# US 12,001,486 B2 · Identifying reference data in a source data set > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,001,486 B2 - Document ID: US12001486B2 - Status: Granted - Kind: B2 - Application number: 16/583,372 - Filed: 2019-09-26 - Published: 2021-04-01 - Granted: 2024-06-04 - Earliest claimed: 2019-09-26 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Martin Oberhofer; Joerg Rehr; Ivan Matthew Milman; Sushain Pandit - Web page: https://patentgrep.com/patent/US12001486B2 ## CPC - G06F 16/906: Clustering; Classification - G06F 16/907: Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually - G06F 18/2148: {characterised by the process organisation or structure, e.g. boosting cascade} - G06F 18/2178: {based on feedback of a supervisor} - G06F 40/106: Display of layout of documents; Previewing - G06F 40/216: using statistical methods - G06F 40/258: Heading extraction; Automatic titling; Numbering - G06F 40/279: Recognition of textual entities - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06N 3/008: based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour - G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N 20/00: Machine learning - G06N 20/20: Ensemble learning - G06V 30/19173: {Classification techniques} - G06V 30/412: Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract A method, system, and computer program product for identifying reference data values in a source data set. The method may include inputting a block of attribute values to a predefined machine learning model. The method may also include receiving an indication of a presentation layout of the block of the attribute values and an associated reference data extraction method. The method may also include determining a reading direction of the block of values. The method may also include identifying one or more inspection areas in the reading direction of the block of values. The method may also include determining sets of the one or more inspection areas that share a common presentation feature. The method may also include identifying tokens in an inspection area. The method may also include determining if the inspection area includes reference data values. The method may also include outputting the reference data values. ## Claims ### Claim 1 A computer-implemented method comprising: inputting a block of attribute values to a predefined machine learning model; receiving, from the machine learning model, an indication of a presentation layout of the block of the attribute values and an associated reference data extraction method; determining a reading direction of the block of values; identifying one or more inspection areas, from the block of values, in the reading direction of the block of values; determining a set of the one or more inspection areas that shares a common presentation feature; identifying tokens in the set of the one or more inspection areas; determining if the set of the one or more inspection areas comprises reference data values, wherein the determining uses a frequency distribution of the identified tokens; and outputting the reference data values. ### Claim 2 (depends on 1) The method of claim 1, wherein the identified tokens of the set of the one or more inspection areas are a same type. ### Claim 3 (depends on 1) The method of claim 1, wherein the presentation layout is a table, further comprising: testing if each inspection area in a first row and a first column of a table comprises predefined header keywords. ### Claim 4 (depends on 3) The method of claim 3, further comprising: determining that the first row comprises at least one predefined header keyword; and marking the first row as a header row. ### Claim 5 (depends on 3) The method of claim 3, further comprising: determining that the first column comprises at least one predefined header keyword; and marking the first column as a header column. ### Claim 6 (depends on 1) The method of claim 1, wherein the block of values comprises at least one of: a table, an image, and a structure between hashtags of a data stream. ### Claim 7 (depends on 1) The method of claim 1, wherein outputting the reference data values comprises pushing the reference data values to a reference data management system. ### Claim 8 (depends on 1) The method of claim 1, wherein the presentation layout is a table, and wherein determining the reading direction of the block of values comprises: processing a first row and a first column of the table, the processing comprising: determining if the first row and the first column comprise header information; and in response to determining that at least one of the first row and the first column comprise header information, determining the reading direction of the table based on the header information. ### Claim 9 (depends on 1) The method of claim 1, wherein determining if the set of the one or more inspection areas comprises reference data values comprises: determining whether the identified tokens can be classified into a reference data class or a non-reference data class; and in response to determining that the identified tokens cannot be automatically classified into the reference data class or the non-reference data class, prompting a user to provide a classification of the identified tokens. ### Claim 10 (depends on 1) The method of claim 1, further comprising: receiving a training set indicative of sets of features, wherein each set of features is: descriptive of the presentation layout, and labeled to indicate the presentation layout and the reference data extraction method specific to the presentation layout. ### Claim 11 (depends on 10) The method of claim 10, wherein each set of features comprises at least one of: data indicative of the presentation layout, a number of attribute values, a background class of the presentation layout, a legend type of a legend of the presentation layout, and a font type and font size of characters presented in the presentation layout. ### Claim 12 A system having one or more computer processors, the system configured to: input a block of attribute values to a predefined machine learning model; receive, from the machine learning model, an indication of a presentation layout of the block of the attribute values and an associated reference data extraction method; determine a reading direction of the block of values; identify one or more inspection areas, from the block of values, in the reading direction of the block of values; determine a set of the one or more inspection areas that shares a common presentation feature; identify tokens in the set of the one or more inspection areas; determine if the set of the one or more inspection areas comprises reference data values, wherein the determining uses a frequency distribution of the identified tokens; and output the reference data values. ### Claim 13 (depends on 12) The system of claim 12, wherein the identified tokens of the set of the one or more inspection areas are a same type. ### Claim 14 (depends on 12) The system of claim 12, wherein the presentation layout is a table, further configured to: test if each inspection area in a first row and a first column of a table comprises predefined header keywords. ### Claim 15 (depends on 14) The system of claim 14, further configured to: determine that the first row comprises at least one predefined header keyword; and mark the first row as a header row. ### Claim 16 (depends on 14) The system of claim 14, further configured to: determine that the first column comprises at least one predefined header keyword; and mark the first column as a header column. ### Claim 17 A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a server to cause the server to perform a method, the method comprising: inputting a block of attribute values to a predefined machine learning model; receiving, from the machine learning model, an indication of a presentation layout of the block of the attribute values and an associated reference data extraction method; determining a reading direction of the block of values; identifying one or more inspection areas, from the block of values, in the reading direction of the block of values; determining a set of the one or more inspection areas that shares a common presentation feature; identifying tokens in the set of the one or more inspection areas; determining if the set of the one or more inspection areas comprises reference data values, wherein the determining uses a frequency distribution of the identified tokens; and outputting the reference data values. ### Claim 18 (depends on 17) The computer program product of claim 17, wherein the presentation layout is a table, and wherein determining the reading direction of the block of values comprises: processing a first row and a first column of the table, the processing comprising: determining if the first row and the first column comprise header information; and in response to determining that at least one of the first row and the first column comprise header information, determining the reading direction of the table based on the header information. ### Claim 19 (depends on 17) The computer program product of claim 17, further comprising: receiving a training set indicative of sets of features, wherein each set of features is: descriptive of the presentation layout, and labeled to indicate the presentation layout and the reference data extraction method specific to the presentation layout. ### Claim 20 (depends on 19) The computer program product of claim 19, wherein each set of features comprises at least one of: data indicative of the presentation layout, a number of attribute values, a background class of the presentation layout, a legend type of a legend of the presentation layout, and a font type and font size of characters presented in the presentation layout.