Showing papers on "Multi-document summarization published in 2001"

Generic text summarization using relevance measure and latent semantic analysis

Abstract: In this paper, we propose two generic text summarization methods that create text summaries by ranking and extracting sentences from the original documents. The first method uses standard IR methods to rank sentence relevances, while the second method uses the latent semantic analysis technique to identify semantically important sentences, for summary creations. Both methods strive to select sentences that are highly ranked and different from each other. This is an attempt to create a summary with a wider coverage of the document's main content and less redundancy. Performance evaluations on the two summarization methods are conducted by comparing their summarization outputs with the manual summaries generated by three independent human evaluators. The evaluations also study the influence of different VSM weighting schemes on the text summarization performances. Finally, the causes of the large disparities in the evaluators' manual summarization results are investigated, and discussions on human text summarization patterns are presented.

Finding topic words for hierarchical summarization

Abstract: Hierarchies have long been used for organization, summarization, and access to information. In this paper we define summarization in terms of a probabilistic language model and use the definition to explore a new technique for automatically generating topic hierarchies by applying a graph-theoretic algorithm, which is an approximation of the Dominating Set Problem. The algorithm efficiently chooses terms according to a language model. We compare the new technique to previous methods proposed for constructing topic hierarchies including subsumption and lexical hierarchies, as well as the top TF.IDF terms. Our results show that the new technique consistently performs as well as or better than these other techniques. They also show the usefulness of hierarchies compared with a list of terms.

Experiments in Single and Multi-Document Summarization Using MEAD

Abstract: In this paper, we describe four experiments in text summarization. The first experiment involves the automatic creation of 120 multi-document summaries and 308 single-document summaries from a set of 30 clusters of related documents. We present official results from a multi-site manual evaluation of the quality of the summaries. The second experiment is about the identification by human subjects of cross-document structural relationships such as identity, paraphrase, elaboration, and fulfillment. The third experiment focuses on a particular cross-document structural relationship, namely subsumption. The last experiment asks human judges to determine which of the input articles in a given cluster were used to produce individual sentences of a manual summary. We present numerical evaluations of all four experiments. All automatic summaries have been produced by MEAD, a flexible summarization system under development at the University of Michigan.

A new approach to unsupervised text summarization

Abstract: The paper presents a novel approach to unsupervised text summarization. The novelty lies in exploiting the diversity of concepts in text for summarization, which has not received much attention in the summarization literature. A diversity-based approach here is a principled generalization of Maximal Marginal Relevance criterion by Carbonell and Goldstein \cite{carbonell-goldstein98}.We propose, in addition, aninformation-centricapproach to evaluation, where the quality of summaries is judged not in terms of how well they match human-created summaries but in terms of how well they represent their source documents in IR tasks such document retrieval and text categorization.To find the effectiveness of our approach under the proposed evaluation scheme, we set out to examine how a system with the diversity functionality performs against one without, using the BMIR-J2 corpus, a test data developed by a Japanese research consortium. The results demonstrate a clear superiority of a diversity based approach to a non-diversity based approach.

Multidocument summarization via information extraction

Abstract: We present and evaluate the initial version of RIPTIDES, a system that combines information extraction, extraction-based summarization, and natural language generation to support user-directed multidocument summarization.

Sentence ordering in multidocument summarization

Abstract: The problem of organizing information for multidocument summarization so that the generated summary is coherent has received relatively little attention. In this paper, we describe two naive ordering techniques and show that they do not perform well. We present an integrated strategy for ordering information, combining constraints from chronological order of events and cohesion. This strategy was derived from empirical observations based on experiments asking humans to order information. Evaluation of our augmented algorithm shows a significant improvement of the ordering over the two naive techniques we used as baseline.

NewsInEssence: a system for domain-independent, real-time news clustering and multi-document summarization

Abstract: NEWSINESSENCE is a system for finding, visualizing and summarizing a topic-based cluster of news stories. In the generic scenario for NEWSINESSENCE, a user selects a single news story from a news Web site. Our system then searches other live sources of news for other stories related to the same event and produces summaries of a subset of the stories that it finds, according to parameters specified by the user.

