Verifying Text Summaries of Relational Datasets

As a result of widespread use of internet as an outlet to reach the information, people are struggling with a bizarre number of false claims and half-truths. In this environment, people need a way to check the claims by inspecting the relevant data and documents but the amount of claims that are made and the speed of transmission of these claims in the cyberspace are not proportional to humans’ capability to check the truthiness of these claims. In this project, I have tried to implement a query interface which is based on text summaries of relational data sets. The main goal of the tool is that it tries to focus on natural language claims and to translate them into an SQL query and check the query result with the claimed information. As a result of this process, the tool finds the candidate claims which are seemed to be inconsistent with the information in the database. Basically, the system is based on a probabilistic model which maps the claimed keywords to a probability distribution which represents the correctness of a claim by using query translations and executing these queries over the dataset provided. By this way, claimed facts in a document is checked against the actual data.

Poster: https://drive.google.com/file/d/1shT9EJs3FlKuTW3h9oK_BMdLPcTUHldW/view?usp=sharing

Video: https://drive.google.com/file/d/1D_Qkjk2ABapjG8AqtpvF7Kjfs8V_gyR5/view?usp=sharing