Citation Sentence Research Articles

Generating Paraphrase Using Simulated Annealing for Citation Sentences

The paraphrase generator for citation sentences is used to produce several sentence alternatives to avoid plagiarism. Furthermore, the generation results need to pay attention to semantic similarity and lexical divergence standards. This study proposed the StoPGEN model as an algorithm for generating citation paraphrase sentences with stochastic output. The generation process is guided by an objective function using a simulated annealing algorithm to maintain the properties of semantic similarity and lexical divergence. The objective function is created by combining the two factors that maintain these properties. This study combined METEOR and PINC Scores in a linear weighting function that can be adjusted for its value tendency in one of the matrix functions. The dataset of citation sentences that had been labeled with paraphrases was used to test StoPGEN and other models for comparison. The StoPGEN model, with the citation sentences dataset, produced a BLEU score of 55.37, outperforming the bidirectional LSTM method with a value of 28.93. StoPGEN was also tested using Quora data by changing the language source in the architecture section resulting in a BLEU score of 22.37, outperforming UPSA 18.21. In addition, the qualitative evaluation results of the citation sentence generation based on respondents obtained an acceptance value of 50.80.

Heterogeneous deep graph convolutional network with citation relational BERT for COVID-19 inline citation recommendation

The outbreak of COVID-19 brings almost the biggest explosions of scientific literature ever. Facing such volume literature, it is hard for researches to find desired citation when carrying out COVID-19 related research, especially for junior researchers. This paper presents a novel neural network based method, called citation relational BERT with heterogeneous deep graph convolutional network (CRB-HDGCN), for COVID-19 inline citation recommendation task. The CRB-HDGCN contains two main stages. The first stage is to enhance the representation learning of BERT model for COVID-19 inline citation recommendation task through CRB. To achieve the above goal, an augmented citation sentence corpus, which replaces the citation placeholder with the title of the cited papers, is used to lightly retrain BERT model. In addition, we extract three types of sentence pair according citation relation, and establish sentence prediction tasks to further fine-tune the BERT model. The second stage is to learn effective dense vector of nodes among COVID-19 bibliographic graph through HDGCN. The HDGCN contains four layers which are essentially all sub neural networks. The first layer is initial embedding layer which generates initial input vectors with fixed size through CRB and a multilayer perceptron. The second layer is a heterogeneous graph convolutional layer. In this layer, we expand traditional homogeneous graph convolutional network into heterogeneous by subtly adding heterogeneous nodes and relations. The third layer is a deep attention layer. This layer uses trainable project vectors to reweight the node importance simultaneously according to both node types and convolution layers, which further promotes the performance of learnt node vectors. The last decoder layer recovers the graph structure and let the whole network trainable. The recommendation is finally achieved by integrating the high performance heterogeneous vectors learnt from CRB-HDGCN with the query vectors. We conduct experiments on the CORD-19 and LitCovid datasets. The results show that compared with the second best method CO-Search, CRB-HDGCN improves MAP, MRR, P@100 and R@100 with 21.8%, 22.7%, 37.6% and 21.2% on CORD-19, and 29.1%, 25.9%, 15.3% and 11.3% on LitCovid, respectively.

An NLP-based citation reason analysis using CCRO

In recent scientific advances, Artificial Intelligence and Natural Language Processing are the major contributors to classifying documents and extracting information. Classifying citations in different classes have gathered a lot of attention due to the large volume of citations available in different digital libraries. Typical citation classification uses sentiment analysis, where various techniques are applied to citations texts to mainly classify them in “Positive”, “Negative” and “Neutral” sentiments. However, there can be innumerable reasons why an author selects another research for citation. Citations’ Context and Reasons Ontology—CCRO uses a clear scientific method to articulate eight basic reasons for citing by using an iterative process of sentiment analysis, collaborative meanings, and experts' opinions. Using CCRO, this research paper adopts an ontology-based approach to extract citation's reasons and instantiate ontology classes and properties on two different corpora of citation sentences. One corpus of citation sentences is a publicly available dataset, while the other is our own manually curated. The process uses a two-step approach. The first part is an interface to manually annotate each citation text in the selected corpora on CCRO properties. A team of carefully selected annotators has annotated each citation to achieve a high inter-annotator agreement. The second part focuses on the automatic extraction of these reasons. Using Natural Language Processing, Mapping Graph, and Reporting Verb in a citation sentence, citation's reason is extracted and mapped onto a CCRO property. After comparing both manual and automatic mapping, accuracy is calculated. Based on experiments and results, accuracy is calculated for both publicly available and own corpora of citation sentences.

