scholarly journals Text summarization based on multi-head self-attention mechanism and pointer network

Author(s):  
Dong Qiu ◽  
Bing Yang

AbstractExisting text summarization methods mainly rely on the mapping between manually labeled standard summaries and the original text for feature extraction, often ignoring the internal structure and semantic feature information of the original document. Therefore, the text summary extracted by the existing model has the problems of grammatical structure errors and semantic deviation from the original text. This paper attempts to enhance the model’s attention to the inherent feature information of the source text so that the model can more accurately identify the grammatical structure and semantic information of the document. Therefore, this paper proposes a model based on the multi-head self-attention mechanism and the soft attention mechanism. By introducing an improved multi-head self-attention mechanism in the model coding stage, the training model enables the correct summary syntax and semantic information to obtain higher weight, thereby making the generated summary more coherent and accurate. At the same time, the pointer network model is adopted, and the coverage mechanism is improved to solve out-of-vocabulary and repetitive problems when generating abstracts. This article uses CNN/DailyMail dataset to verify the model proposed in this article and uses the ROUGE indicator to evaluate the model. The experimental results show that the model in this article improves the quality of the generated summary compared with other models.

Information ◽  
2020 ◽  
Vol 11 (2) ◽  
pp. 78 ◽  
Author(s):  
Tulu Tilahun Hailu ◽  
Junqing Yu ◽  
Tessfu Geteye Fantaye

Text summarization is a process of producing a concise version of text (summary) from one or more information sources. If the generated summary preserves meaning of the original text, it will help the users to make fast and effective decision. However, how much meaning of the source text can be preserved is becoming harder to evaluate. The most commonly used automatic evaluation metrics like Recall-Oriented Understudy for Gisting Evaluation (ROUGE) strictly rely on the overlapping n-gram units between reference and candidate summaries, which are not suitable to measure the quality of abstractive summaries. Another major challenge to evaluate text summarization systems is lack of consistent ideal reference summaries. Studies show that human summarizers can produce variable reference summaries of the same source that can significantly affect automatic evaluation metrics scores of summarization systems. Humans are biased to certain situation while producing summary, even the same person perhaps produces substantially different summaries of the same source at different time. This paper proposes a word embedding based automatic text summarization and evaluation framework, which can successfully determine salient top-n sentences of a source text as a reference summary, and evaluate the quality of systems summaries against it. Extensive experimental results demonstrate that the proposed framework is effective and able to outperform several baseline methods with regard to both text summarization systems and automatic evaluation metrics when tested on a publicly available dataset.


Webology ◽  
2021 ◽  
Vol 18 (1) ◽  
pp. 77-91
Author(s):  
J. Shobana ◽  
M. Murali

Nowadays online reviews play an important role by giving an helping hand to the customers to know about other customer’s opinions about the product they are going to purchase. This also guides the organizations as well as government sectors to increase their quality of product and services. So automatic review summarization becomes more important rather than summarizing it manually as it saves time. The aim of this work is to produce a comprehensive summary which includes all key content from the source text. The Proposed Automatic Review Summarization model with improved attention mechanism increases the semantic knowledge and thus improves the summary’s eminence. This encoder-decoder model aims to generate summary in an abstractive way. The Pointer generator mechanism solves the problem of rare words which are out-of-vocabulary and the repetition issues are overcome by coverage mechanism. Experiments were conducted on Amazon’s mobile reviews dataset reveals that the proposed methodology generated more accurate abstractive review summarization when compared with existing techniques. The performance of the summary report is measured using the evaluation metric ROUGE.


2020 ◽  
Vol 10 (16) ◽  
pp. 5422 ◽  
Author(s):  
So-Eon Kim ◽  
Nazira Kaibalina ◽  
Seong-Bae Park

The advent of the sequence-to-sequence model and the attention mechanism has increased the comprehension and readability of automatically generated summaries. However, most previous studies on text summarization have focused on generating or extracting sentences only from an original text, even though every text has a latent topic category. That is, even if a topic category helps improve the summarization quality, there have been no efforts to utilize such information in text summarization. Therefore, this paper proposes a novel topical category-aware neural text summarizer which is differentiated from legacy neural summarizers in that it reflects the topic category of an original text into generating a summary. The proposed summarizer adopts the class activation map (CAM) as topical influence of the words in the original text. Since the CAM excerpts the words relevant to a specific category from the text, it allows the attention mechanism to be influenced by the topic category. As a result, the proposed neural summarizer reflects the topical information of a text as well as the content information into a summary by combining the attention mechanism and CAM. The experiments on The New York Times Annotated Corpus show that the proposed model outperforms the legacy attention-based sequence-to-sequence model, which proves that it is effective at reflecting a topic category into automatic summarization.


