III-CTX-Small: Adaptive Integration of Structured and Unstructured Data from Many Sources in a Biological Domain

该项目涉及设计和构建一个知识库（KB）系统，该系统学习更准确地集成与单个科学家的研究需求相关的许多异构信息源。该系统的工作原理是将多种类型的数据（包括非结构化文本）松散地集成到一个单一类型的有向图中，然后使用允许“无模式相似性查询”的查询语言查询该图。这些查询指定了一组查询术语(例如：关键字，知识库中的实体等)和期望输出的约束（例如目标数据类型）。查询的结果是KB实体的排序列表，按与查询词的相似度排序。在查询之后，用户可以选择将建议答案排序列表中的任何子集标记为“相关”或“不相关”。这些标签推动了一个学习阶段，其目标是产生更好的排名。目前正在研究的学习类型包括基于em的参数转换，学习判别排序，以及学习重构图（通过添加或删除边缘或顶点）。在工作生物学家的实验室中收集的问题被用来评估这些学习方法。这个项目最广泛的影响是在学习整合异构数据源（包括自由文本和结构化数据）的问题上。然而，如果成功的话，知识库系统将在生物研究界产生广泛的影响；特别是，我们相信这种自适应的个人知识库系统将是对现有生物知识库的有价值的补充。关于这个项目的更多信息可以在https://www.cs.cmu.edu/~wcohen/querendipity/找到

This project involves the design and building ofa knowledge base (KB) system thatlearns to more accurately integrate the many heterogeneous sources ofinformation that are relevant to a single scientist's research needs.The system works by loosely integrating data of many sorts (includingunstructured text) into a single typed directed graph, and thenquerying the graph using a query language that allows "schema-freesimilarity queries". These queries specify a set of query terms(e.g. keywords, entities in the KB, etc) and constraints on thedesired output (e.g. a target data type). The result of a query is aranked list of KB entities, ordered by similarity to the query terms.After a query, a user can optionally label any subset of the rankedlist of suggested answers as ``relevant'' or ``non-relevant''. Theselabels drive a learning phase, the goal of which is to produce abetter ranking. Types of learning currently being investigatedinclude EM-based parameter turning, learning to discriminativelyre-rank, and learning to restructure the graph (by adding or deletingedges or vertexes). Queries collected in the laboratory of workingbiologists are used to evaluate these learning methods.The broadest impact of this project is on the problem of learning tointegrate heterogeneous data sources (including free text andstructured data). However, if successful, the KB system will havebroad impact in the biological research community; in particular, webelieve that adaptive personal KB systems of this sort will be avaluable complement to existing biological KBs.Further information on this project can be found athttp://www.cs.cmu.edu/~wcohen/querendipity/