An analysis of in-text citations based on fractional counting

• The majority of in-text citations were independent. • The majority of references that had no independent mentions were mentioned only once. • Approximately 20 % of the references did not independently contribute to the citing paper. • Most of the multiple mentioned references had high mention frequencies according to two counting methods. With the development of citation analysis, the analysis of in-text citations is getting more important. There can be many references in the bibliography of a paper, but the way that each reference is mentioned within the full text of a paper is different. Some references are mentioned together with other references, and some references are mentioned alone. That is, a citation sentence can include only one reference or several references. However, the citation sentence gives readers a description. Thus, it is necessary to examine in-text citations by considering the way that each reference is mentioned within the full text. From this point of view, we introduce two counting methods (full counting and fractional counting) to examine in-text citations and compare the two counting methods. The number of in-text citations according to full counting was approximately 1.448 times larger than that according to fractional counting. The results show that the majority of in-text citations are independent, and the majority of references that have no independent mentions are mentioned only once. The results also show that most of the multiple mentioned references have high mention frequencies according to both full counting and fractional counting.

Correction to: ADAM-17 is expressed in the inflammatory myopathy and is involved with interstitial lung disease.

The original version of this article, unfortunately, contained errors. Figure citation, caption, image and updated sentence in the Result section are now presented correctly in this article.

Automatic sentence extraction for the detection of scientific paper relations

The relations between scientific papers are very useful for researchers to see the interconnection between scientific papers quickly. By observing the inter-article relationships, researchers can identify, among others, the weaknesses of existing research, performance improvements achieved to date, and tools or data typically used in research in specific fields. So far, methods that have been developed to detect paper relations include machine learning and rule-based methods. However, a problem still arises in the process of sentence extraction from scientific paper documents, which is still done manually. This manual process causes the detection of scientific paper relations longer and inefficient. To overcome this problem, this study performs an automatic sentences extraction while the paper relations are identified based on the citation sentence. The performance of the built system is then compared with that of the manual extraction system. The analysis results suggested that the automatic sentence extraction indicates a very high level of performance in the detection of paper relations, which is close to that of manual sentence extraction.

A content-based citation analysis study based on text categorization

Publications and citations are important components for measuring research performance. Academics receive incentives, tenures, or awards from the number of citations they receive; however, the use of citations for research/er evaluation purposes can give rise to unethical practices and manipulation. Consequently, it is necessary to change the current approach to the use of citations. The main aim of this study was to conduct a content-based citation analysis study for Turkish citations. To achieve this aim, 423 peer-reviewed articles, the associated 12,881 references, and 101,019 sentences published in library and information science literature in Turkey were thoroughly examined. The citations were divided into four main categories; citation meaning, citation purpose, citation shape, and citation array. Then, each category was further divided into sub-categories. A tagging process with inter-annotator agreement was conducted and citation categories for the citation sentences determined. Weka software was used to apply the text categorization methods. The automatic citation sentence classification achieved at least a 90% success rate for all citation classes, which proved that using computational linguistics to evaluate citation contexts developing new techniques was possible and gave more detailed results.

어휘 자질 기반 기계 학습을 사용한 한국어 암묵 인용문 인식

Implicit citation sentence recognition is to locate citation sentences which lacks explicit citation markers, from articles' full-text. State-of-the-art approaches exploit word ngrams, clue words, researcher's surnames, mentions of previous methods, and distance relative to nearest explicit citation sentences, etc., reaching over 50% performance. However, most previous works have been conducted on English. As for Korean, a rule-based method using positive/negative clue patterns was reported to attain the performance of 42%, requiring further improvement. This study attempted to learn to recognize implicit citation sentences from Korean literatures' full-text using Korean lexical features. Different lexical feature units such as Eojeol, morpheme, and Eumjeol were evaluated to determine proper lexical features for Korean implicit citation sentence recognition. In addition, lexical features were combined with the position features representing backward/forward proximities to explicit citation sentences, improving the performance up to over 50%.