2019 ◽  
Vol 25242644 ◽  
pp. 38-41
Author(s):  
Maryna Aleksandrovych

This article contains a summary of some practical issues of translated texts editing with clear examples. Copy editing of translated texts is different from copy editing of texts written in the native language, because the focus of the work shifts from how to deliver an author’s message in the most appropriate way to how to deliver author’s text, written in the native language, in another language (Ukrainian) in the most appropriate way. The first question is the versatility: does the copy editor need to know the language of the original text in order to do effective copy editing. And she/he should at least understand the basic features of the language of the original text such as phonetics, grammar and syntax. Also a copy editor should pay particular attention to such aspects as: at the lexical level – false friends, transliteration of proper nouns, excess of possessive pronouns, translation or adaptation of lexical gaps; at the syntactic level – copulative verb, word order in a sentence, contrastive stress in a phrase, address words, syntax simplification. A necessary aspect is the unification of certain elements in the translated text: address words, units of measurement (length, weight, area, time, volume, etc.), transliterated proper and common nouns. Described in this article principles of transliteration, unification, adaptation, lexical and syntactic aspects of copy editing of translated texts will help to improve the quality of translated books into Ukrainian.


2021 ◽  
pp. 1-12
Author(s):  
Lv YE ◽  
Yue Yang ◽  
Jian-Xu Zeng

The existing recommender system provides personalized recommendation service for users in online shopping, entertainment, and other activities. In order to improve the probability of users accepting the system’s recommendation service, compared with the traditional recommender system, the interpretable recommender system will give the recommendation reasons and results at the same time. In this paper, an interpretable recommendation model based on XGBoost tree is proposed to obtain comprehensible and effective cross features from side information. The results are input into the embedded model based on attention mechanism to capture the invisible interaction among user IDs, item IDs and cross features. The captured interactions are used to predict the match score between the user and the recommended item. Cross-feature attention score is used to generate different recommendation reasons for different user-items.Experimental results show that the proposed algorithm can guarantee the quality of recommendation. The transparency and readability of the recommendation process has been improved by providing reference reasons. This method can help users better understand the recommendation behavior of the system and has certain enlightenment to help the recommender system become more personalized and intelligent.


2020 ◽  
Vol 29 (20) ◽  
pp. 1178-1185
Author(s):  
Hani Hasan ◽  
Sunil Mamtora ◽  
Nimish Shah

The demand for performing intravitreal injections has increased in recent years, prompting the need for more nurse training in their administration. The Great Western Hospitals NHS Trust in Swindon has developed a structured nurse training programme and now has 8 independent nurse injectors trained to undertake injections independently; nurse practitioners now contribute upwards of 85% of the total number of injections. The authors have also demonstrated the financial benefits of using injection assistant devices and shown the positive impact such devices have on training. In September 2019, the authors organised the first course to offer nurses and doctors hands-on experience in administering injections, using the Swindon training model to provide participants with a structured approach to learn how to perform intravitreal injections safely. Nurses made up 96% of participants; the remainder were doctors and managers; 6% had never performed an intravitreal injection; of units where they had, disposable drapes and a speculum were used in 71% of these. The number of injections performed per session at participants' units at the time they attended the course was: 17 or more injections=46%, 13–14=39%, and 11–12=15%. The course was rated 8.9/10 overall for content, with 85% very likely to recommend it to colleagues. All participants indicated that using the Swindon model made them feel confident to deliver injections safely. The authors demonstrated that using a structured training protocol and intravitreal assistant device improves the quality of nurse training and increases confidence in administering intravitreal injections.


2021 ◽  
Vol 3 (2) ◽  
pp. 227-246
Author(s):  
Pin Wang

Abstract This paper analyses and compares the systemic functional features of the Sanskrit original text and the Chinese and English translations of the Buddhist scripture Heart Sutra, focusing on the ideational components that are manifest on the strata of discourse semantics and lexicogrammar. Results show that there are both expected equivalence and significant differences among the Sanskrit original text and the two translated texts. The accounts for the equivalence and differences are twofold (on two hierarchies): in terms of instantiation, the translators go along different re-instantiation routes in finding corresponding potentials between the source text and their respective target texts; in terms of individuation, the English and Chinese translators’ personal and social identity has an immediate influence on their respective reproductions of the text.