Event detection and summarization in American football broadcast video

Abstract: We propose a framework for event detection and summary generation in football broadcast video. First, we formulate summarization as a play detection problem, with play being defined as the most basic segment of time during which the ball is being played. Then we propose both deterministic and probabilistic approaches to the detection of the plays. The detected plays are concatenated to generate a compact, time-compressed summary of the original video. Such a summary is complete in the sense that it contains every meaningful action of the underlying game, and it also servers as a much better starting point for higher-level summarization and other analyses than the original video does. Based on the summary, we also propose an audio-based hierarchical summarization method. Experimental results show the proposed methods work very well on consumer grade platforms.© (2001) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Automatic generation of concise summaries of spoken dialogues in unrestricted domains

Abstract: Automatic summarization of open domain spoken dialogues is a new research area. This paper introduces the task, the challenges involved, and presents an approach to obtain automatic extract summaries for multi-party dialogues of four different genres, without any restriction on domain. We address the following issues which are intrinsic to spoken dialogue summarization and typically can be ignored when summarizing written text such as newswire data: (i) detection and removal of speech disfluencies; (ii) detection and insertion of sentence boundaries; (iii) detection and linking of cross-speaker information units (question-answer pairs). A global system evaluation using a corpus of 23 relevance annotated dialogues containing 80 topical segments shows that for the two more informal genres, our summarization system using dialogue specific components significantly outperforms a baseline using TFIDF term weighting with maximum marginal relevance ranking (MMR).

Recent developments in text summarization

Abstract: With the explosion in the quantity of on-line text and multimedia information in recent years, demand for text summarization technology is growing. Increased pressure for technology advances is coming from users of the web, on-line information sources, and new mobile devices, as well as from the need for corporate knowledge management. Commercial companies are increasingly starting to offer text summarization capabilities, often bundled with information retrieval tools. In this paper, I will discuss the significance of some recent developments in summarization technology.

Domain-specific informative and indicative summarization for information retrieval

Abstract: In this paper, we propose the use of multido ument summarization as a post-pro essing step in do ument retrieval We examine the use of the summary as a repla ement to the standard ranked list The form of the summary is novel be ause it has both informative and indi ate elements, designed to help di erent users perform their tasks better Our summary uses the do uments' topi al stru ture as a ba kbone for its own stru ture, as it was deemed the most useful do ument feature in our study of a orpus of summaries

An Extrinsic Evaluation for Question-Biased Text Summarization on QA tasks

Abstract: This paper proposes a Question-Biased Text Summarization (QBTS) approach that is useful for question-answering systems. QBTS is an extension of Query-Biased Text Summarization in the sense that summarization is biased not only by the question, which corresponds to the query, but also by the prospective answers to the question. We conducted text summarization experiments based on QA tasks and confirmed the effectiveness of our method in obtaining short summaries.

Internet as Corpus : Automatic Construction of a Swedish News Corpus

Abstract: Today, with digitally stored information available in abundance, even for many minor languages, this information must by some means be filtered and extracted in order to avoid drowning in it. Automatic summarization is one such technique, where a computer summarizes a longer text to a shorter non-rendundant form. Apart from the major languages of the world there are a lot of languages for which large bodies of data aimed at language technology research to a high degree are lacking. There might also not be resources available to develop such bodies of data, since it is usually time consuming and requires substantial manual labor, hence being expensive. Nevertheless, there will still be a need for automatic text summarization for these languages in order to subdue this constantly increasing amount of electronically produced text. This thesis thus sets the focus on automatic summarization of text and the evaluation of summaries using as few human resources as possible. The resources that are used should to as high extent as possible be already existing, not specifically aimed at summarization or evaluation of summaries and, preferably, created as part of natural literary processes. Moreover, the summarization systems should be able to be easily assembled using only a small set of basic language processing tools, again, not specifically aimed at summarization/evaluation. The summarization system should thus be near language independent as to be quickly ported between different natural languages. The research put forth in this thesis mainly concerns three computerized systems, one for near language independent summarization – The HolSum summarizer; one for the collection of large-scale corpora – The KTH News Corpus; and one for summarization evaluation – The KTH eXtract Corpus. These three systems represent three different aspects of transferring the proposed summarization method to a new language. One aspect is the actual summarization method and how it relates to the highly irregular nature of human language and to the difference in traits among language groups. This aspect is discussed in detail in Chapter 3. This chapter also presents the notion of “holistic summarization”, an approach to self-evaluative summarization that weighs the fitness of the summary as a whole, by semantically comparing it to the text being summarized, before presenting it to the user. This approach is embodied as the text summarizer HolSum, which is presented in this chapter and evaluated in Paper 5. A second aspect is the collection of large-scale corpora for languages where few or none such exist. This type of corpora is on the one hand needed for building the language model used by HolSum when comparing summaries on semantic grounds, on the other hand a large enough set of (written) language use is needed to guarantee the randomly selected subcorpus used for evaluation to be representative. This topic briefly touched upon in Chapter 4, and detailed in Paper 1. The third aspect is, of course, the evaluation of the proposed summarization method on a new language. This aspect is investigated in Chapter 4. Evaluations of HolSum have been run on English as well as on Swedish, using both well established data and evaluation schemes (English) as well as with corpora gathered “in the wild” (Swedish). During the development of the latter corpora, which is discussed in Paper 4, evaluations of a traditional sentence ranking text summarizer, SweSum, have also been run. These can be found in Paper 2 and 3. This thesis thus contributes a novel approach to highly portable automatic text summarization, coupled with methods for building the needed corpora, both for training and evaluation on the new language.

