TEGRA is a research system developed at MSR, to automatically segment data that may not come with explitic column-delimiters (e.g., lists/tables on the web, and data embedded in texts), into relational tables.
We develop an algorithm based on a concept we call ``global-alignment'', which can perform this task at high accuracy, as described in our SIGMOD 15 paper TEGRA: Table Extraction by Global Record Alignment in detail.
This repo has the code used in our TEGRA paper, which is now released under MIT licsense.