The Searchbench - Combining Sentence-semantic, Full-text and Bibliographic Search in Digital Libraries

We describe a novel approach to precise searching in the full content of digital libraries. The Searchbench (for search workbench) is based on sentence-wise syntactic and semantic natural language processing (NLP) of both born-digital and scanned publications in PDF format. The term born-digital means natively digital, i.e. prepared electronically using typesetting systems such as LaTeX, OpenOffice, and the like. In the Searchbench, queries can be formulated as (possibly underspecified) statements, consisting of simple subject-predicate-object constructs such as ‘algorithm improves word alignment’. This reduces the number of false hits in large document collections when the search words happen to appear close to each other, but are not semantically related. The method also abstracts from passive voice and predicate synonyms. Moreover, negated statements can be excluded from the search results, and negated antonym predicates again count as synonyms (e.g. not include = exclude).In the Searchbench, a sentence-semantic search can be combined with search filters for classical full-text, bibliographic metadata and automatically computed domain terms. Auto-suggest fields facilitate text input. Queries can be bookmarked or emailed. Furthermore, a novel citation browser in the Searchbench allows graphical navigation in citation networks. These have been extracted automatically from metadata and paper texts. The citation browser displays short phrases from citation sentences at the edges in the citation graph and thus allows students and researchers to quickly browse publications and immerse into a new research field. By clicking on a citation edge, the original citation sentence is shown in context, and optionally also in the original PDF layout.To showcase the usefulness of our research, we have a applied it to a collection of currently approx. 25,000 open access research papers in the field of computational linguistics and language technology, the ACL Anthology ( http://aclweb.org/anthology). The Searchbench user interface is a web application running in every modern, JavaScript-enabled web browser, also on smart phones and tablet computers. The system is a free and public service at http://aclasb.dfki.de. Because the NLP technology is domain-independent, it could also be applied to newspaper texts, technical documentation, or scientific publications from other disciplines. The aim of this paper is to make the benefits of this new, language technology based approach known in library research and related fields.This article summarises 9 peer reviewed publications from the past three years that have been published in international conferences and workshops in the area of computational linguistics, and tries to present them in an appropriate way to the LIBER audience. The original papers contain more details and are freely available from the author’s homepage[1] or via the Searchbench[2].

On what it takes to predict perceptual difficulty in cross-language vowel perception

According to current models of cross-language speech perception and second language (L2) speech learning, perceptual difficulty depends upon how listeners map the phonetic segments of the foreign language onto their native language (L1) categories. A number of indirect methods have been suggested for predicting perceptual difficulty, e.g., contrastive phonemic analyses and acoustic comparisons of L1 and L2 sounds. This presentation summarizes recent studies which directly assessed perceptual assimilation patterns for North German vowels (as perceived by American English listeners) and for British English vowels (as perceived by Danish listeners). Vowels were produced in CVC syllables in various consonantal contexts in both citation form and sentence frames. As expected, contrastive phonemic analyses were not good predictors of assimilation patterns. Somewhat surprisingly, acoustic comparisons didn’t always predict assimilation patterns either. In particular, measures of acoustic similarity from ‘‘canonical’’ forms didn’t predict acoustic similarity for coarticulated vowels, nor did these measures predict perceptual similarity. Our results strongly suggest that perceptual difficulty in cross-language speech perception cannot be predicted using indirect methods, nor can difficulty be predicted by perceptual studies of citation form utterances. Instead, successful predictions of perceptual difficulty must be based on direct assessments of assimilation patterns. [Work supported by NIDCD.]

Effects of speech style on the perceptual assimilation of American English vowels by Japanese speakers.

Of specific interest in this study was the extent to which speech style (citation form versus sentence form) affected perceptual assimilation of AE vowels by Japanese listeners. Four adult male speakers produced 11 AE vowels in /hVba/ bisyllables in both citation form (lists) and in the sentence ‘‘I hear the /hVb/ on the tape.’’ Response categories, designated in Katakana symbols, included five short vowels, five long vowels, and eight vowel combinations (e.g., ou, ei, ja:). Twenty‐four Japanese listeners selected the response category which was most similar to the /hV/ syllable they heard and rated its ‘‘goodness of fit.’’ Results indicated that, in general, AE vowels were assimilated to the Japanese vowel category closest in articulatory vowel space. However, intrinsic duration information was utilized much more consistently when the syllables were embedded in a carrier sentence. That is, intrinsically long (or diphthongized) AE vowels were assimilated to long‐vowel or two‐vowel response categories more often in the sentence condition. Acoustical analysis indicated that some, but not all, of the variation in assimilation patterns could be explained by F1/F2 variation. Implications for Japanese L2 learners of English are discussed. [Work supported by NIDCD.]

The relation between the middle tone and empty category principle violations in Krio

An examination of monosyllabic words in Krio reveals the existence of three distinct tones: high. mid and low. words of more than one syllable. However, only two tones are attested: high and low. The mid tone is, in fact, found only in monosyllabic forms in two very specific contexts: citation forms and sentence final forms. In other contexts. the mid tone ia replaced by the high tone. This suggests that the mid tone is merely a phonetic variant of the high tone. The interaction of tones and syntax is considered.