An unsupervised approach to generating generic summaries of documents

We present an optimization-based unsupervised approach to automatic document summarization. In the proposed approach, text summarization is modeled as a Boolean programming problem. This model generally attempts to optimize three properties, namely, (1) relevance: summary should contain informative textual units that are relevant to the user; (2) redundancy: summaries should not contain multiple textual units that convey the same information; and (3) length: summary is bounded in length. The approach proposed in this paper is applicable to both tasks: single- and multi-document summarization. In both tasks, documents are split into sentences in preprocessing. We select some salient sentences from document(s) to generate a summary. Finally, the summary is generated by threading all the selected sentences in the order that they appear in the original document(s). We implemented our model on multi-document summarization task. When comparing our methods to several existing summarization methods on an open DUC2005 and DUC2007 data sets, we found that our method improves the summarization results significantly. This is because, first, when extracting summary sentences, this method not only focuses on the relevance scores of sentences to the whole sentence collection, but also the topic representative of sentences. Second, when generating a summary, this method also deals with the problem of repetition of information. The methods were evaluated using ROUGE-1, ROUGE-2 and ROUGE-SU4 metrics. In this paper, we also demonstrate that the summarization result depends on the similarity measure. Results of the experiment showed that combination of symmetric and asymmetric similarity measures yields better result than their use separately.     

An intelligent web-page classifier with fair feature-subset selection

The explosion of on-line information has given rise to many manually constructed topic hierarchies (such as Yahoo!!). But with the current growth rate in the amount of information, manual classification in topic hierarchies results in an immense information bottleneck. Therefore, developing an automatic classifier is an urgent need. However, classifiers suffer from enormous dimensionality, since the dimensionality is determined by the number of distinct keywords in a document corpus. More seriously, most classifiers are either working slowly or they are constructed subjectively without any learning ability. In this paper, we address these problems with a fair feature-subset selection (FFSS) algorithm and an adaptive fuzzy learning network (AFLN) for classification. The FFSS algorithm is used to reduce the enormous dimensionality. It not only gives fair treatment to each category but also has ability to identify useful features, including both positive and negative features. On the other hand, the AFLN provides extremely fast learning ability to model the uncertain behavior for classification so as to correct the fuzzy matrix automatically. Experimental results show that both FFSS algorithm and the AFLN lead to a significant improvement in document classification, compared to alternative approaches.     

Empirical comparison of image retrieval color similarity methods with human judgment

It is important to verify assumptions and methods of image retrieval against actual human behavior. A study was conducted to compare similarity methods of color histograms against human assessment of similarity. The similarity methods tested include basic histogram intersection, center histogram matching, locality histogram matching, and size-weighted histogram matching. 161 subjects participated in the empirical study. The findings, based on Spearman correlation analysis, showed that both the basic histogram intersection method and size-weighted histogram are very close to human assessment of similarity (Spearman correlation coefficient of 0.915). The other two are not close to human judgment on similarity. This study illustrates an alternative approach to evaluating matching algorithms. Unlike the usual measures of recall and precision, this approach emphasizes human validation. Fewer images are required with the use of statistical testing.     

Using Web resources to construct multilingual medical thesaurus for cross-language medical information retrieval

Due to the language barrier, non-English users are unable to retrieve the most updated medical information from the U.S. authoritative medical websites, such as PubMed and MedlinePlus. However, currently, there is no any cross-language medical information retrieval (CLMIR) system that can help Chinese-speaking consumers cross the language barrier in finding useful English medical information. A few CLMIR systems utilize MeSH (Medical Subject Headings) to help overcome the language barrier. Unfortunately, the traditional Chinese version of MeSH is currently unavailable.In this paper, we employ a semi-automatic term translation method to construct a Chinese–English MeSH by exploiting abundant multilingual Web resources, including Web anchor texts and search–result pages. Through this method, we have developed a Chinese–English Mesh Compilation System to assist knowledge engineers in compiling a Chinese–English medical thesaurus with more than 19,000 entries. Furthermore, this thesaurus has been used to develop a prototypical system for cross-language medical information retrieval, MMODE, which can help consumers retrieve top-quality English medical information using Chinese terms.     

