Graph based re-composition of document fragments for name entity recognition under exploitation of enterprise databases

1. A non-transitory computer-readable storage medium tangibly storing machine-readable instructions thereon, which when executed by the machine, cause the machine to:

receive a document containing text or unstructured data, wherein the document is a representation of a physical business object stored in a document storage unit;

identify and extract a plurality of text segments and structure of the document using a set of tagging and natural language processing techniques;

query the extracted plurality of text segments against a set of inverted indices, wherein the set of inverted indices comprises attributes of a set of predefined data structure entities and associated relationships based on weight values in an entity model structure;

for a text segment from the extracted plurality of text segments, store matching attributes of the set of predefined data structure entities and associated relationships;

construct a set of entity graphs including a plurality of entity nodes connected by a weighted edge representing the matching attributes and associated relationships between the plurality of text segments and the attributes of the set of predefined data structure entities, between the attributes and the set of predefined data structure entities, and between data structure entities in the set of predefined data structure entities;

rank the set of entity graphs based on a sum of the weight values associated with the matching attributes and associated relationships; and

select higher scored one or more entity graphs of the set of entity graphs based on the ranking.

2. The non-transitory computer-readable storage medium of claim 1 having instructions that when executed further cause the machine to:

parse the document via a parser; and

in response to parsing the document, extract the text segment of the document.

3. The non-transitory computer-readable storage medium of claim 1 having instructions that when executed further cause the machine to:

apply text recognition techniques to the document; and

in response to applying the text recognition techniques, obtain the structure of the document.

4. The non-transitory computer-readable storage medium of claim 1 having instructions that when executed further cause the machine to:

determine the matching attributes and associated relationships for at least one of the queried plurality of text segments; and

determine the weight values for the matching attributes and associated relationships between the at least one of the queried plurality of text segments and the set of predefined data structure entities to construct the set of entity graphs.

5. The non-transitory computer-readable storage medium of claim 4 having instructions that when executed further cause the machine to:

create a set of unambiguous graph variants for each graph in the set of entity graphs; and

create a set of groups, each group including a subset of the set of entity graphs and representing interpretations of the at least one of the queried plurality of text segments.

6. The non-transitory computer-readable storage medium of claim 5 having instructions that when executed further cause the machine to:

sum the weight values for the matching attributes and associated relationships in each entity graph of the set of groups; and

select an entity graph from each group of the set of groups, the entity graph having a best weight value and representing a best interpretation of the at least one of the queried plurality of text segments.

7. The non-transitory computer-readable storage medium of claim 1 , wherein the document comprises an invoice, the predefined data structure entities comprises attributes of a purchase order, and the set of entity graphs comprises the matching attributes and associated relationships between the invoice and the purchase order.

8. A computing system comprising:

a set of documents stored in a document storage unit, a document from the set of documents comprising text or unstructured data;

a database storage unit that stores a set of inverted indices including predefined data structure entities organized in an entity model structure, wherein an entity from the set of predefined entities has a set of characteristics including attributes and associated relationships based on weight values; and

a processor in communication with the database storage unit and the document storage unit, the processor to:

identify and extract a plurality of text segments and structure of the document using a set of tagging and natural language processing techniques;

query the extracted plurality of text segments against a set of inverted indices, wherein the set of inverted indices comprises attributes of a set of predefined data structure entities and associated relationships based on weight values in an entity model structure;

construct a set of entity graphs including a plurality of entity nodes connected by a weighted edge representing the matching attributes and associated relationships between the plurality of text segments and the attributes of the set of predefined data structure entities, between the attributes and the set of predefined data structure entities, and between entities in the set of predefined data structure entities;

rank the constructed set of entity graphs based on a sum of the weight values associated with the matching attributes and associated relationships; and

select higher scored one or more entity graphs of the set of entity graphs based on the ranking.

9. The computer system of claim 8 further comprising a parser to parse the document and to extract the text segment of the document.

10. The computer system of claim 8 further comprising a package of text recognition techniques to obtain the structure of the document.

11. The computer system of claim 8 further comprising:

the set of entity graphs built upon the matching attributes and associated relationships, the set of entity graphs including a plurality of possible interpretations of the plurality of text segments;

a set of unambiguous graph variants for each entity graph in the set of entity graphs; and

a set of groups, each group including a subset of the set of entity graphs and representing interpretations of the plurality of text segments.

12. The computer system of claim 8 further comprising an integrated data storage unit that includes a set of structures of integrated data from matching the set of predefined data structure entities with the plurality of text segments.

13. The computer system of claim 8 , wherein the document comprises an invoice, the predefined data structure entities comprises attributes of a purchase order, and the set of entity graphs comprises the matching attributes and associated relationships between the invoice and the purchase order.

14. A computer implemented method comprising:

receiving a document containing text or unstructured data, wherein the document is a representation of a physical business object stored in a document storage unit;

identifying and extracting a plurality of text segments and structure of the document using a set of tagging and natural language processing techniques;

querying the extracted plurality of text segments against a set of inverted indices, wherein the set of inverted indices comprises attributes of a set of predefined data structure entities and associated relationships based on weight values in an entity model structure;

for a text segment from the extracted plurality of text segments, storing matching attributes of the set of predefined data structure entities and associated relationships;

constructing a set of entity graphs including a plurality of entity nodes connected by a weighted edge representing the matching attributes and associated relationships between the plurality of text segments and the attributes of the set of predefined data structure entities, between the attributes and the set of predefined data structure entities, and between data structure entities in the set of predefined entities;

ranking the set of entity graphs based on a sum of the weight values associated with the matching attributes and associated relationships; and

selecting higher scored one or more entity graphs of the set of entity graphs based on the ranking.

15. The method of claim 14 further comprising:

parsing the document via a parser; and

in response to parsing the document, extracting the text segment of the document.

16. The method of claim 14 further comprising:

applying text recognition techniques to the document; and

in response to applying the text recognition techniques, obtaining the structure of the document.

17. The method of claim 14 , further comprising:

determining the matching attributes and associated relationships for at least one of the queried plurality of text segments; and

determining the weight values for the matching attributes and associated relationships between the at least one of the queried plurality of text segments and the set of predefined data structure entities to construct the set of entity graphs.

18. The method of claim 17 , further comprising:

creating a set of unambiguous graph variants for each entity graph in the set of entity graphs; and

creating a set of groups, each group including a subset of the set of entity graphs and representing interpretations of the at least one of the queried plurality of text segments.

19. The method of claim 18 , further comprising:

summing the weight values for the matching attributes and associated relationships in each entity graph of the set of groups; and

selecting entity graph from each group of the set of groups, the entity graph having a best weight value and representing a best interpretation of the at least one of the queried plurality of text segments.

20. The method of claim 14 , wherein the document comprises an invoice, the predefined data structure entities comprises attributes of a purchase order, and the set of entity graphs comprises the matching attributes and associated relationships between the invoice and the purchase order.