Multi-Document Summarization and Visualization in the Informedia Digital Video Library

Abstract: The Informedia Digital Video Library project provided a technological foundation for full content indexing and retrieval of video and audio media. The library now contains over 2000 hours of video and is growing daily. A good query engine is not sufficient for information retrieval because often the candidate result sets grow in number as the library grows. Video digests summarize sets of stories from the library, providing users with a visual mechanism for interactive browsing and query refinement. These digests are generated dynamically under the direction of the user based on automatically derived metadata from the video library.

Development of a Swedish Corpus for Evaluating Summarizers and other IR-tools

Abstract: We are presenting the construction of a Swedish corpus aimed at research1on Information Retrieval, Information Extraction, Named Entity Recognitionand Multi Text Summarization, we will also present ...

Towards the use of prosodic information for spoken document retrieval

Abstract: When human talk, they exchange a lot of audio information along with the words. Computers, however, don’t hear between the lines, which is one reason why speech recognition application can seem so frustratingly stupid. The technical reason for this is in the Hidden Markow Model (HMM) that most speech recognition systems employ for speech processing. HMM rely only on tiny, 10 millisecond slices of speech and while this works well for picking up words, it cannot capture the conceptual cues that span words, phrases or sentences. In fact, a pause at the end of a sentence or a drop in pitch spans 10-100 times large than what HMM can capture. Prosodic information is slowly being recognised as an important source of information in speech understanding. Prosody includes the duration, pitch and energy of speech. Duration, or the way people stretch or speed certain parts of speech is, arguably, the most important component of prosodic information. Studies in psycho-linguistic have shown that people use the duration of speech sounds in certain ways to emphasise the information content of what they are saying. Currently, spoken document retrieval (SDR) is carried out using Information Retrieval (IR) techniques on transcripts of spoken documents. These transcripts are produced using speech recognition systems, often employing HMM, that disregard prosodic information, concentrating instead in producing the most faithful possible recognition of the spoken words. Although this approach has proven quite successful [1], we believe that prosodic information should be used for a more effective indexing and retrieval of spoken documents in all those situations in which it would be impossible to obtain acceptable levels of speech recognition accuracy. The work briefly presented in this paper aims at: 1) establishing a correspondence between prosodic information and semantic information to define ways of identifying “acoustic” keywords; 2) studying the relationship between “prosodic in-

Combined-media video tracking for summarization

Abstract: Video summarization is receiving increasing attention due to the large amount of video content made available on the Internet. In this paper we present a novel idea to track video from multiple sources for video summarization. An algorithm that takes advantage of both video and close caption text information for video scene clustering is described. Experimental results are given followed by discussion on future directions.

Self supervised learning for automatic text summarization by text span extraction

Abstract: We describe a system for automatic text summarization that operates by extracting the most relevant sentences from documents with regard to a query. The lack of labeled corpora makes it difficult to develop automatic techniques for summarization. We propose to use a self-supervised method which does not rely on the availability of labeled corpora for learning to rank sentences for the summary. The method operates in two steps: first a statistical similarity based system which does not require any training is developed, second a classifier is trained using self-supervised learning in order to improve this baseline method. This idea is evaluated on the Reuters news-wire corpus and compared to other strategies.

Yet Another Summarization System with Two Modules using Empirical Knowledge.

