Developments in deep learning and machine learning overall has created a plethora of opportunities for easier training of automatic text summarization (ATS) models for producing summaries with higher quality. ATS can be split into extractive and abstractive tasks; extractive models extract sentences from the original text to create summaries. On the contrary, abstractive models generate novel sentences to create summaries. While extractive summaries are often preferred over abstractive ones, summaries created by extractive models trained on Swedish texts often lack cohesion, which affects the readability and overall quality of the summary. Therefore, there is a need to improve the process of training ATS models in terms of cohesion, while maintaining other text qualities such as content coverage. This thesis explores and implements methods at the data-processing stage aimed at improving cohesion of generated summaries. The methods are based around Sentence-BERT for creating advanced sentence embeddings that can be used to rank sentences in a text in terms of if it should be included in the extractive summary or not. Three models are trained using different methods and evaluated using ROUGE, BERTScore for measuring content coverage and Coh-Metrix for measuring cohesion. The results of the evaluation suggest that the methods can indeed be used to create more cohesive summaries, although content coverage was reduced, which gives rise to the potential for extensive future exploration of further implementation.

Background: Communication plays a crucial role in supply chain management (SCM) as it facilitates the flow of information, materials, and goods across various stages of the supply chain. In the context of supply planning, each planner manages thousands of supply chain entities and spends a lot of time reading and responding to high volumes of emails related to part orders, delays, and backorders that can lead to information overload and hinder workflow and decision-making. Therefore, streamlining communication and enhancing email management are essential for optimizing supply chain efficiency. Objectives: This study aims to create an automated system that can summarize email conversations between planners, suppliers, and other stakeholders. The goal is to increase communication efficiency using Natural Language Processing (NLP) algorithms to extract important information from lengthy conversations. Additionally, the study will explore the effectiveness of using conditional random fields (CRF) to filter out irrelevant content during preprocessing. Methods: We chose four advanced pre-trained abstractive dialogue summarization models, BART, PEGASUS, T5, and CODS, and evaluation metrics, ROUGE and BERTScore, to compare their performance in effectively summarizing our email conversations. We used CRF to preprocess raw data from around 400 planner-supplier email conversations to extract important sentences in a dialogue format and label them with specific dialogue act tags. We then manually summarized the 400 conversations and fine-tuned the four chosen models. Finally, we evaluated the models using ROUGE and BERTScore metrics to determine their similarity to human references. Results: The results show that the performance of the summarization models has significantly improved after fine-tuning the models with domain-specific data. The BART model achieved the highest ROUGE-1 score of 0.65, ROUGE-L score of 0.56, and BERTScore of 0.95 compared to other models. Additionally, CRF-based preprocessing proved to be crucial in extracting essential information and minimizing unnecessary details for the summarization process. Conclusions: This study shows that advanced NLP techniques can make supply chain communication workflows more efficient. The BART-based email summarization tool that we created showed great potential in giving important insights and helping planners deal with information overload.

No description available.

