Modern search engines display a summary for each ranked document that is returned in response to a query. These summaries typically include a snippet - a collection of text fragments from the underlying document - tha...
详细信息
ISBN:
(纸本)9781450302470
Modern search engines display a summary for each ranked document that is returned in response to a query. These summaries typically include a snippet - a collection of text fragments from the underlying document - that has some relation to the query that is being answered. In this study we investigate how 10 humans construct snippets: participants first generate their own natural language snippet, and then separately extract a snippet by choosing text fragments, for four queries related to two documents. By mapping their generated snippets back to text fragments in the source document using eye tracking data, we observe that participants extract these same pieces of text around 73% of the time when creating their extractive snippets. In comparison, we notice that automated approaches for extracting snippets only use these same fragments 22% of the time. However, when the automated methods are evaluated using a position-independent bag-of-words approach, as typically used in the research literature for evaluating snippets, they appear to be much more competitive, with only a 24 point difference in coverage, compared to the human extractive snippets. While there is a 51 point difference when word position is taken into account. In addition to demonstrating this large scope for improvement in snippet generation algorithms with our novel methodology, we also offer a series of observations on the behaviour of participants as they constructed their snippets. Copyright 2010 ACM.
The widening performance gap between processor and disk demands innovations. Several optimization techniques, (e.g., read caching, write buffering, prefetching), have been proposed to narrow this gap. Read caching is ...
详细信息
ISBN:
(纸本)9780889868205
The widening performance gap between processor and disk demands innovations. Several optimization techniques, (e.g., read caching, write buffering, prefetching), have been proposed to narrow this gap. Read caching is the least effective technique among them in resource-poor condition. In this paper, we propose a novel technique which splits the disk cache into metadata and data to improve read miss ratio. The proposal is based on observations that metadata is smaller in size and more frequently accessed than its data. Split caches can reduce the interference between metadata and data. Our study shows that by splitting the disk cache the effective read miss ratio can be improved by 20%, which in turn would bring about 16% performance improvements in response time. Furthermore, the total size of the split caches is about 35%-97% of the unified cache to reach the same read miss ratio.
We present PlayGo, a comprehensive tool for scenario-based programming, built around the language of live sequence charts and the play-in/play-out approach [7], which includes a compiler into AspectJ code and means fo...
详细信息
This paper examines the role of social networks and artefacts in technology appropriation in an educational environment. We use a qualitative research method to illustrate that social networks and accompanying artefac...
详细信息
Usually, in a group decision context, the importance level, confidence degree and amount of knowledge are very different among individuals. So, when all the individuals have to reach agreement, is quite important to m...
详细信息
This paper describes a method for detecting event trigger words in biomedical text based on a word sense disambiguation (WSD) approach. We first investigate the applicability of existing WSD techniques to trigger word...
详细信息
With the recent spreading of social networks and digital communities, the Web has a new business model for providing free content-sharing services. Using the power of the masses and the collective intelligence, Web 2....
详细信息
ISBN:
(纸本)9781601321480
With the recent spreading of social networks and digital communities, the Web has a new business model for providing free content-sharing services. Using the power of the masses and the collective intelligence, Web 2.0 sites have created enormous e-commerce opportunities for marketing and advertising. Thus, there is still a necessity of developing tools to help users to reach good decisions in such frameworks. We propose to manage this collective intelligence with a Decision Support System (DSS). It allows that users with similar user profiles that the potential customer can advice him/her according with their own experiences in their e-commerce activities. From user preferences the DSS obtains a collective advice that helps customers to make decisions related with their e-commerce activities.
This paper presents the novel task of best topic word selection, that is the selection of the topic word that is the best label for a given topic, as a means of enhancing the interpretation and visualisation of topic ...
This paper presents the novel task of best topic word selection, that is the selection of the topic word that is the best label for a given topic, as a means of enhancing the interpretation and visualisation of topic models. We propose a number of features intended to capture the best topic word, and show that, in combination as inputs to a reranking model, we are able to consistently achieve results above the baseline of simply selecting the highest-ranked topic word. This is the case both when training in-domain over other labelled topics for that topic model, and cross-domain, using only labellings from independent topic models learned over document collections from different domains and genres.
Multi-category bootstrapping algorithms were developed to reduce semantic drift. By extracting multiple semantic lexicons simultaneously, a category's search space may be restricted. The best results have been ach...
详细信息
We introduce integrated parking management system (IPMS) based on wireless sensor network. The system was designed with the object of a large scale deployment in existing facilities, and has been actually deployed to ...
详细信息
暂无评论