textbooks on reading performance, behavior, and perception

Although many app-based textbooks are available for students, reading have not been thoroughly outlined. This study aimed to understand how changes from paper to electronic textbooks have affected the academic reading task, investigate student users’ perceptions of in-app components and screen sizes, and identify issues affecting in-app components and task requirements. A mixed factorial design experiment was employed. Results showed that there were no significant changes in comprehension and time spent reading between print text and the iPad. Yet, student highlighting, notetaking, and reading behavior and perception significantly changed based on condition. In addition, students struggled to use in-app components and found them frustrating especially when accounting for sentence splitting. The findings presented can assist in understanding the changes in student reading behavior, which can be used to improve interface design of future e-textbooks.

Fact-Aware Sentence Split and Rephrase with Permutation Invariant Training

Sentence Split and Rephrase aims to break down a complex sentence into several simple sentences with its meaning preserved. Previous studies tend to address the issue by seq2seq learning from parallel sentence pairs, which takes a complex sentence as input and sequentially generates a series of simple sentences. However, the conventional seq2seq learning has two limitations for this task: (1) it does not take into account the facts stated in the long sentence; As a result, the generated simple sentences may miss or inaccurately state the facts in the original sentence. (2) The order variance of the simple sentences to be generated may confuse the seq2seq model during training because the simple sentences derived from the long source sentence could be in any order.To overcome the challenges, we first propose the Fact-aware Sentence Encoding, which enables the model to learn facts from the long sentence and thus improves the precision of sentence split; then we introduce Permutation Invariant Training to alleviate the effects of order variance in seq2seq learning for this task. Experiments on the WebSplit-v1.0 benchmark dataset show that our approaches can largely improve the performance over the previous seq2seq learning approaches. Moreover, an extrinsic evaluation on oie-benchmark verifies the effectiveness of our approaches by an observation that splitting long sentences with our state-of-the-art model as preprocessing is helpful for improving OpenIE performance.

Automated transformation of NL to OCL constraints via SBVR

This paper presents a neoteric method to automatically generate object constraint language (OCL) constraints from natural language (NL) statements. To support the software practitioners in using OCL, we present a novel method. The aim of this method is to produce a framework so that the user of unified modelling language (UML) tool can write constraints and pre/post conditions in natural language like English and the framework converts such natural language expressions to equivalent OCL statements. Here, the state of art of the two well-known technologies, namely open natural language processing (OpenNLP) and Semantics of Business Vocabulary and Rules (SBVR) are used. OpenNLP is used as a pre-processing phase to process the natural language statements. Pre-processing includes sentence splitting, tokenisation and parts of speech (POS) tagging. Then in the second phase, i.e., the transformation phase SBVR is used to automatically transform the preprocessed natural language statements to SBVR specifications. The main aim of this research is to provide automated tool support for model processing tasks in UML models via SBVR to model transform the input SBVR specifications to OCL specifications as explained in model driven architecture (MDA).

A corpus study of splitting and joining sentences in translation

The use of corpora in translation studies has risen dramatically over recent years, contributing towards a growing body of empirical research focussing not only on what differentiates translated from non-translated language, but also on the changes or shifts that translators make from source to target texts. Most of these studies are centred on sub-sentential elements, such as the contrastive use of particular lexis or grammar. However, translation shifts can transcend the level of the sentence. For example, sentences can be split or joined in translation, or there can be complex shifts that combine the two. While there is some research on sentence splitting, there do not seem to be many studies about sentence joining, or indeed sentence splitting and joining together. This study seeks to address this gap. Using a bi-directional parallel corpus of Portuguese and English fiction, over 90,000 source-text sentences and their corresponding text in translation were analysed from a quantitative perspective, and a closer look was taken at a sample of over 1,000 parallel text segments involving sentence joining and splitting. The main findings were that in both translation directions (1) there was a strong tendency for sentence preservation, (2) the differences between sentence splitting and joining were not significant, and (3) changes in sentence boundaries were predominantly associated with the standardisation or normalisation of syntax and a tendency for explicitation.

Effectiveness Level of Online Plagiarism Detection Tools in Arabic

Plagiarism affects education quality, academic research results and publishers reputation. Consequently, many online plagiarism tools have been developed to detect and reduce such affects. However, most of these tools were evaluated according to their abilities to reveal different rates of plagiarism in English text. While evaluating their capability in detecting different plagiarism rates from different patterns in Arabic text is still vague. This paper aims to evaluate the efficiency level of online academic plagiarism detection tools (PlagScan, iThenticate and CheckForPlagiarism.net) in detecting different plagiarism patterns’ amounts in Arabic language. A comparison was made between, PlagScan, iThenticate and CheckForPlagiarism.net, detection capabilities by merging university theses and dissertations with eight plagiarism patterns (whole document, some parts, insertion, sentence split or join, phrase reordering, syntax, lexical and morpho-syntactic) with the ratio between 90% , 30% and 10% respectively. Experiment’s results showed that iThenticate is the most efficient online plagiarism detection tool in Arabic for eight plagiarism patterns between 90% and 80% ratio Arabic language. While none of the three online plagiarism detection tools are efficient for less than 80% plagiarized text from any of the eight plagiarism patterns. Hence, mechanism enhancements and consideration to the Arabic anguage structure are recommended for online plagiarism detection tool in Arabic.

