Natural language processing for extracting conveyance graphs

• US 9,251,139 B2
• Filed: 04/08/2015
• Issued: 02/02/2016
• Est. Priority Date: 04/08/2014
• Status: Active Grant

• Current Assignee: Titleflow LLC
• Original Assignee: Titleflow LLC
• Inventors: Bateman, David T., Phillips, Aaron, Plagens, Andrew E., Drennan, J. Charles, Langdon, Wendell H.
• Primary Examiner(s): Desir, Pierre-Louis
• Assistant Examiner(s): KOVACEK, DAVID M
• Application Number:
  US14/681,719
• Publication Number:
  US 20150286630A1
• Time in Patent Office: 300 Days
• Field of Search: 704 1- 9, 704/10, 704/231, 704/270, 705/1.1, 705/4, 705/7.11, 705 727- 728, 705/14.11, 705 30- 45, 705310-316, 705/342
• US Class Current: 1/1
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 40/237 Lexical tools
• CPC Class Codes: G06F 40/279 Recognition of textual entities
• CPC Class Codes: G06Q 30/018 Certifying business or products
• CPC Class Codes: G06Q 30/0645 Rental transactions; Leasing transactions
• CPC Class Codes: G06Q 50/00 Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism healthcare informatics G16H
• CPC Class Codes: G06Q 50/02 Agriculture; Fishing; Forestry; Mining
• CPC Class Codes: G06Q 50/163 Real estate management
• CPC Class Codes: G06Q 50/165 Land development
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables
• CPC Class Codes: G06V 30/418 Document matching, e.g. of document images
• CPC Class Codes: G10L 21/00 Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility G10L19/00 takes precedence

Abstract

Provided is a process for extracting conveyance records from unstructured text documents, the process including: obtaining, with one or more processors, a plurality of documents describing, in unstructured form, one or more conveyances of interest in real property; determining, with one or more processors, for each of the documents, a respective jurisdiction; selecting, with one or more processors, from a plurality of language processing models for the English language, a respective language processing model for each of the documents based on the respective determined jurisdiction; extracting, with one or more processors, for each of the documents, a plurality of structured conveyance records from each of the plurality of documents by applying the language processing model selected for the respective document based on the jurisdiction associated with the document; and storing, with one or more processors, the extracted, structured conveyance record in memory.

Claims

1. A method for extracting conveyance records from unstructured text documents, the method comprising:

obtaining, with one or more processors, a plurality of scanned, optical-character-recognized (OCR) documents, each having OCR-produced English language text describing, in unstructured form, one or more conveyances of interest in real property, wherein each document is associated with metadata identifying a jurisdiction in which the respective real property is located;

determining, with one or more processors, for each of the documents, a respective jurisdiction based on the metadata;

selecting, with one or more processors, from a plurality of language processing models for the English language, a respective language processing model for each of the documents based on the respective determined jurisdiction,

wherein a first language processing model is selected for at least some of the documents associated with a first jurisdiction and a second language processing model, different from the first language processing model, is selected for at least some of the documents associated with a second jurisdiction that is different from the first jurisdiction, and

wherein each language processing model is configured to extract structured data from unstructured text, and wherein each language processing model is configured to detect different terminology used in different jurisdictions with different frequencies;

extracting, with one or more processors, for each of the documents, from the respective OCR-produced English language text describing, in unstructured form, one or more conveyances of interest in real property, a plurality of structured conveyance records from each of the plurality of documents by applying the language processing model selected for the respective document based on the jurisdiction associated with the document,

wherein each extracted conveyance record identifies land in which an interest is conveyed by the respective document, identifies a grantor of the conveyance, identifies a grantee of the conveyance, and identifies the interest conveyed; and

storing, with one or more processors, the extracted, structured conveyance records in memory.

2. The method of claim 1, comprising:

determining, with one or more processors, for each of the documents, a respective date based on the metadata; and

selecting, with one or more processors, from a plurality of language processing models for the English language, a respective language processing model for each of the documents based on the respective determined date and/or jurisdiction.

3. The method of claim 1, wherein the plurality of language processing models comprises a language processing model containing hand-coded rules.

4. The method of claim 1, wherein the plurality of language processing models comprises a language processing model trained, at least in part, on previous analyses of documents.

5. The method of claim 1, comprising:

obtaining, with one or more processors, a training set of conveyance records and documents from which the conveyance records were manually populated; and

training, with one or more processors, at least in part, a language processing model based on the training set.

6. The method of claim 5, comprising:

subsampling, with one or more processors, into a plurality of subsamples, the training set of conveyance records and documents from which the conveyance records were manually populated;

training, with one or more processors, a plurality of language processing models on respective ones of the subsamples; and

aggregating the plurality of language processing models into a single aggregate language processing model.

7. The method of claim 1, wherein the one or more language processing models comprise a language processing model trained on a training set of conveyance records and documents from which the conveyance records were extracted.

8. The method of claim 1, wherein the extracted conveyance record is presented in one or more of a title flow form, a tabulation of ownership form, a title chain form, a runsheet form, or a combination thereof.

9. The method of claim 1, wherein the extracted conveyance record includes a plurality of parameters, including an identifier of a grantor, an identifier of a grantee, an identifier of an interest conveyed, and wherein at least one of the parameters is associated with a link to a supporting document from which the parameter is extracted.

10. The method of claim 1, wherein extracting a conveyance record comprises:

detecting a conveyance record parameter;

inserting an anchor into a PDF file depicting a supporting document in which the conveyance record parameter is detected; and

associating a link to the PDF file and the anchor with the detected parameter in the conveyance record.

