Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ uBibliorum Repositor...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
versions View all 1 versions
addClaim

Automatic Text Summarization

Authors: Saraiva, João Pedro Abrantes;

Automatic Text Summarization

Abstract

O texto é um dos utensílios mais importantes de transmissão de ideias entre os seres humanos. Pode ser de vários tipos e o seu conteúdo pode ser mais ou menos fácil de interpretar, conforme a quantidade de informação relevante sobre o assunto principal. De forma a facilitar o processamento pelo leitor existe um mecanismo propositadamente criado para reduzir a informação irrelevante num texto, chamado sumarização de texto. Através da sumarização criam­se versões reduzidas do text original e mantém­se a informação do assunto principal. Devido à criação e evolução da Internet e outros meios de comunicação, surgiu um aumento exponencial de documentos textuais, evento denominado de sobrecarga de informação, que têm na sua maioria informação desnecessária sobre o assunto que retratam. De forma a resolver este problema global, surgiu dentro da área científica de Processamento de Linguagem Natural, a sumarização automática de texto, que permite criar sumários automáticos de qualquer tipo de texto e de qualquer lingua, através de algoritmos computacionais. Desde a sua criação, inúmeras técnicas de sumarização de texto foram idealizadas, podendo ser classificadas em dois tipos diferentes: extractivas e abstractivas. Em técnicas extractivas, são transcritos elementos do texto original, como palavras ou frases inteiras que sejam as mais ilustrativas do assunto do texto e combinadas num documento. Em técnicas abstractivas, os algoritmos geram elementos novos. Nesta dissertação pesquisaram­se, implementaram­se e combinaram­se algumas das técnicas com melhores resultados de modo a criar um sistema completo para criar sumários. Relativamente às técnicas implementadas, as primeiras três são técnicas extractivas enquanto que a ultima é abstractiva. Desta forma, a primeira incide sobre o cálculo das frequências dos elementos do texto, atribuindo­se valores às frases que sejam mais frequentes, que por sua vez são escolhidas para o sumário através de uma taxa de compressão. Outra das técnicas incide na representação dos elementos textuais sob a forma de nodos de um grafo, sendo atribuidos valores de similaridade entre os mesmos e de seguida escolhidas as frases com maiores valores através de uma taxa de compressão. Uma outra abordagem foi criada de forma a combinar um mecanismo de análise das caracteristicas do texto com métodos baseados em inteligência artificial. Nela cada frase possui um conjunto de caracteristicas que são usadas para treinar um modelo de rede neuronal. O modelo avalia e decide quais as frases que devem pertencer ao sumário e filtra as mesmas através deu uma taxa de compressão. Um sumarizador abstractivo foi criado para para gerar palavras sobre o assunto do texto e combinar num sumário. Cada um destes sumarizadores foi combinado num só sistema. Por fim, cada uma das técnicas pode ser avaliada segundo várias métricas de avaliação, como por exemplo a ROUGE. Segundo os resultados de avaliação das técnicas, com o conjunto de dados DUC, os nossos sumarizadores obtiveram resultados relativamente parecidos com os presentes na comunidade cientifica, com especial atenção para o codificador­descodificador que em certos casos apresentou resultados promissores.

Writing text was one of the first ever methods used by humans to represent their knowledge. Text can be of different types and have different purposes. Due to the evolution of information systems and the Internet, the amount of textual information available has increased exponentially in a worldwide scale, and many documents tend to have a percentage of unnecessary information. Due to this event, most readers have difficulty in digesting all the extensive information contained in multiple documents, produced on a daily basis. A simple solution to the excessive irrelevant information in texts is to create summaries, in which we keep the subject’s related parts and remove the unnecessary ones. In Natural Language Processing, the goal of automatic text summarization is to create systems that process text and keep only the most important data. Since its creation several approaches have been designed to create better text summaries, which can be divided in two separate groups: extractive approaches and abstractive approaches. In the first group, the summarizers decide what text elements should be in the summary. The criteria by which they are selected is diverse. After they are selected, they are combined into the summary. In the second group, the text elements are generated from scratch. Abstractive summarizers are much more complex so they still need a lot of research, in order to represent good results. During this thesis, we have investigated the state of the art approaches, implemented our own versions and tested them in conventional datasets, like the DUC dataset. Our first approach was a frequency­based approach, since it analyses the frequency in which the text’s words/sentences appear in the text. Higher frequency words/sentences automatically receive higher scores which are then filtered with a compression rate and combined in a summary. Moving on to our second approach, we have improved the original TextRank algorithm by combining it with word embedding vectors. The goal was to represent the text’s sentences as nodes from a graph and with the help of word embeddings, determine how similar are pairs of sentences and rank them by their similarity scores. The highest ranking sentences were filtered with a compression rate and picked for the summary. In the third approach, we combined feature analysis with deep learning. By analysing certain characteristics of the text sentences, one can assign scores that represent the importance of a given sentence for the summary. With these computed values, we have created a dataset for training a deep neural network that is capable of deciding if a certain sentence must be or not in the summary. An abstractive encoder­decoder summarizer was created with the purpose of generating words related to the document subject and combining them into a summary. Finally, every single summarizer was combined into a full system. Each one of our approaches was evaluated with several evaluation metrics, such as ROUGE. We used the DUC dataset for this purpose and the results were fairly similar to the ones in the scientific community. As for our encoder­decode, we got promising results.

Country
Portugal
Related Organizations
Keywords

Automatic Text Summarization, Eletrónica e Informática, Abstractive Summarization:deep Learning, Extractive Summarization, Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Natural Language Processing

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
    OpenAIRE UsageCounts
    Usage byUsageCounts
    visibility views 76
    download downloads 50
  • 76
    views
    50
    downloads
    Powered byOpenAIRE UsageCounts
Powered by OpenAIRE graph
Found an issue? Give us feedback
visibility
download
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
views
OpenAIRE UsageCountsViews provided by UsageCounts
downloads
OpenAIRE UsageCountsDownloads provided by UsageCounts
0
Average
Average
Average
76
50
Green
Related to Research communities