IP Verse

Methods For Automated Filling Of Columns In Spreadsheets

Patent No. US11132492 (titled "Methods For Automated Filling Of Columns In Spreadsheets") was filed by Certara Usa Inc on Oct 6, 2020.

What is this patent about?

’492 is related to the field of electronic document processing, specifically addressing the challenge of efficiently populating spreadsheets. Traditional methods involve manual data entry or using macros with predefined formulas. These approaches are either time-consuming or limited in their ability to handle complex, unstructured data. The patent aims to improve spreadsheet functionality by automating data entry using machine learning.

The underlying idea behind ’492 is to leverage natural language processing (NLP) to automatically answer a series of related questions and populate a spreadsheet column with the answers. Instead of relying on formulas that reference other cells, the invention uses a template question containing variables that are instantiated with values from other columns in the spreadsheet. The NLP system then answers each instantiated question, and the answers are used to fill the corresponding cells in the target column.

The claims of ’492 focus on a method and system for automatically populating entries in a table. This involves receiving a template question with variables, where each variable corresponds to a 'variable column' in the table. The system then automatically identifies alphanumeric responses to instantiations of the template question using a machine learning module . Finally, the system populates a 'smart column' with these responses and automatically updates the smart column when the variable columns are edited.

In practice, a user defines a template question like "Who is the CEO of [Company Name]?" where "Company Name" is a variable linked to a column containing a list of company names. The system then generates a series of questions, such as "Who is the CEO of Acme Inc.?", "Who is the CEO of Widgets R Us?", and so on. A question answering (QA) system , potentially using models like BERT, processes each question and retrieves the answer from a dataset or the internet. These answers are then automatically inserted into the corresponding cells of the 'smart column'.

This approach differs significantly from traditional spreadsheet macros. Macros rely on predefined formulas and relationships between cells, whereas ’492 uses NLP to derive answers from unstructured data sources. This allows for more flexible and intelligent data population, as the system can handle complex questions and adapt to changes in the data. Furthermore, the system can automatically update the 'smart column' when the data in the 'variable columns' is modified, ensuring that the information remains current and accurate. This dynamic updating is a key advantage over static macros.

How does this patent fit in bigger picture?

Technical landscape at the time

In the late 2010s when ’492 was filed, question answering systems were increasingly leveraging deep learning models. At a time when NLP tasks were commonly addressed using large pre-trained models and fine-tuning techniques, question answering systems commonly relied on large datasets for training and evaluation. When hardware or software constraints made the deployment of large models non-trivial, techniques such as model compression and efficient inference were actively being explored.

Novelty and Inventive Step

The claims were rejected in a non-final office action. The rejections were based on 35 U.S.C. 112(b) for indefiniteness and 35 U.S.C. 103 for obviousness over prior art. The prosecution record does not describe the technical reasoning or specific claim changes that led to allowance.

Claims

There are 11 claims in total. Claims 1 and 7 are independent. The independent claims are directed to a method and a system for automatically populating entries in a table using a template question and a machine learning module. The dependent claims generally add specificity to the method or system, such as detailing the machine learning module or associated datasets.

Key Claim Terms New

Definitions of key terms used in the patent claims.

Term (Source)Support for SpecificationInterpretation
Machine learning module
(Claim 1, Claim 7)		“Presented herein are systems, methods, and architectures related to populating electronic documents, and, in particular, automatically filling columns in a spreadsheet, using a machine learning module. In certain embodiments, the machine learning module comprises natural language processing (NLP) software, for example, Question Answer (QA) software.”A module that automatically identifies alphanumeric responses to instantiations of a template question. It can include natural language processing (NLP) software.
Smart column
(Claim 1, Claim 7)		“Presented herein is a user interface referenced below as “smart columns,” which appear in spreadsheets referenced herein as “smart tables.” Instead of defining output according to a formula based on input variables gathered from other cells of the spreadsheet, as with macro-enabled spreadsheet software, smart tables define a series of unique (albeit patterned) questions for which answers are found (e.g., using NLP software) to fill corresponding cells of smart columns.”A column in a table that is automatically populated with answers generated by a machine learning module (e.g., NLP software) in response to instantiations of a template question. Each cell in the smart column contains an answer corresponding to a specific instantiation of the template question.
Template question
(Claim 1, Claim 7)		“As illustrated in Table 1 below, a template question is used to define the smart column where a variable defined by square brackets (e.g., [X]) is used to populate each cell in the smart column with a new question formed from the template question and the variable term from the alternative column. For example, “Who is the CEO of [Company Names]?” will fill each cell with the values from the Company Names column. Answers for each of the newly formed questions are automatically retrieved via the NLP software based on the input question.”A question containing one or more variables, used to generate a series of questions for populating a smart column. Each instantiation of the template question is created by replacing the variables with values from a corresponding variable column.
Variable column
(Claim 1, Claim 7)		“As illustrated in Table 1 below, a template question is used to define the smart column where a variable defined by square brackets (e.g., [X]) is used to populate each cell in the smart column with a new question formed from the template question and the variable term from the alternative column. For example, “Who is the CEO of [Company Names]?” will fill each cell with the values from the Company Names column.”A column in a table that contains a plurality of instantiations of a variable, where each instantiation is listed in a separate row. The values in the variable column are used to populate the template question and generate specific questions for each row of the smart column.

Patent Family

Patent Family

File Wrapper

The dossier documents provide a comprehensive record of the patent's prosecution history - including filings, correspondence, and decisions made by patent offices - and are crucial for understanding the patent's legal journey and any challenges it may have faced during examination.

  • Date

    Description

  • Get instant alerts for new documents

US11132492

CERTARA USA INC
Application Number
US17064230
Filing Date
Oct 6, 2020
Status
Granted
Expiry Date
Oct 6, 2040
External Links
Slate, USPTO, Google Patents