Academic literature on the topic 'Multi-document summarization'

With the rapid growth of the World Wide Web and online information services, more and more information is available and accessible online. Automatic summarization is an indispensable solution to reduce the information overload problem. Multi-document summarization is useful to provide an overview of a topic and allow users to zoom in for more details on aspects of interest. This paper reports three types of multi-document summaries generated for a set of research abstracts, using different summarization approaches: a sentence-based summary generated by a MEAD summarization system that extracts important sentences using various features, another sentence-based summary generated by extracting research objective sentences, and a variable-based summary focusing on research concepts and relationships. A user evaluation was carried out to compare the three types of summaries. The evaluation results indicated that the majority of users (70%) preferred the variable-based summary, while 55% of the users preferred the research objective summary, and only 25% preferred the MEAD summary.

Publicly available data grows exponentially through web services and technological advancements. To comprehend large data-streams multi-document summarization (MDS) can be used. In this research, the area of multi-document summarization is investigated. Multiple systems for extractive multi-document summarization are implemented using modern techniques, in the form of the pre-trained BERT language model for word embeddings and sentence classification. This is combined with well proven techniques, in the form of the TextRank ranking algorithm, the Waterfall architecture and anti-redundancy filtering. The systems are evaluated on the DUC-2002, 2006 and 2007 datasets using the ROUGE metric. Where the results show that the BM25 sentence representation implemented in the TextRank model using the Waterfall architecture and an anti-redundancy technique outperforms the other implementations, providing competitive results with other state-of-the-art systems. A cohesive model is derived from the leading system and tried in a user study using a real-world application. The user study is conducted using a real-time news detection application with users from the news-domain. The study shows a clear favour for cohesive summaries in the case of extractive multi-document summarization. Where the cohesive summary is preferred in the majority of cases.

There are presently a number of NLP tools available which can provide semantic information about a sentence. Connexor Machinese Semantics is one of the most elaborate of such tools in terms of the information it provides. It has been hypothesized that semantic analysis of sentences is required in order to make significant improvements in automatic summarization. Elaborate semantic analysis is still not particularly feasible. In this thesis, I will look at what shallow semantic features are available from an off the shelf semantic analysis tool which might improve the responsiveness of a summary. The aim of this work is to use the information made available as an intermediary approach to improving the responsiveness of summaries. While this approach is not likely to perform as well as full semantic analysis, it is considerably easier to achieve and could provide an important stepping stone in the direction of deeper semantic analysis. As a significant portion of this task we develop mechanisms in various programming languages to view, process, and extract relevant information and features from the data.

Current web search engines, such as Google, Yahoo!, and Bing, rank the set of documents S retrieved in response to a user query Q and display each document with a title and a snippet, which serves as an abstract of the corresponding document in S. Snippets, however, are not as useful as they are designed for, i.e., to assist search engine users to quickly identify results of interest, if they exist, without browsing through the documents in S, since they (i) often include very similar information and (ii) do not capture the main content of the corresponding documents. Moreover, when the intended information need specified in a search query is ambiguous, it is difficult, if not impossible, for a search engine to identify precisely the set of documents that satisfy the user's intended request. Furthermore, a document title retrieved by web search engines is not always a good indicator of the content of the corresponding document, since it is not always informative. All these design problems can be solved by our proposed query-based, web informative summarization engine, denoted Q-WISE. Q-WISE clusters documents in S, which allows users to view segregated document collections created according to the specific topic covered in each collection, and generates a concise/comprehensive summary for each collection/cluster of documents. Q-WISE is also equipped with a query suggestion module that provides a guide to its users in formulating a keyword query, which facilitates the web search and improves the precision and recall of the search results. Experimental results show that Q-WISE is highly effective and efficient in generating a high quality summary for each cluster of documents on a specific topic, retrieved in response to a Q-WISE user's query. The empirical study also shows that Q-WISE's clustering algorithm is highly accurate, labels generated for the clusters are useful and often reflect the topic of the corresponding clustered documents, and the performance of the query suggestion module of Q-WISE is comparable to commercial web search engines.