Made available in DSpace on 2016-06-02T19:05:48Z (GMT). No. of bitstreams: 1 3512.pdf: 1897835 bytes, checksum: 598f309a846cb201fe8f13be0f2e37da (MD5) Previous issue date: 2010-12-21 / Automatic text summarization has been of great interest in Natural Language Processing due to the need of processing a huge amount of information in short time, which is usually delivered through distinct media. Thus, large-scale methods are of utmost importance for synthesizing and making access to information simpler. They aim at preserving relevant content of the sources with little or no human intervention. Building upon the extractive summarizer SuPor and focusing on texts in Portuguese, this MsC work aimed at exploring varied features for automatic summarization. Computational methods especially driven towards textual statistics, graphs and machine learning have been explored. A meaningful extension of the SuPor system has resulted from applying such methods and new summarization models have thus been delineated. These are based either on each of the three methodologies in isolation, or are hybrid. In this dissertation, they are generically named after the original SuPor as SuPor-2. All of them have been assessed by comparing them with each other or with other, well-known, automatic summarizers for texts in Portuguese. The intrinsic evaluation tasks have been carried out entirely automatically, aiming at the informativeness of the outputs, i.e., the automatic extracts. They have also been compared with other well-known automatic summarizers for Portuguese. SuPor-2 results show a meaningful improvement of some SuPor-2 variations. The most promising models may thus be made available in the future, for generic use. They may also be embedded as tools for varied Natural Language Processing purposes. They may even be useful for other related tasks, such as linguistic studies. Portability to other languages is possible by replacing the resources that are language-dependent, namely, lexicons, part-of-speech taggers and stop words lists. Models that are supervised have been so far trained on news corpora. In spite of that, training for other genres may be carried out by interested users using the very same interfaces supplied by the systems. / A tarefa de Sumarização Automática de textos tem sido de grande importância dentro da área de Processamento de Linguagem Natural devido à necessidade de se processar gigantescos volumes de informação disponibilizados nos diversos meios de comunicação. Assim, mecanismos em larga escala para sintetizar e facilitar o acesso a essas informações são de extrema importância. Esses mecanismos visam à preservação do conteúdo mais relevante e com pouca ou nenhuma intervenção humana. Partindo do sumarizador extrativo SuPor e contemplando o Português, este trabalho de mestrado visou explorar variadas características de sumarização pela utilização de métodos computacionais baseados em estatísticas textuais, grafos e aprendizado de máquina. Esta exploração consistiu de uma extensão significativa do SuPor, pela definição de novos modelos baseados nessas três abordagens de forma individual ou híbrida. Por serem originários desse sistema, manteve-se a relação com seu nome, o que resultou na denominação genérica SuPor-2. Os diversos modelos propostos foram, então, comparados entre si em diversos experimentos, avaliando-se intrínseca e automaticamente a informatividade dos extratos produzidos. Foram realizadas também comparações com outros sistemas conhecidos para o Português. Os resultados obtidos evidenciam uma melhora expressiva de algumas variações do SuPor-2 em relação aos demais sumarizadores extrativos existentes para o Português. Os sistemas que se evidenciaram superiores podem ser disponibilizados no futuro para utilização geral por usuários comuns ou ainda para utilização como ferramentas em outras tarefas do Processamento de Língua Natural ou em áreas relacionadas. A portabilidade para outras línguas é possível com a substituição dos recursos dependentes de língua, como léxico, etiquetadores morfossintáticos e stoplist Os modelos supervisionados foram treinados com textos jornalísticos até o momento. O treino para outros gêneros pode ser feito pelos usuários interessados através dos próprios sistemas desenvolvidos

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 / Universidade Federal de Minas Gerais / 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. / 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.

No description available.

Abstractive text summarization aims to condense long textual documents into a short, human-readable form while preserving the most important information from the source document. A common approach to training summarization models is by using maximum likelihood estimation with the teacher forcing strategy. Despite its popularity, this method has been shown to yield models with suboptimal performance at inference time. This work examines how using alternative, task-specific training signals affects the performance of summarization models. Two novel training signals are proposed and evaluated as part of this work. One, a novelty metric, measuring the overlap between n-grams in the summary and the summarized article. The other, utilizing a discriminator model to distinguish human-written summaries from generated ones on a word-level basis. Empirical results show that using the mentioned metrics as rewards for policy gradient training yields significant performance gains measured by ROUGE scores, novelty scores and human evaluation. / Abstraktiv textsammanfattning syftar på att korta ner långa textdokument till en förkortad, mänskligt läsbar form, samtidigt som den viktigaste informationen i källdokumentet bevaras. Ett vanligt tillvägagångssätt för att träna sammanfattningsmodeller är att använda maximum likelihood-estimering med teacher-forcing-strategin. Trots dess popularitet har denna metod visat sig ge modeller med suboptimal prestanda vid inferens. I det här arbetet undersöks hur användningen av alternativa, uppgiftsspecifika träningssignaler påverkar sammanfattningsmodellens prestanda. Två nya träningssignaler föreslås och utvärderas som en del av detta arbete. Den första, vilket är en ny metrik, mäter överlappningen mellan n-gram i sammanfattningen och den sammanfattade artikeln. Den andra använder en diskrimineringsmodell för att skilja mänskliga skriftliga sammanfattningar från genererade på ordnivå. Empiriska resultat visar att användandet av de nämnda mätvärdena som belöningar för policygradient-träning ger betydande prestationsvinster mätt med ROUGE-score, novelty score och mänsklig utvärdering.

