Searchable table extraction

A computer system comprising:

• one or more data processors;
• and
• one or more non-transitory computer-readable storage media storing instructions that, when executed by the one or more data processors, cause the one or more data processors to perform a method including:
• extracting, by using a table extraction model on an input document image, a plurality of tables as a plurality of objects, the plurality of tables including a plurality of cells, respectively;
• receiving, as an output the table extraction model, the plurality of tables in an extensible Markup Language (XML) format and outputting cell level information of the plurality of cells;
• filtering the plurality of tables based on a plurality of keywords and a condition comprising one from among AND and OR, to reduce a number of the plurality of tables, the filtering further including:
• for each of the plurality of tables, determining whether the table comprises at least one from among a first keyword and a second keyword among the plurality of keywords,
• for the condition being AND, based on the table containing the first keyword and the second keyword, outputting the table, as a filtered table among filtered tables, and, based on the table not containing the first keyword and the second keyword, not outputting the table, and
• for the condition being OR, based on the table containing the first keyword or the second keyword, outputting the table, as a filtered table among the filtered tables;
• outputting the filtered tables that include the at least one from among the first keyword and the second keyword based on the condition, among the plurality of tables;
• executing a query on the filtered tables, the query including one or more keys composed based on the cell level information associated with the filtered tables;
• identifying one or more cells among the plurality of cells based on the one or more keys;
• and
• providing target content that corresponds to the one or more cells, wherein the target content is a response to the query.

A computer-program product tangibly embodied in one or more non-transitory computer-readable storage media including instructions configured to cause one or more data processors to perform a method including:

• extracting, by using a table extraction model on an input document image, a plurality of tables as a plurality of objects, the plurality of tables including a plurality of cells, respectively;
• receiving, as an output the table extraction model, the plurality of tables in an eXtensible Markup Language (XML) format and outputting cell level information of the plurality of cells;
• filtering the plurality of tables based on a plurality of keywords and a condition comprising one from among AND and OR, to reduce a number of the plurality of tables, the filtering further including:
• for each of the plurality of tables, determining whether the table comprises at least one from among a first keyword and a second keyword among the plurality of keywords,
• for the condition being AND, based on the table containing the first keyword and the second keyword, outputting the table, as a filtered table among filtered tables, and, based on the table not containing the first keyword and the second keyword, not outputting the table, and
• for the condition being OR, based on the table containing the first keyword or the second keyword, outputting the table, as a filtered table among the filtered tables;
• outputting the filtered tables that include the at least one from among the first keyword and the second keyword based on the condition, among the plurality of tables;
• executing a query on the filtered tables, the query including one or more keys composed based on the cell level information associated with the filtered tables;
• identifying one or more cells among the plurality of cells based on the one or more keys;
• and
• providing target content that corresponds to the one or more cells, wherein the target content is a response to the query.