研究内容
- When using the Natural language interfaces to databases, users may be surprised by the absence of certain expected results. To this end, we propose to demonstrate NLProveNAns, a system that allows non-expert users to view explanations for non-answers of interest.(在使用数据库的自然语言接口时,用户可能会对缺少某些预期结果感到惊讶。为此,我们建议演示 NLProveNAns,这是一个允许非专家用户查看非答案解释的系统。)
- the systems can provide explanations in one of two flavors corresponding to two different why-not provenance models: a short explanation based on thefrontier picky model, and a detailed explanation based on the why-not polynomial model.(系统可以提供两种类型的解释,对应于两种不同的“为什么不是“出处模型: 基于边界挑剔模型的简短解释,以及基于“为什么不是”多项式模型的详细解释。)
- enrich NaLIR with support for explanations of non-answers.(通过支持非答案的解释来丰富NaLIR)
技术方案
- we focus on explaining non-answers through the parts of the query (a query operator or a set thereof) that were responsible for the answers omission.(我们主要通过查询的部分(查询操作符或其集合)来解释非答案,这些部分是导致答案缺失的原因。)
系统架构
NLProveNAns is implemented in JAVA, and runs on Windows 10. It uses MySQL as its underlying database system and uses two previously developed system prototypes, namely NaLIR and NLProv.(NLProveNAns 是用JAVA实现的,运行在Windows 10上。它使用MySQL作为底层数据库系统,并使用了两个以前开发的系统原型,即NaLIR和NLProv。)
- First, the user inputs a query in natural language and chooses between the two provenance models. This query is fed to the modified NaLIR implementation which parses the sentence and generates an SQL query. This query is then evaluated by SelP.(首先,用户以自然语言输入查询并在两个来源模型之间进行选择。 这个查询被提供给修改后的 NaLIR 实现,它解析句子并生成一个 SQL 查询。 这个查询然后由 SelP 评估。)
- The user is then presented with the results of the initial query along with natural language explanations for the result set tuples, generated by NLProv.(然后向用户呈现初始查询的结果以及由 NLProv 生成的结果集元组的自然语言解释。)
- After viewing the results, the user specifies a why-not question.(在查看结果之后,用户指定一个“为什么不”的问题。)
- This question is parsed by NLProveNAns and using the information stored while processing the initial query, it computes an answer for the why-not question using the chosen provenance model, and uses this answer to produce a word highlighting answer.(这个问题由 NLProveNAns 解析,并使用在处理初始查询时存储的信息,它使用选择的出处模型计算为什么不的问题的答案,并使用这个答案生成一个单词突出显示的答案。)
如何做实验
demonstrate the system prototype using a sample of the publications database of Microsoft Academic Search.(使用 Microsoft Academic Search的出版物数据库示例来演示系统原型)