IP Library › Granted Patent US 12,436,981
Granted Patent B2
US 12,436,981 · App. 18/065,621 · Granted Oct 7, 2025

Transmforming table-to-text using agglomerative clustering

Inventor: Kunal Sawarkar (Franklin Park, NJ)
Assignee: International Business Machines Corporation
G06F16/3344G06F40/10G06N20/00
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,436,981
App. No.
18/065,621
Granted
Oct 7, 2025
Kind
B2
Abstract

A method and system of generating inferential text are provided. The method includes ingesting a data set that includes at least one structured hierarchical or multidimensional table for a particular domain. The method includes processing the ingested data set that includes the at least one structured hierarchical or multidimensional table for the particular domain by applying a generated machine learning model. The method includes generating inferential natural language text based on applying the machine learning model. The method includes outputting the generated inferential natural language text in a sequence format.

Claims (54)

1. A method, comprising:

ingesting, by a computing system, a data set that includes at least one structured hierarchical or multidimensional table for a particular domain;

processing, by the computing system, the ingested data set that includes the at least one structured hierarchical or multidimensional table for the particular domain by applying a generated machine learning model;

generating, by the computing system, inferential natural language text based on applying the machine learning model;

assigning, by the computing system, a row identifier and a unique token for each row of the at least one structured hierarchical or multidimensional table;

converting, by the computing system, each row of the at least one structured hierarchical or multidimensional table into a sentence with the assigned row identifier and unique token;

mapping, by the computing system, using agglomerative clustering techniques, the at least one structured hierarchical or multidimensional table and an associated hierarchical structure;

determining, by the computing system, based on mapping the at least one structured hierarchical or multidimensional table and the associated hierarchical structure, context for each row of the at least one structured hierarchical or multidimensional table;

determining, based on the determined context for each row of the at least one table, a relationship between each row of the at least one structured hierarchical or multidimensional table; and

outputting, by the computing system, the generated inferential natural language text in a sequence format.

2. The method of claim 1 , wherein the at least one structured hierarchical or multidimensional table is a multidimensional clustering table.

3. The method of claim 1 , wherein determining the relationship between each row further comprises using, by the computing system, a recurrent neural network based on semi-supervised model or a sequence model.

4. The method of claim 1 , wherein determining the relationship between each row further comprises using, by the computing system, a deep neural network natural language generation model as a pyramid function.

5. The method of claim 1 , the method further comprising inputting, by the computing system, the generated inferential text to a feedback loop configured to compare the generated inferential text to historically generated inferential text.

6. The method of claim 1 , wherein:

determining a relationship between each row further comprises inputting each unique token to a neural network;

the neural network comprises an input layer, at least one hidden layer, and an output layer; and

the output layer is a SoftMax function.

7. A computing system comprising:

a processor;

a storage device coupled to the processor;

program instructions stored in the storage device, wherein an execution of the instructions by the processor configures the processor to perform acts comprising:

ingesting, by the processor, a data set that includes at least one structured hierarchical or multidimensional table for a particular domain;

processing the ingested data set that includes the at least one structured hierarchical or multidimensional table for the particular domain by applying a generated machine learning model;

generating inferential natural language text based on applying the machine learning model;

assigning a row identifier and unique token for each row of the at least one structured hierarchical or multidimensional table;

converting for each row of the at least one structured hierarchical or multidimensional table into a sentence with the assigned row identifier and unique token;

mapping, using agglomerative clustering techniques, the at least one hierarchical or multidimensional structured table and an associated hierarchical structure;

determining based on mapping the at least one structured hierarchical or multidimensional table and the associated hierarchical structure, context for each row of the at least one structured hierarchical or multidimensional table; and

determining, based on the determined context for each row of the at least one structured table, a relationship between each row of the at least one structured table; and

outputting the generated inferential natural language text in a sequence format.

8. The computer system of claim 7 , wherein the determining the relationship between each row further comprises using a recurrent neural network based on semi-supervised model or a sequence model.

9. The computer system of claim 7 , wherein the determining the relationship between each row further comprises using, by the computing system, a deep neural network natural language generation model as a pyramid function.

10. The computer system of claim 7 , wherein the execution of the instructions by the processor further configures the processor to perform an act comprising inputting the generated inferential text to a feedback loop configured to compare the generated inferential text to historically generated inferential text.

11. The computer system of claim 7 , wherein

determining a relationship between each row further comprises inputting each unique token to a neural network;

the neural network comprises an input layer, at least one hidden layer, and an output layer; and

the output layer is a SoftMax function.

12. The computer system of claim 7 , wherein the at least one structured hierarchical or multidimensional table is a multidimensional clustering table.

13. A computer program product comprising:

one or more non-transitory computer-readable storage devices and program instructions stored on at least one of the one or more non-transitory storage devices, the program instructions executable by a processor, the program instructions comprising instructions to:

ingest, by the processor, a data set that includes at least one structured hierarchical or multidimensional table for a particular domain;

process the ingested data set that includes the at least one structured hierarchical or multidimensional table for the particular domain by applying a generated machine learning model;

generate an inferential natural language text based on applying the machine learning model;

assigning a row identifier and unique token for each row of the at least one structured hierarchical or multidimensional table;

converting each row of the at least one structured hierarchical or multidimensional table into a sentence with the assigned row identifier and unique token;

mapping, using at least one agglomerative clustering technique, the at least one structured hierarchical or multidimensional table and an associated hierarchical structure;

determining based on a mapping of the at least one structured hierarchical or multidimensional table and the associated hierarchical structure, context for each row of the at least one structured hierarchical or multidimensional table;

determining, based on the determined context for each row of the at least one structured table, a relationship between each row of the at least one structured table; and

output the generated inferential natural language text in a sequence format.