Abstract: We previously proposed a summarization system, GREEN, for Japanese newspaper editorials However, GREEN is not suitable for summarizing ordinal newspaper articles which are different from newspaper editorials To participate in subtasks A-1 and A-2 of TSC (text Summarization Challenge) in NTCIR-2, we developed a new summarization system from scratch which copes with both ordinal articles and editorials in a Japanese newspaper The new summarization system resulted in good evaluations: the mean value of all evaluations held the foremost place among ten systems in subtask A-1 and nine systems in subtask A-2, respectively

Summarization of news speech with unknown topic boundary

Abstract: TV viewers want to grasp the contents of the news program in a short time due to the increasing number of news channels. Conventional summarization methods based on extraction of the important sentences from each topic included in the news speech is insufficient because the important sentences can not always be extracted from each topic due to unknown topic boundary. To solve this problem, in this paper, we propose a summarization method of TV news program by segmenting the news speech into topics and then extracting the important sentence from each topic.

Interactive phrase browsing within compressed text

Abstract: Introduction. Several techniques have been described for pattern searching in compressed text [de Moura et al., 1998, Manber, 1997]. In these methods the compression regime is usually modified in some way to suit the needs of the searching process, and compression effectiveness compromised. Furthermore, the searching mechanism used is usually based upon linear-search regular expression pattern matching, meaning that it can be expensive to handle large files. Finally, the whole nature of the exact-match searching process means that a small misconception in the construction of the pattern to be searched can result in a belief that the information sought is not present when in fact it is. Nevill-Manning et al. [1997] describe PHIND, a phrase-browsing approach to searching, in which a user “explores” the collection, and narrows down their region of interest step by step until their information need is met. In this presentation we build upon this approach, and show how the RE-PAIR compression regime [Larsson and Moffat, 2000] naturally supports a similar browsing capability, and also is such that locations in the text at which target phrases appear can be quickly identified, decompressed, and returned to the user. And because by-products of the compression process are used by the browser, no explicit index is required.

A System for Text Summarization Based on Word Importance Measures.

Abstract: This paper presents a text summarization system based on word importance measures. It produces a summary through a part or all of the following four steps. First, we assign local and global scores to the nouns in each sentence according to the functional roles they play and the significancy they assume in the text for summarization. Second, based on the local and global scores, we calculate the importance value of each sentence. Third, the sentences with the importance values big enough for a summary are selected. Fourth and finally, a coherency test is given and we make the final decision for inclusion or exclusion of the selected sentences into the summary. We compare the summaries thus produced with the ones by humans and other systems.

How Small a Distinction among Summaries can the Evaluation Method Identify

Abstract: This paper describes the summarization methods used by the team FLAB (gid040) for text summarization tasks in the NTCIR-2 workshop. The focus is on the effectiveness of an extrinsic evaluation based on relevance assessment in information retrieval with reference to the evaluation results obtained for a task B. The team FLAB submitted two types of summaries for the task: baseline summaries and thematic hierarchy based summaries. Statistical analysis of the results of five types of summaries comprising these two types of submitted summaries and three types of official baseline summaries suggests that the difference between the former two was too small for the evaluation to identify. This suggests that the task might better be performed under the condition that only a few subjects could complete the given task, e.g., with only a certain short time limit given for assessment.

Dreams for Some and Nightmares for Others

Abstract: Through two dreams, past and current, an ideal online information retrieval system is depicted, including full text online access, real time reference assistance via the Internet, and automatic summarization of all papers and chapters. This is the near future of information retrieval and processing. Selected resources on automatic summarization are reported and some thoughts on implications are offered.

Monologue Summarization: Extraction of Important Sentences for TV News Commentary Programs.

Abstract: The extraction of important sentences is a key technique for automatic summarization. Whereas most research in this area has targeted written language, we are conducting research on spoken language monologues such as presentations and TV news commentary programs. We collected 50 TV news commentary programs, and experimented with the extraction of important sentences from transcriptions. We used two extraction methods. The first one uses word statistics, and the second one uses the surface features of the sentences. In order to use the latter method, we analyzed the transcriptions and obtained surface features related to the importance of the sentences. The experiments showed that the latter method was better than the former one especially when extracting small sets of sentences. We also mention the ambiguity of judgment by individuals and the contribution of each surface feature to the importance of the sentences.

Machine Learning Methods for Text Summarization

Abstract: This paper addresses the problem of creating a summary by extracting a set of sentences that are likely to represent the content of a document. A small scale experiment is conducted leading to the compilation of an evaluation corpus for the Greek language. Three models of sentence extraction are then described, along the lines of shallow linguistic analysis, feature combination and machine learning