2021 ◽  
Vol 14 (1) ◽  
pp. 65-86 ◽  
Author(s):  
Ewa Kujawska-Lis

Heart of Darkness, due to its semantic complexity, interpretative openness and universal thematic interests, has been frequently intersemiotically adapted in a variety of media, encompassing radio broadcast, films, opera, graphic narratives and video games, as well as rewritten in the form of interlingual translations and refracted, with refractions including reviews and critical assessments, but also textual versions radically different from the source text. This article considers selected reinterpretations of Conrad’s text and comments briefly on how in each case the adaptation illustrates a fusion of Conrad’s vision with that of the adapter, hence (trans)fusion, and how this may give a new life to the source text via interpretative shifts. The article presents case studies: the film adaptation – Francis Ford Coppola’s Apocalypse Now (1979), Tarik O’Regan’s one-act chamber opera (both the United Kingdom in 2011 and the US staging in 2015), the graphic narrative by Catherine Anyango and David Zane Mairowitz (2010) and Jacek Dukaj’s Polish language version Serce ciemności (2017). This selection is governed by the variety of media and by the dissimilar approaches of the adapters to their source text. What is evident based on these variants is the role of the adapter as a creative participant in the process of transmitting the ideas of the original text, often updating them to make them relevant to new recipients of various cultural backgrounds. Additionally, reinterpretations and recontextualizations of the novella result directly from adaptive strategies specific to a given medium.


2021 ◽  
Author(s):  
Meiqi Ji ◽  
Ruiling Fu ◽  
Tongtong Xing ◽  
Fulian Yin

2016 ◽  
pp. 91
Author(s):  
Reginaldo Francisco

http://dx.doi.org/10.5007/1980-4237.2014n16p91O teórico e crítico de tradução francês Antoine Berman afirma que as traduções literárias em suas formas tradicionais e dominantes representam um ato culturalmente etnocêntrico, isto é, que traz tudo à sua própria cultura, às suas normas e valores, buscando fazer com que se esqueça que se trata de uma tradução. Para se opor a essa prática dominante, o autor propõe uma tradução que não esconda o elemento estrangeiro na obra traduzida, e que para isso seja fiel à “letra” (lettre) do original. Essa oposição é muito conhecida também nos termos utilizados pelo teórico norte-americano Lawrence Venuti, que fala em “domesticação” (domestication) e “estrangeirização” (foreignization) para se referir respectivamente às práticas tradutórias que ocultam as diferenças culturais, adaptando tudo à cultura de chegada, e àquelas que mantêm a estranheza do texto original e da cultura de partida. Interpretações mais radicais das ideias desses autores podem levar a pensar a tradução como dividida nessas duas possibilidades, e muitas vezes à escolha de uma delas como ideal e a outra como condenável. Entretanto, assim como com dicotomias mais antigas (literal x livre, equivalência formal x equivalência dinâmica, etc.), também estas não são duas categorias estanques, podendo haver diferentes combinações de ambas na tradução de um mesmo texto, além de estratégias híbridas ou soluções que não representam nem uma nem outra posição. Neste trabalho discuto a problematização dessa dicotomia, incluindo exemplos de minha tradução do italiano para o português do livro infantojuvenil O diário de Gian Burrasca, de Luigi Bertelli (Vamba).ABSTRACTFrench translation theorist and critic Antoine Berman states that in their traditional and dominant forms literary translations represent a culturally ethnocentric act, which adapts everything to its own culture, standards and values, seeking to make readers forget that they are reading a translation. To oppose this dominant practice, the author suggests a kind of translation that would not hide the foreign element in the translated work, one that is faithful to the “letter” (lettre) of the original text. A similar opposition to that / to Berman’s is also well-known through the terms “domestication” and “foreignization” as defined by American theorist Lawrence Venuti, who uses them to refer to translation practices that on one hand conceal cultural differences, adapting everything to the target culture, and on the other keep the strangeness of both source text and culture in the translation. Radical interpretations of these authors’ ideas may lead to the misconception that translation is divided into those two possibilities, and often to the judgement that one of them is ideal and the other condemnable. Nevertheless, as with other older dichotomies (literal vs. free translation, formal vs. dynamic equivalence, etc.), these are not clearly distinguishable and opposed categories. There may be different combinations of them in the translation of a text, as well as hybrid strategies or solutions that do not represent either one of them. In this paper I discuss the problems of such dichotomy, drawing examples from my translation of Luigi Bertelli’s book Il giornalino di Gian Burrasca from Italian to Portuguese.Keywords: foreignization; domestication; dichotomy.


Sign in / Sign up

Export Citation Format

Share Document