In this work we investigate the application of entity type models in extractive multi-document summarization using the automatic caption generation for images of geo-located entities (e.g. Westminster Abbey, Loch Ness, Eiffel Tower) as an application scenario. Entity type models contain sets of patterns aiming to capture the ways the geo-located entities are described in natural language. They are automatically derived from texts about geo-located entities of the same type (e.g. churches, lakes, towers). We collect texts about geo-located entities from Wikipedia because our investigation show that the information humans associate with entity types positively correlates with the information contained in Wikipedia articles about the same entity types. We integrate entity type models into a multi-document summarizer and use them to address the two major tasks in extractive multi-document summarization: sentence scoring and summary composition. We experiment with three different representation methods for entity type models: signature words, n-gram language models and dependency patterns. We first propose that entity type models will improve sentence scoring, i.e. they will help to assign higher scores to sentences which are more relevant to the output summary than to those which are not. Secondly, we claim that summary composition can be improved using entity type models. We follow two different approaches to integrate the entity type models into our multi-document summarizer. In the first approach we use the entity type models in combination with existing standard summarization features to score the sentences. We also manually categorize the set of patterns by the information types they describe and use them to reduce redundancy and to produce better flow within the summary. The second approach aims to eliminate the need for manual intervention and to fully automate the process of summary generation. As in the first approach the sentences are scored using standard summarization features and entity type models. However, unlike the first approach we fully automate the process of summary composition by simultaneously addressing the redundancy and flow aspects of the summary. We evaluate the summarizer with integrated entity type models relative to (1) a summarizer using standard text related features commonly used in summarization and (2) the Wikipedia location descriptions. The latter constitute a strong baseline for automated summaries to be evaluated against. The automated summaries are evaluated against human reference summaries using ROUGE and human readability evaluation, as is a common practice in automatic summarization. Our results show that entity type models significantly improve the quality of output summaries over that of summaries generated using standard summarization features andWikipedia baseline summaries. The representation of entity type models using dependency patterns is superior to the representations using signature words and n-gram language models.

The huge amount of textual data produced everyday by scientists, journalists and Web users, allows investigating many different aspects of information stored in the published documents. Data mining and information retrieval techniques are exploited to manage and extract information from huge amount of unstructured textual data. Text mining also known as text data mining is the processing of extracting high quality information (focusing relevance, novelty and interestingness) from text by identifying patterns etc. Text mining typically involves the process of structuring input text by means of parsing and other linguistic features or sometimes by removing extra data and then finding patterns from structured data. Patterns are then evaluated at last and interpretation of output is performed to accomplish the desired task. Recently, text mining has got attention in several fields such as in security (involves analysis of Internet news), for commercial (for search and indexing purposes) and in academic departments (such as answering query). Beyond searching the documents consisting the words given in a user query, text mining may provide direct answer to user by semantic web for content based (content meaning and its context). It can also act as intelligence analyst and can also be used in some email spam filters for filtering out unwanted material. Text mining usually includes tasks such as clustering, categorization, sentiment analysis, entity recognition, entity relation modeling and document summarization. In particular, summarization approaches are suitable for identifying relevant sentences that describe the main concepts presented in a document dataset. Furthermore, the knowledge existed in the most informative sentences can be employed to improve the understanding of user and/or community interests. Different approaches have been proposed to extract summaries from unstructured text documents. Some of them are based on the statistical analysis of linguistic features by means of supervised machine learning or data mining methods, such as Hidden Markov models, neural networks and Naive Bayes methods. An appealing research field is the extraction of summaries tailored to the major user interests. In this context, the problem of extracting useful information according to domain knowledge related to the user interests is a challenging task. The main topics have been to study and design of novel