Summon架构-华南师范大学图书馆.ppt

myocardial infarction：心肌梗塞 * * * 鉴于Summon系统一直以来的优异表现，原来越多的内容提供商选择与Summon合作，目前，全球已经有超过7600家出版社为Summon提供内容 Nearly all participating publishers provide full text for indexing! * This is a partial list * * 将图书馆的馆藏放在统一的架构、单一的索引下进行探索是是Summon的独特之处，也是它取得成功的关键 这种统一的索引结构 (Unified Index) 影响到探索服务的各个方面：速度，易用性，相关性排序，个性化设定的灵活度，研究资料的完整性，系统的稳定性等等 Normalization, Correction and Deduplication of metadata across all sources. Not just from a single content vendor or platform. Not reliant on thin metadata from citation sources like CrossRef. All of this work happens prior to indexing. * While this model over-simplifies the complexity associated with applying relevance ranking algorithms across a body content as large as the Summon index, Summon offers advanced relevancy tuning that is unmatched by any other discovery service. First it is important to understand how Summon works. The fact that we do return results from a single, unified index across the full breadth of the library’s collection, means that we actually do have all of the results available to apply a relevance ranking algorithm to. Of course Summon uses Dynamic Relevance Ranking – the types of things you’d expect such as term frequency, word proximity, field weighting, etc.; but we also use Static Ranking. Just as Google uses “page rank” to boost web pages that are hit more often, Summon uses Static Rank to boost items by different factors. Local content and books might get boosted, as would scholarly or more recent articles, while a less scholarly item might get downgraded in the relevancy. All results come back in a single, unified result set. But the user is still given unparalleled options for using facets and limiters that apply to all of the content equally allowing the user to zoom in on the most relevant results. In addition Summon can even provide context sensitive recommendations to other content sources, research guides, etc. that may aide the user. * * * * * * The only discovery solution that automat