Web Content Mining of Hepatitis-C Community Forums for Identification of Most Frequently Transpiring Lexis

This paper presents the identification of most frequently emerging tokens in the online discussions in community forums on Hepatitis-C, a common disease in many parts of the world. Technically, this has been accomplished by populating the text corpus using the online Web text content. More than 10,000 topics in 2300 posts, collected over a period of seven months, were used for the research purpose. The vector space document model through Bag-of-Words for syntactic parsing and tokenisation or sentence splitting was implemented. The initial number of the lexis was nearly 11,40,000 while removal of noise, stop-words and duplication yielded nearly 15,000 entries. The paper presents the top 100 most frequently transpiring lexis and establishes that the course of treatment, symptoms of the disease and effectiveness of the medicines are key three areas of discussion in the peer online community forum of patients. The results presented here are for Hepatitis-C forum but hold true for other similar diseases as well. To the best of available literature, the paper is first formal attempt to study and analyse the online behaviour of patients.

Minimum redundancy and maximum relevance for single and multi-document Arabic text summarization

Automatic text summarization aims to produce summaries for one or more texts using machine techniques. In this paper, we propose a novel statistical summarization system for Arabic texts. Our system uses a clustering algorithm and an adapted discriminant analysis method: mRMR (minimum redundancy and maximum relevance) to score terms. Through mRMR analysis, terms are ranked according to their discriminant and coverage power. Second, we propose a novel sentence extraction algorithm which selects sentences with top ranked terms and maximum diversity. Our system uses minimal language-dependant processing: sentence splitting, tokenization and root extraction. Experimental results on EASC and TAC 2011 MultiLingual datasets showed that our proposed approach is competitive to the state of the art systems.

‘Lösen Sie Schachtelsätze möglichst auf’: The Impact of Editorial Guidelines on Sentence Splitting in German Business Article Translations

Sentence splitting is assumed to occur mainly in translations from languages that prefer a hierarchical discourse structure, such as German, to languages that prefer an incremental structure. This article challenges that assumption by presenting findings from a diachronic corpus study of English–German business article translations, which shows that sentence splitting has long been common in German business translation, and perhaps because of editorial guidelines, has increased strongly over the course of the 25 years under analysis. A corpus of unedited draft versions and published versions of the same translations is used to show that sentence splitting is also affected by editors where translators did not split the sentence. The evidence suggests that sentence splitting may be a strategy of explicitation in translation rather than a phenomenon triggered mainly in translation into languages with incremental discourse structures. The observed increase in sentence splitting in German may indicate a shift by which meaning relations are increasingly made between sentences using cohesive resources of reference rather than within sentences using grammatical devices such as hypotaxis or parataxis.

A modular framework for biomedical concept recognition

BackgroundConcept recognition is an essential task in biomedical information extraction, presenting several complex and unsolved challenges. The development of such solutions is typically performed in an ad-hoc manner or using general information extraction frameworks, which are not optimized for the biomedical domain and normally require the integration of complex external libraries and/or the development of custom tools.ResultsThis article presents Neji, an open source framework optimized for biomedical concept recognition built around four key characteristics: modularity, scalability, speed, and usability. It integrates modules for biomedical natural language processing, such as sentence splitting, tokenization, lemmatization, part-of-speech tagging, chunking and dependency parsing. Concept recognition is provided through dictionary matching and machine learning with normalization methods. Neji also integrates an innovative concept tree implementation, supporting overlapped concept names and respective disambiguation techniques. The most popular input and output formats, namely Pubmed XML, IeXML, CoNLL and A1, are also supported. On top of the built-in functionalities, developers and researchers can implement new processing modules or pipelines, or use the provided command-line interface tool to build their own solutions, applying the most appropriate techniques to identify heterogeneous biomedical concepts. Neji was evaluated against three gold standard corpora with heterogeneous biomedical concepts (CRAFT, AnEM and NCBI disease corpus), achieving high performance results on named entity recognition (F1-measure for overlap matching: species 95%, cell 92%, cellular components 83%, gene and proteins 76%, chemicals 65%, biological processes and molecular functions 63%, disorders 85%, and anatomical entities 82%) and on entity normalization (F1-measure for overlap name matching and correct identifier included in the returned list of identifiers: species 88%, cell 71%, cellular components 72%, gene and proteins 64%, chemicals 53%, and biological processes and molecular functions 40%). Neji provides fast and multi-threaded data processing, annotating up to 1200 sentences/second when using dictionary-based concept identification.ConclusionsConsidering the provided features and underlying characteristics, we believe that Neji is an important contribution to the biomedical community, streamlining the development of complex concept recognition solutions. Neji is freely available at http://bioinformatics.ua.pt/neji.

Textrous!: Extracting Semantic Textual Meaning from Gene Sets

The un-biased and reproducible interpretation of high-content gene sets from large-scale genomic experiments is crucial to the understanding of biological themes, validation of experimental data, and the eventual development of plans for future experimentation. To derive biomedically-relevant information from simple gene lists, a mathematical association to scientific language and meaningful words or sentences is crucial. Unfortunately, existing software for deriving meaningful and easily-appreciable scientific textual ‘tokens’ from large gene sets either rely on controlled vocabularies (Medical Subject Headings, Gene Ontology, BioCarta) or employ Boolean text searching and co-occurrence models that are incapable of detecting indirect links in the literature. As an improvement to existing web-based informatic tools, we have developed Textrous!, a web-based framework for the extraction of biomedical semantic meaning from a given input gene set of arbitrary length. Textrous! employs natural language processing techniques, including latent semantic indexing (LSI), sentence splitting, word tokenization, parts-of-speech tagging, and noun-phrase chunking, to mine MEDLINE abstracts, PubMed Central articles, articles from the Online Mendelian Inheritance in Man (OMIM), and Mammalian Phenotype annotation obtained from Jackson Laboratories. Textrous! has the ability to generate meaningful output data with even very small input datasets, using two different text extraction methodologies (collective and individual) for the selecting, ranking, clustering, and visualization of English words obtained from the user data. Textrous!, therefore, is able to facilitate the output of quantitatively significant and easily appreciable semantic words and phrases linked to both individual gene and batch genomic data.