data representations and data mining algorithms useful for managing and extracting knowledge from unstructured documents. This thesis describes an effort to investigate the application of data mining approaches, firmly established in the subject of transactional data (e.g., frequent itemset mining), to textual documents. Frequent itemset mining is a widely exploratory technique to discover hidden correlations that frequently occur in the source data. Although its application to transactional data is well-established, the usage of frequent itemsets in textual document summarization has never been investigated so far. A work is carried on exploiting frequent itemsets for the purpose of multi-document summarization so a novel multi-document summarizer, namely ItemSum (Itemset-based Summarizer) is presented, that is based on an itemset-based model, i.e., a framework comprise of frequent itemsets, taken out from the document collection. Highly representative and not redundant sentences are selected for generating summary by considering both sentence coverage, with respect to a sentence relevance score, based on tf-idf statistics, and a concise and highly informative itemset-based model. To evaluate the ItemSum performance a suite of experiments on a collection of news articles has been performed. Obtained results show that ItemSum significantly outperforms mostly used previous summarizers in terms of precision, recall, and F-measure. We also validated our approach against a large number of approaches on the DUC’04 document collection. Performance comparisons, in terms of precision, recall, and F-measure, have been performed by means of the ROUGE toolkit. In most cases, ItemSum significantly outperforms the considered competitors. Furthermore, the impact of both the main algorithm parameters and the adopted model coverage strategy on the summarization performance are investigated as well. In some cases, the soundness and readability of the generated summaries are unsatisfactory, because the summaries do not cover in an effective way all the semantically relevant data facets. A step beyond towards the generation of more accurate summaries has been made by semantics-based summarizers. Such approaches combine the use of general-purpose summarization strategies with ad-hoc linguistic analysis. The key idea is to also consider the semantics behind the document content to overcome the limitations of general-purpose strategies in differentiating between sentences based on their actual meaning and context. Most of the previously proposed approaches perform the semantics-based analysis as a preprocessing step that precedes the main summarization process. Therefore, the generated summaries could not entirely reflect the actual meaning and context of the key document sentences. In contrast, we aim at tightly integrating the ontology-based document analysis into the summarization process in order to take the semantic meaning of the document content into account during the sentence evaluation and selection processes. With this in mind, we propose a new multi-document summarizer, namely Yago-based Summarizer, that integrates an established ontology-based entity recognition and disambiguation step. Named Entity Recognition from Yago ontology is being used for the task of text summarization. The Named Entity Recognition (NER) task is concerned with marking occurrences of a specific object being mentioned. These mentions are then classified into a set of predefined categories. Standard categories include “person”, “location”, “geo-political organization”, “facility”, “organization”, and “time”. The use of NER in text summarization improved the summarization process by increasing the rank of informative sentences. To demonstrate the effectiveness of the proposed approach, we compared its performance on the DUC’04 benchmark document collections with that of a large number of state-of-the-art summarizers. Furthermore, we also performed a qualitative evaluation of the soundness and readability of the generated summaries and a comparison with the results that were produced by the most effective summarizers. A parallel effort has been devoted to integrating semantics-based models and the knowledge acquired from social networks into a document summarization model named as SociONewSum. The effort addresses the sentence-based generic multi-document summarization problem, which can be formulated as follows: given a collection of news articles ranging over the same topic, the goal is to extract a concise yet informative summary, which consists of most salient document sentences. An established ontological model has been used to improve summarization performance by integrating a textual entity recognition and disambiguation step. Furthermore, the analysis of the user-generated content coming from Twitter has been exploited to discover current social trends and improve the appealing of the generated summaries. An experimental evaluation of the SociONewSum performance was conducted on real English-written news article collections and Twitter posts. The achieved results demonstrate the effectiveness of the proposed summarizer, in terms of different ROUGE scores, compared to state-of-the-art open source summarizers as well as to a baseline version of the SociONewSum summarizer that does not perform any UGC analysis. Furthermore, the readability of the generated summaries has also been analyzed.

