研究內容

• In this demo, we present NaLIR, a generic interactive natural language interface for querying relational databases.（在這個演示中，我們展示了NaLIR，一個用於查詢關係資料庫的通用互動式自然語言介面）
• NaLIR can accept a logically complex English language sentence as query input. This query is first translated into a SQL query. Then, an RDBMS is used to evaluate the translated SQL query and return the results to the user.
  （它接受英語句子作為查詢輸入之後，先將此查詢轉換為一個SQL 查詢，然後使用 RDBMS 評估，並將結果返回給使用者）

系統設計思路

1. We use an off-the-shelf natural language parser to obtain the linguistic understanding (represented by a parse tree) of the input sentence.（我們使用現成的自然語言解析器來獲得輸入句子的語言理解(由解析樹表示)）
2. transform the linguistic understanding to a database’s understanding. In order to make sure that the sentence is correctly understood, we explain the database’s understanding back to the user in natural language.
   （將語言理解轉換為資料庫理解，此過程中為確保正確理解句子，會用自然語言向用戶解釋資料庫理解）
3. the understanding is translated to a SQL query statement and evaluated by an RDBMS. （將資料庫理解轉換為 SQL 查詢語句，並由 RDBMS 進行評估）

系統架構

• The query translation part, which includes parse tree node mapper, structure adjustor and translator, is responsible for making full use of the exist
  ing information provided by the user and generate the correct SQL query statement.（查詢翻譯部分，包括解析樹節點對映器、結構調整器和翻譯器，負責充分利用使用者提供的現有資訊，生成正確的SQL查詢語句）
• The interactive communicator, is responsible for obtaining more information from the user and making sure that the system correctly understands her query intent.（互動通訊部分：負責從使用者那裡獲得更多資訊，並確保系統正確理解使用者的查詢意圖）
  • it explains to the user how her query is processed（它向用戶解釋如何處理查詢）
  • it shows multiple understandings for the user to choose from, which facilitates the user in eliminating misunderstandings（它顯示了多種理解供使用者選擇，方便使用者消除誤解）

研究基礎

• There is a stream of such research on keyword search, in which, given a set of keywords, instead of finding the data relevant to these keywords, they try to interpret the query intent behind the keywords in the view of a formal query language.（有一種關於關鍵詞搜尋的研究，他們給出一組關鍵詞，不是尋找與這些關鍵詞相關的資料，而是試圖從一種形式化的查詢語言的角度來解釋關鍵詞背後的查詢意圖）

如何做實驗

• show the use of NaLIR against a number of real application scenarios including Microsoft Academic Search, Yahoo! Movies, and DBLP collection.（在一些真實的應用場景中展示NaLIR的使用，這些場景包括……）
• The demonstration will consist of two phases（演示將包括兩個階段）
  • the user will run the queries in the query log, which contains a set of successfully processed NL queries. The chosen NL query can also serve as a query template for the user to make small modifications.（使用者將在查詢日誌中執行查詢，其中包含一組成功處理的NL查詢。所選的NL查詢還可以作為查詢模板，供使用者進行小的修改）
  • the user will be free to run their own queries.（使用者可以自由地執行自己的查詢）