11. The method of claim 1, comprising:

receiving, from a client device, a request to view a given conveyance record;

sending, to the client device, the given conveyance record and an identifier of a link to a document from which at least part of the given conveyance record was extracted;

receiving, from the client device, a request for content corresponding to the identifier of the link; and

in response to receiving the request for content, sending, to the client device, at least a portion of the document from which at least part of the given conveyance record was extracted.

12. The method of claim 11, comprising:

receiving updates to at least some of the conveyance records;

calculating a score for a given one of the language processing models based on an amount of the updates to conveyance records having data extracted by the given one of the language processing models; and

ranking the given one of the language processing models relative to other ones of the language processing models based on the score.

13. The method of claim 12, comprising:

selecting a new language processing model based on the ranking; and

extracting a parameter of a conveyance records with the selected language processing model.

14. The method of claim 1, comprising:

generating, with one or more processors, recommendations and/or alerts based on the extracted conveyance records.

15. The method of claim 1, wherein at least some of the conveyance records form a directed ownership graph having a tree structure with interest holders serving as nodes of the graph and conveyances serving as edges of the graph, and wherein the method comprises:

recursing at least part of the ownership tree, determining for at least some of the edges whether a corresponding node from which the respective edge extends possess the interest conveyed by the respective edge.

16. The method of claim 15, wherein determining for at least some of the edges whether a corresponding node from which the respective edge extends possess the interest conveyed by the respective edge comprises:

tracing the interest conveyed by the respective edge to a root node of the tree structure.

17. The method of claim 1, wherein at least some of the conveyance records form a directed ownership graph having a tree structure with interest holders serving as nodes of the graph and conveyances serving as edges of the graph, and wherein the method comprises:

identifying leaf nodes of the graph; and

determining whether the leaf nodes sum to 100% ownership of an interest in property.

18. The method of claim 17, comprising:

determining that the some leaf nodes do not sum to 100% ownership of an interest in property; and

in response to the determination, sending a warning that title is not fully documented.

19. The method of claim 1, comprising:

obtaining a training set of documents having tagged text from which corresponding parameters of conveyance records are extracted;

calculating frequencies with which a plurality of n-grams occurs within an amount of tokens from the tagged text; and

assigning weights to the n-grams in a language model based on the calculated frequencies.

20. A system configured to extract conveyance records from unstructured text documents, the system comprising:

one or more computer processors; and

storage media, storing machine-readable instructions that, when executed by at least some of the one or more processors, cause operations comprising;

obtaining a plurality of scanned, optical-character-recognized (OCR) documents, each having OCR-produced English language text describing, in unstructured form, one or more conveyances of interest in real property, wherein each document is associated with metadata identifying a jurisdiction in which the respective real property is located;

determining, for each of the documents, a respective jurisdiction based on the metadata;

selecting from a plurality of language processing models for the English language, a respective language processing model for each of the documents based on the respective determined jurisdiction,

wherein a first language processing model is selected for at least some of the documents associated with a first jurisdiction and a second language processing model, different from the first language processing model, is selected for at least some of the documents associated with a second jurisdiction that is different from the first jurisdiction, and

wherein each language processing model is configured to extract structured data from unstructured text, and wherein each language processing model is configured to detect different terminology used in different jurisdictions with different frequencies;

extracting, for each of the documents, from the respective OCR-produced English language text describing, in unstructured form, one or more conveyances of interest in real property, a plurality of structured conveyance records from each of the plurality of documents by applying the language processing model selected for the respective document based on the jurisdiction associated with the document,

wherein each extracted conveyance record identifies a plot of land in which an interest is conveyed by the respective document, identifies a grantor of the conveyance, identifies an grantee of the conveyance, and identifies the interest conveyed; and

storing the extracted structured conveyance records in memory.

21. The system of claim 20, the operations comprising:

determining for each of the documents, a respective date based on the metadata; and

selecting from a plurality of language processing models for the English language, a respective language processing model for each of the documents based on the respective determined date and/or jurisdiction.

22. The system of claim 20, wherein the plurality of language processing models comprises a language processing model containing hand-coded rules.

23. The system of claim 20, wherein the plurality of language processing models comprises a language processing model trained, at least in part, on previous analyses of documents.

24. The system of claim 20, the operations comprising:

obtaining a training set of conveyance records and documents from which the conveyance records were manually populated; and

training, with one or more processors, at least in part, a language processing model based on the training set.

25. The system of claim 24, the operations comprising:

subsampling into a plurality of subsamples, the training set of conveyance records and documents from which the conveyance records were manually populated;

training a plurality of language processing models on respective ones of the subsamples; and

aggregating the plurality of language processing models into a single aggregate language processing model.

26. The system of claim 20, wherein the one or more language processing models comprise a language processing model trained on a training set of conveyance records and documents from which the conveyance records were extracted.

27. The system of claim 20, wherein the extracted conveyance record is presented in one or more of a title flow form, a tabulation of ownership form, a title chain form, a runsheet form, or a combination thereof.

28. The system of claim 20, wherein the extracted conveyance record includes a plurality of parameters, including an identifier of a grantor, an identifier of a grantee, an identifier of an interest conveyed, and wherein at least one of the parameters is associated with a link to a supporting document from which the parameter is extracted.

29. The system of claim 20, wherein extracting a conveyance record comprises:

detecting a conveyance record parameter;

inserting an anchor into a PDF file depicting a supporting document in which the conveyance record parameter is detected; and

associating a link to the PDF file and the anchor with the detected parameter in the conveyance record.

30. The system of claim 20, the operations comprising:

receiving, from a client device, a request to view a given conveyance record;

sending, to the client device, the given conveyance record and an identifier of a link to a document from which at least part of the given conveyance record was extracted;

receiving, from the client device, a request for content corresponding to the identifier of the link; and

in response to receiving the request for content, sending, to the client device, at least a portion of the document from which at least part of the given conveyance record was extracted.