Nonostante lo sviluppo scientifico in molti task di Natural Language Processing (NLP), il task di multi-document summarization rimane meno esplorato, lasciando ampi margini di miglioramento. Nella comunità medica e scientifica, questo task trova applicazione nella generazione delle revisioni sistematiche della letteratura, che sono articoli che riassumono molti studi su uno stesso argomento. In questo lavoro di tesi si esplorano le principali tecnologie in ambito di NLP e text generation, concentrandosi in particolare su BlenderBot 2, un'architettura sviluppata da Facebook AI che rappresenta lo stato dell'arte tra i modelli di chatbot. Le tecnologie studiate sono poi state estese e adattate al task di abstractive multi-document summarization in campo medico per affrontare la generazione automatica delle revisioni sistematiche della letteratura, proponendo un nuovo approccio che miri a risolvere alcuni dei problemi più comuni dei modelli utilizzati in tale ambito.

Made available in DSpace on 2016-06-02T20:25:21Z (GMT). No. of bitstreams: 1 5583.pdf: 2165924 bytes, checksum: 9508776d3397fc5a516393218f88c50f (MD5) Previous issue date: 2013-08-30<br>Universidade Federal de Minas Gerais<br>The multi-document human summarization (MHS), which is the production of a manual summary from a collection of texts from different sources on the same subject, is a little explored linguistic task. Considering the fact that single document summaries comprise information that present recurrent features which are able to reveal summarization strategies, we aimed to investigate multi-document summaries in order to identify MHS strategies. For the identification of MHS strategies, the source texts sentences from the CSTNews corpus (CARDOSO et al., 2011) were manually aligned to their human summaries. The corpus has 50 clusters of news texts and their multi-document summaries in Portuguese. Thus, the alignment revealed the origin of the selected information to compose the summaries. In order to identify whether the selected information show recurrent features, the aligned (and nonaligned) sentences were semi automatically characterized considering a set of linguistic attributes identified in some related works. These attributes translate the content selection strategies from the single document summarization and the clues about MHS. Through the manual analysis of the characterizations of the aligned and non-aligned sentences, we identified that the selected sentences commonly have certain attributes such as sentence location in the text and redundancy. This observation was confirmed by a set of formal rules learned by a Machine Learning (ML) algorithm from the same characterizations. Thus, these rules translate MHS strategies. When the rules were learned and tested in CSTNews by ML, the precision rate was 71.25%. To assess the relevance of the rules, we performed 3 different kinds of intrinsic evaluations: (i) verification of the occurrence of the same strategies in another corpus, and (ii) comparison of the quality of summaries produced by the HMS strategies with the quality of summaries produced by different strategies. Regarding the evaluation (i), which was automatically performed by ML, the rules learned from the CSTNews were tested in a different newspaper corpus and its precision was 70%, which is very close to the precision obtained in the training corpus (CSTNews). Concerning the evaluating (ii), the quality, which was manually evaluated by 10 computational linguists, was considered better than the quality of other summaries. Besides describing features concerning multi-document summaries, this work has the potential to support the multi-document automatic summarization, which may help it to become more linguistically motivated. This task consists of automatically generating multi-document summaries and, therefore, it has been based on the adjustment of strategies identified in single document summarization or only on not confirmed clues about MHS. Based on this work, the automatic process of content selection in multi-document summarization methods may be performed based on strategies systematically identified in MHS.<br>A sumarização humana multidocumento (SHM), que consiste na produção manual de um sumário a partir de uma coleção de textos, provenientes de fontes-distintas, que abordam um mesmo assunto, é uma tarefa linguística até então pouco explorada. Tomando-se como motivação o fato de que sumários monodocumento são compostos por informações que apresentam características recorrentes, a ponto de revelar estratégias de sumarização, objetivou-se investigar sumários multidocumento com o objetivo de identificar estratégias de SHM. Para a identificação das estratégias de SHM, os textos-fonte (isto é, notícias) das 50 coleções do corpus multidocumento em português CSTNews (CARDOSO et al., 2011) foram manualmente alinhados em nível sentencial aos seus respectivos sumários humanos, relevando, assim, a origem das informações selecionadas para compor os sumários. Com o intuito de identificar se as informações selecionadas para compor os sumários apresentam características recorrentes, as sentenças alinhadas (e não-alinhadas) foram caracterizadas de forma semiautomática em função de um conjunto de atributos linguísticos identificados na literatura. Esses atributos traduzem as estratégias de seleção de conteúdo da sumarização monodocumento e os indícios sobre a SHM. Por meio da análise manual das caracterizações das sentenças alinhadas e não-alinhadas, identificou-se que as sentenças selecionadas para compor os sumários multidocumento comumente apresentam certos atributos, como localização das sentenças no texto e redundância. Essa constatação foi confirmada pelo conjunto de regras formais aprendidas por um algoritmo de Aprendizado de Máquina (AM) a partir das mesmas caracterizações. Tais regras traduzem, assim, estratégias de SHM. Quando aprendidas e testadas no CSTNews pelo AM, as regras obtiveram precisão de 71,25%. Para avaliar a pertinência das regras, 2 avaliações intrínsecas foram realizadas, a saber: (i) verificação da ocorrência das estratégias em outro corpus, e (ii) comparação da qualidade de sumários produzidos pelas estratégias de SHM com a qualidade de sumários produzidos por estratégias diferentes. Na avaliação (i), realizada automaticamente por AM, as regras aprendidas a partir do CSTNews foram testadas em um corpus jornalístico distinto e obtiveram a precisão de 70%, muito próxima da obtida no corpus de treinamento (CSTNews). Na avaliação (ii), a qualidade, avaliada de forma manual por 10 linguistas computacionais, foi considerada superior à qualidade dos demais sumários de comparação. Além de descrever características relativas aos sumários multidocumento, este trabalho, uma vez que gera regras formais (ou seja, explícitas e não-ambíguas), tem potencial de subsidiar a Sumarização Automática Multidocumento (SAM), tornando-a mais linguisticamente motivada. A SAM consiste em gerar sumários multidocumento de forma automática e, para tanto, baseava-se na adaptação das estratégias identificadas na sumarização monodocumento ou apenas em indícios, não comprovados sistematicamente, sobre a SHM. Com base neste trabalho, a seleção de conteúdo em métodos de SAM poderá ser feita com base em estratégias identificadas de forma sistemática na SHM.