PhD Research Seminar: Synthesizing database queries from natural language specifications with neural networks
When: March 2, 9:30–10:50
First talk: SPARQLing Database Queries from Intermediate Question Decompositions
Speaker: Irina Saparina, second-year PhD student, Faculty of Computer Science
This talk will be devoted to the task of generating a database query from a natural-language description of the user’s intent. Most approaches currently generate database queries using encoder-decoder neural networks trained with supervision, requiring a fully annotated training set. Datasets for this task consist of questions with corresponding queries and databases, which are challenging to collect. I will present our approach that uses intermediate question representations to produce executable queries without relying on query annotation.
Second talk: Searching for Better Database Queries from Predictions of Neural Models
Speaker: Ramil Yarullin, third-year PhD student, Faculty of Computer Science
The task of generating a database query from a question in natural language suffers from ambiguity and insufficiently precise description of the goal. The problem is amplified when the system needs to generalize to databases unseen at training. We will discuss how to search for better database queries using external criteria that evaluate the generated queries. The criteria vary from checking that a query executes without errors to verifying the query on a set of tests.