Configurational and nonconfigurational interactions between odorants in binary mixtures.

Studies on odor mixture perception suggest that although odor components can often be identified in mixtures, mixtures can also give rise to novel perceptual qualities that are not present in the components. Using an olfactory habituation task, the authors evaluated how the perceptual similarity between components in a mixture affects the perceptual quality of the mixture itself. Rats perceived binary mixtures composed of similar components as different from their 2 components, whereas binary mixtures composed of dissimilar components were perceived as very similar to their components. Results show that for both types of mixtures, pretraining to Component A reduces subsequent learning about Component B in rats trained in the presence of A. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The time course of spoken word learning and recognition: Studies with artificial lexicons.

The time course of spoken word recognition depends largely on the frequencies of a word and its competitors, or neighbors (similar-sounding words). However, variability in natural lexicons makes systematic analysis of frequency and neighbor similarity difficult. Artificial lexicons were used to achieve precise control over word frequency and phonological similarity. Eye tracking provided time course measures of lexical activation and competition (during spoken instructions to perform visually guided tasks) both during and after word learning, as a function of word frequency, neighbor type, and neighbor frequency. Apparent shifts from holistic to incremental competitor effects were observed in adults and neural network simulations, suggesting such shifts reflect general properties of learning rather than changes in the nature of lexical representations. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Category Norms as a Function of Culture and Age: Comparisons of Item Responses to 105 Categories by American and Chinese Adults.

Understanding how aging influences cognition across different cultures has been hindered by a lack of standardized, cross-referenced verbal stimuli. This study introduces a database of such item-level stimuli for both younger and older adults, in China and the United States, and makes 3 distinct contributions. First, the authors specify which item categories generalize across age and/or cultural groups, rigorously quantifying differences among them. Second, they introduce novel, powerful methods to measure between-group differences in freely generated ranked data, the rank-ordered logit model and Hellinger Affinity. Finally, a broad archive of tested, cross-linguistic stimuli is now freely available to researchers: data, similarity measures, and all stimulus materials for 105 categories and 4 culture-by-age groups, comprising over 10,000 fully translated unique item responses. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

建筑制图标准中图形符号适宜性、区分性的研究

应用线段长短、间距大小视觉反应的量化规律，提出设计图形符号大小的原则——既更使其大小满足“长”“短”“大”“小”的心理要求，又要互相之间在大小上有区分性。根据图形组织的相似、接近、连续三原则提出设计图形符号形状的原则——图形符号之间尽量减少相似、接近、连续，以保证较好的区分性。通过分析发现，现行建筑制图标准中的图形符号绝大部分在大小和形状具有适宜性、区分性。     

SIMILARITY LAW OF FLUCTUATING PRESSURE WITHIN HYDRAULIC JUMP AREA

Through a series of model tests of five scales for 2-D free hydraulic jump, the da-ta of fluctuating pressure acting on the floor level within the hydraulic jump were obtained. Dur-ing the experiments, Froude number varied from 2.94 to 8.61, and Reynolds number rangedfrom 2×10~4 to 6×10~5. Experiment results indicate that the amplitude scale of fluctuating pres-sure is the length scale of model, i. e. P'=L, which agrees with gravity similarity law: Thefrequency scale of the fluctuating pressure is a unity i. e., f=1, which does not satisfy thegravity similarity law.     

Examining the effects of caching,compression and popup filtering on internet surfing

Over the years the number of Web users has increased dramatically unfortunately leading to the inherent problem of congestion. This can affect each user's surfing experience. This paper investigates download times associated with a web request, identifies where delays occur, and provides guidelines which can be followed by web developers to enable a faster and more efficient download and service for their users. We conclude that delays in Internet traffic and congestion can often be attributed to poorly developed sites (e.g. excessive image size). Copyright © 2005 John Wiley & Sons, Ltd.