A Natural Language Interface for Querying General and Individual Knowledge论文学习
研究目的
To make the joint analysis of general and individual knowledge accessible to the public, it is desirable to provide an interface that translates the user questions, posed in natural language (NL), into the formal query languages that crowd mining platforms support.（为了让公众能够对一般和个人知识进行联合分析，最好提供一个界面，将用户在自然语言(NL)中提出的问题翻译成人群挖掘平台支持的正式查询语言。）
研究内容
They develop a principled approach for the translation of NL questions that mix general and individual knowledge, into formal queries.This approach is demonstrated for translating NL questions into OASSIS-QL.（因此，我们开发了一种原则性的方法，用于将混合一般知识和个人知识的 NL 问题转换为正式查询。本文演示了将NL问题转换为OASSIS-QL的方法）
研究基础
In recent work, we introduced crowd mining as a novel approach for answering user questions about a mix of individual and general knowledge, using crowdsourcing techniques. In particular, we have implemented the OASSIS platform, which supports a declarative query language, OASSIS-QL, enabling users to specify their information needs. （在最近的工作中，我们引入了人群挖掘作为一种新方法，使用众包技术回答用户关于个人和一般知识混合的问题。特别是，我们实现了OASSIS平台，该平台支持一种声明性查询语言OASSIS- ql，使用户能够指定他们的信息需求。）
问题解决思路
To support the distinct query constructs associated with these two types of knowledge, the NL question must be partitioned and translated using different means; yet eventually all the translated parts should be seamlessly combined to a well-formed query.（特别是，为了支持与这两种类型的知识相关联的不同查询结构，NL问题必须使用不同的方法进行分割和翻译;但是最终所有翻译的部分应该无缝地结合到一个格式良好的查询中。）
技术方案框架
The input NL question is converted, by standard NL parsing tools, into a well-defined structure that captures the semantic roles of text parts. A new Individual eXpression (IX) Detector then serves to decompose the structure into its general and individual parts. Each general and individual part is separately processed, and in particular, an existing General Query Generator is used to process the general query parts, whereas the individual parts are processed by our new modules. The processed individual and general query parts are integrated, via another new module, to form the final output query.(输入的NL问题被标准的NL解析工具转换为定义良好的结构，然后，一个新的个体表达式(IX)检测器将结构分解为通用部分和独立部分，单独处理这两部分，使用现有的通用查询生成器来处理通用查询部件，而独立部件则由我们的新模块处理。将处理后的单个查询和通用查询部分集成，通过另一个新模块，形成最终的输出查询。)
- A major contribution here is in providing a semantic and syntactic definition of an IX, as well a means for detecting and extracting IXs. This is done via declarative selection patterns combined with dedicated vocabularies.(论文里提供了IX的语义和语法定义，以及检测和提取IX的方法，是通过结合专用词汇表的声明性选择模式实现的)
- After executing the IX Detector, the framework now processes these IXs and the rest of the query separately to form the building blocks of the query - in the case of OASSIS-QL, the SPARQL-like triples.(在执行了IX检测器之后，框架现在会分别处理这些IX和查询的其余部分，以形成查询的构建块——在OASSIS-QL中，是类似sparql的三元组)
如何做实验
- We have implemented these techniques in a novel prototype system, NL2CM (Natural Language interface to Crowd Mining). (我们在一个新的原型系统NL2CM (Crowd Mining的自然语言接口)中实现了这些技术。)
- NL parsing: we show only the experimental results for Stanford Parser 3.3.0, trained over the Penn Treebank Corpus using the EnglishPCFG model.
- GQG: FREyA
- To examine the quality of the translation, we have allowed users to feed questions into NL2CM in a real usage scenario, and analyzed the resulting queries. Then, we have tested the applicability of our approach to real questions on wide-ranging topics, by using NL2CM to translate questions from Yahoo! Answers, the question andanswer platform.为了检查翻译的质量，我们允许用户在真实的使用场景中向NL2CM提交问题，并分析结果查询。然后，我们通过使用NL2CM翻译来自雅虎问答平台的问题，测试了我们的方法对广泛话题的真实问题的适用性。
- As a baseline, we compare the results of NL2CM with the two basic alternatives: Opinion mining and Ontology matching(作为基准，我们将NL2CM的结果与两个基本替代方案进行比较)
