Identifying entities in semi-structured content

• US 10,353,905 B2
• Filed: 04/24/2015
• Issued: 07/16/2019
• Est. Priority Date: 04/24/2015
• Status: Active Grant

• Current Assignee: Salesforce.com, Inc.
• Original Assignee: Salesforce.com, Inc.
• Inventors: Ramachandran, Govardana Sachithanandam, Mustafa, Hawro
• Primary Examiner(s): Truong, Cam Y T
• Application Number:
  US14/695,996
• Publication Number:
  US 20160314123A1
• Time in Patent Office: 1,544 Days
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/81 Indexing, e.g. XML tags; Data structures therefor; Storage structures
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Identifying entities in semi-structured content is described. A system assigns a corresponding entity type based on a corresponding entity type score for each token in a sequence of tokens in semi-structured content, based on multiple entity types, wherein each token is a corresponding character set. The system assigns a corresponding boundary type based on a corresponding boundary type score for each token in the sequence of tokens, based on a begin boundary type or a continue boundary type. The system identifies an entity based on a corresponding entity type score and a corresponding boundary type for each token in the sequence of tokens. The system outputs the sequence of tokens as an identified set of entities based on the identified entity.

Claims

1. A system for identifying entities in semi-structured content, the system comprising:

one or more processors; and

a non-transitory computer readable medium storing a plurality of instructions, which when executed, cause the one or more processors to;

identify a sequence of tokens in the semi-structured content based on assigning an information score to each token;

assign, by a first layer of classifications that is executed by a machine learning model of a database system, an entity type for each token in the sequence of tokens based on an entity score representing a probability that the token corresponds to the entity type, the entity type being one of a plurality of entity types and the entity score being a maximum score for correspondence between the token and any of the plurality of entity types;

assign, by a second layer of classifications that is executed by the machine learning model, a structure score for each token in the sequence of tokens based on the token matching one of a plurality of structure types;

re-assign, by the second layer, the corresponding entity type and corresponding entity score for each token in the sequence of tokens matching one of the structure types;

assign, by a third layer of classifications that is executed by the machine learning model, a boundary type for each token in the sequence of tokens based on a boundary type score, the boundary type being one of a begin boundary type and a continue boundary type;

identify, by a fourth layer of classifications that is executed by the machine learning model, an entity based on;

i) the entity type and the boundary type for each token, and;

ii) the structure score for each token; and

output the sequence of tokens as an identified set of entities based on the identified entity.

2. The system of claim 1, comprising further instructions, which when executed, cause the one or more processors to identifying contiguous tokens with collective information scores, which is higher than another information score for other contiguous tokens, as the sequence of tokens.

3. The system of claim 1, wherein assigning the entity type to each token in the sequence of tokens in the semi-structured content comprises identifying the sequence of tokens as corresponding to an entity type assigned to a last token of the token sequence of tokens.

4. The system of claim 1, wherein assigning a structure score to each token in the sequence of tokens is based on one of an email structure type, a phone structure type, and a website structure type, and wherein the boundary type for each token in the sequence of tokens is further based on the structure score for each token in the sequence of tokens.

5. The system of claim 1, comprising further instructions, which when executed, cause the one or more processors to merge adjacent tokens corresponding to a same entity type to generate a single compound entity type.

6. The system of claim 1, wherein the entity comprises a sequence of tokens from a same line of the semi-structured content.

7. A computer program product comprising a non-transitory computer-readable medium storing computer-readable program code which when executed by one or more processors, cause the one or more processors to:

identify, by a database system, a sequence of tokens in semi-structured content based on assigning an information score to each token;

assign, by a first layer of classifications that is executed by a machine learning model of the database system, an entity type for each token in the sequence of tokens based on an entity score representing a probability that the token corresponds to the entity type, the entity type being one of a plurality of entity types and the entity score being a maximum score for correspondence between the token and any of the plurality of entity types;

assign, by a second layer of classifications that is executed by the machine learning model, a structure score for each token in the sequence of tokens based on the token matching one of a plurality of structure types;

re-assign, by the second layer, the corresponding entity type and corresponding entity score for each token in the sequence of tokens matching one of the structure types;

assign, by a third layer of classifications that is executed by the machine learning model, a boundary type for each token in the sequence of tokens based on a boundary type score, the boundary type being one of a begin boundary type and a continue boundary type;

identify, by a fourth layer of classifications that is executed by the machine learning model, an entity based on;

i) the entity type and the boundary type for each token, and;

ii) the structure score for each token; and

output, by the database system, the sequence of tokens as an identified set of entities based on the identified entity.

8. The computer program product of claim 7, wherein the program code comprises instructions to identify, by the database system, contiguous tokens with collective information scores, which is higher than another information score for other contiguous tokens, as the sequence of tokens.

9. The computer program product of claim 7, wherein assigning the entity type to each token in the sequence of tokens in the semi-structured content comprises identifying the sequence of tokens as corresponding to an entity type assigned to a last token of the sequence of tokens.

10. The computer program product of claim 7, wherein the program code comprises instructions wherein assigning a structure score to each token in the sequence of tokens is based on one of an email structure type, a phone structure type, and a website structure type, and wherein the boundary type for each token in the sequence of tokens is further based on the structure score for each token in the sequence of tokens.

11. The computer program product of claim 7, wherein the program code comprises instructions to merge, by the database system, adjacent tokens corresponding to a same entity type to generate a single compound entity type.

12. The computer program product of claim 7, wherein the entity comprises a sequence of tokens from a same line of the semi-structured content.

13. A computer-implemented method for identifying entities in semi-structured content in an on-demand database system, the method comprising:

identifying, by a database system, a sequence of tokens in the semi-structured content based on assigning an information score to each token;

assigning, by a first layer of classifications that is executed by a machine learning model of the database system, an entity type for each token in the sequence of tokens based on an entity score representing a probability that the token corresponds to the entity type, the entity type being one of a plurality of entity types and the entity score being a maximum score for correspondence between the token and any of the plurality of entity types;

assigning, by a second layer of classifications that is executed by the machine learning model of the database system, a structure score for each token in the sequence of tokens based on the token matching one of a plurality of structure types;

re-assigning, by the second layer, the corresponding entity type and the corresponding entity score for each token in the sequence of tokens matching one of the structure types;

assigning, by a third layer of classifications that is executed by the machine learning model of the database system, a boundary type for each token in the sequence of tokens based on a boundary type score, the boundary type being one of a begin boundary type and a continue boundary type;

identifying, by a fourth layer of classifications that is executed by the machine learning model of the database system, an entity based on;

i) the entity type and the boundary type for each token, and;

ii) the structure score for each token; and

outputting, by the database system, the sequence of tokens as an identified set of entities based on the identified entity.

14. The method of claim 13, wherein the method further comprises identifying, by the database system, contiguous tokens with collective information scores, which is higher than another information score for other contiguous tokens, as the sequence of tokens.

15. The method of claim 13, wherein assigning the entity type to each token in the sequence of tokens in the semi-structured content comprises identifying the sequence of tokens as corresponding to an entity type assigned to a last token of the sequence of tokens.

16. The method of claim 13, wherein assigning, by the database system, a structure score to each token in the sequence of tokens is based on one of an email structure type, a phone structure type, and a website structure type, and wherein the boundary type for each token in the sequence of tokens is further based on the structure score for each token in the sequence of tokens.

17. The method of claim 13, wherein the method further comprises merging, by the database system, adjacent tokens corresponding to a same entity type to generate a single corresponding entity type.