14. The computer program product of claim 13 , wherein the determining the relationship between each row further comprises using a recurrent neural network based on semi-supervised model or a sequence model.

15. The computer program product of claim 13 , wherein determining the relationship between each row further comprises using, by the computing system, a deep neural network natural language generation model as a pyramid function.

16. The computer program product of claim 13 , the program instructions comprising instructions to input, by the computing system, the generated inferential text to a feedback loop configured to compare the generated inferential text to historically generated inferential text.

17. The computer program product of claim 13 , wherein the at least one structured hierarchical or multidimensional table is a multidimensional clustering table.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 13, 2022
From: SAWARKAR, KUNAL
To: INTERNATIONAL BUSINESS MACHINES CORPORATION
Reel/Frame 062077/0693 →
Continuity (1)
Related Publication 20240193191A1 · Jun 13, 2024
References Cited (42)
US 9336772B1 · Salvador · 2016 [cited by examiner]
US 9817919B2 · Aila et al. · 2017 [cited by applicant]
US 10162878B2 · Nuchia et al. · 2018 [cited by applicant]
US 10552129B2 · Salgado et al. · 2020 [cited by applicant]
US 11010545B2 · Guggilla et al. · 2021 [cited by applicant]
US 11182806B1 · Arfa et al. · 2021 [cited by applicant]
US 11301639B2 · Jalali et al. · 2022 [cited by applicant]
US 11386145B2 · Singh · 2022 [cited by applicant]
US 11886827B1 · Margolin · 2024 [cited by examiner]
US 20130185050A1 · Bird · 2013 [cited by examiner]
US 20170235820A1 · Conrad et al. · 2017 [cited by applicant]
US 20200250212A1 · Macartney et al. · 2020 [cited by applicant]
US 20200273453A1 · Mody et al. · 2020 [cited by applicant]
US 20210124724A1 · Bordawekar · 2021 [cited by examiner]
US 20210311937A1 · Bordawekar · 2021 [cited by examiner]
US 20210342447A1 · Sanzgiri et al. · 2021 [cited by applicant]
US 20210349929A1 · Seledkin et al. · 2021 [cited by applicant]
US 20220044134A1 · Joy · 2022 [cited by examiner]
US 20220179896A1 · Farrell et al. · 2022 [cited by applicant]
Rebuffel, Clement, et al. “A hierarchical model for data-to-text generation.” Advances in Information Retrieval: 42nd European Conference on IR Research, ECIR 2020, Lisbon, Portugal, Apr. 14-17, 2020, Proceedings, Part … [cited by examiner]
Bao, Junwei, et al. “Table-to-text: Describing table region with natural language.” Proceedings of the AAAI conference on artificial intelligence. vol. 32. No. 1. 2018. (Year: 2018). [cited by examiner]
Liu, Tianyu, et al. “Hierarchical encoder with auxiliary supervision for neural table-to-text generation: Learning better representation for tables.” Proceedings of the AAAI Conference on Artificial Intelligence. vol. 3… [cited by examiner]
Gong, Heng, et al. “Table-to-text generation with effective hierarchical encoder on three dimensions (row, column and time).” arXiv preprint arXiv:1909.02304 (2019). (Year: 2019). [cited by examiner]
Gong, Heng, et al. “Table-to-text generation via row-aware hierarchical encoder.” Chinese Computational Linguistics: 18th China National Conference, CCL 2019, Kunming, China, Oct. 18-20, 2019, Proceedings 18. Springer I… [cited by examiner]
Moosavi, Nafise Sadat, et al. “Scigen: a dataset for reasoning-aware text generation from scientific tables.” Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2). 202… [cited by examiner]
Yang, Yang, et al. “Table to text generation with accurate content copying.” Scientific reports 11.1 (2021): 22750. (Year: 2021). [cited by examiner]
Gao, L. L. et al., “Selective Inference for Hierarchical Clustering”, arXiv:2012.02936v2 [stat.ME] (2021), 55 pgs. [cited by applicant]
Sun, K. et al., “Tabular Functional Block Detection with Embedding-based Agglomerative Cell Clustering”; CIKM (2021); 10 pgs. [cited by applicant]
Bodrunova, S.S. et al., “Topic Detection Based on Sentence Embeddings and Agglomerative Clustering with Markov Moment”; Future Internet (2020); vol. 12:144, 17 pgs. [cited by applicant]
Ah-Pine, J. “An Efficient and Effective Generic Agglomerative Hierarchical Clustering Approach”; Journal of Machine Learning Research (2018), 43 pgs. [cited by applicant]
Bailoni, A., et al., “GASP, A Generalized Framework for Agglomerative Clustering of Signed Graphs and its Application to Instance Segmentation”; arXiv:1906.11713v2 [cs.CV] (2022); 20pgs. [cited by applicant]
Chen, S. et al., “A General Model for Neural Text Generation from Structured Data”; School of Computer Science and Technology, Harbin Institute of Technology, