Natural language is the ideal choice of users when they are expressing
their search needs. However, although it is quite effective to have natural
language interfaces in web search, it is challenging to have those interfaces
in Relational Database Management Systems (RDBMSs). In
RDBMSs, the querying language is SQL, which is a powerful and flexible language
to express user intent. However, casual users cannot query over relational
databases due to need of having technical background. Therefore, natural
language to SQL translation has become quite crucial to provide such interfaces
to the users.
Having a natural language query interface is the first step to make
towards providing user friendly environment to the users for their search
experience. For this purpose, there are well known paradigms introduced in
information retrieval field which try to make the search more user friendly.
However, these paradigms require certain modifications to be applicable in
NLIDBs.
In this project, we aim to develop different information retrieval
based algorithms to be used in translation context to make it more user
friendly and more effective. The information retrieval paradigms involved in
our algorithms are listed below:
1. Query auto-completion to resolve ambiguities inside the query to
increase translation accuracy,
2. Context aware query recommendation to suggest candidate queries to
the users given the context to improve search experience,
3. Information retrieval based ranking to rank queries to be listed as
candidate translations given a query to substitute for state-of-the-art
translation.
The algorithms will be integrated to a natural language to SQL
translation system from the literature.