No description available.

Over the last decades, there have been remarkable shifts in the area of Information Retrieval (IR) as huge amount of information is increasingly accumulated on the Web. The gigantic information explosion increases the need for discovering new tools that retrieve meaningful knowledge from various complex information sources. Thus, techniques primarily used to search and extract important information from numerous database sources have been a key challenge in current IR systems. Topic modeling is one of the most recent techniquesthat discover hidden thematic structures from large data collections without human supervision. Several topic models have been proposed in various fields of study and have been utilized extensively for many applications. Latent Dirichlet Allocation (LDA) is the most well-known topic model that generates topics from large corpus of resources, such as text, images, and audio.It has been widely used in many areas in information retrieval and data mining, providing efficient way of identifying latent topics among document collections. However, LDA has a drawback that topic cohesion within a concept is attenuated when estimating infrequently occurring words. Moreover, LDAseems not to consider the meaning of words, but rather to infer hidden topics based on a statisticalapproach. However, LDA can cause either reduction in the quality of topic words or increase in loose relations between topics. In order to solve the previous problems, we propose a domain specific topic model that combines domain concepts with LDA. Two domain specific algorithms are suggested for solving the difficulties associated with LDA. The main strength of our proposed model comes from the fact that it narrows semantic concepts from broad domain knowledge to a specific one which solves the unknown domain problem. Our proposed model is extensively tested on various applications, query expansion, classification, and summarization, to demonstrate the effectiveness of the model. Experimental results show that the proposed model significantly increasesthe performance of applications.

Programmers increasingly rely on managed languages (e.g. Java and C#) to develop applications faster and with fewer bugs. Managed languages encourage allocating objects in the heap and rely on automatic memory management (garbage collection) to reclaim objects the program can no longer access. With more objects in the heap, the heap encodes more program state than ever before and offers new opportunities for optimization and analysis. This dissertation shows how to efficiently leverage the managed runtime to perform dynamic heap analysis. Previous heap analysis approaches significantly slow down programs, require special hardware, and/or increase memory consumption by 75% or more. We presents two synergistic techniques—dynamic object sampling (DOS) and heap summarization (HSG)—that mine program state embedded in the heap efficiently enough to use in production and effectively enough to improve performance, find bugs, and increase program understanding. We use these techniques to address three problems: (1) Performance of managed language. Because some objects live for a long time, they incur disproportionate collection costs. We optimize these costs with dynamic pretenuring. Dynamic pretenuring uses DOS to accurately predict allocation site survival rates and uses these predictions to improve performance. (2) Finding bugs. Memory leaks in managed languages occur when a program inadvertently maintains references to objects that it no longer needs. Along with degrading performance and resulting in program crashes, memory leaks cause systematic heap growth. We introduce Cork which uses the simplest type of HSG, a class points-from summary graph (CPFG), to detect systematic heap growth. Cork quickly identifies growing data structures observed in three popular benchmarks (fop, jess, and jbb2000) while adding an average of only 2.3% to total time. Additionally, we use Cork to debug a reported memory leak in Eclipse. (3) Program understanding. For a long time, static analysis has sought to statically summarize the shape of dynamic data structures to aid in program verification and understanding. Unfortunately, it only works on small programs. We introduce ShapeUp which instead characterizes recursive data structures dynamically by discovering data structure shape and degree invariants at runtime. ShapeUp uses DOS and a class field-wise summary graph (CFSG) to track in- and out-degree invariants of data structure nodes. We show how ShapeUp automatically identifies recursive data structures and likely shape invariants. Finally, we monitor discovered shape invariants to detect when a data structure becomes malformed. In summary, this dissertation is the first to leverage the managed runtime to perform dynamic heap analysis both accurately and efficiently. Our results show that the heap contains an enormous amount of program state and that there is much potential for dynamically mining heap characteristics for optimization, debugging, and program understanding. / text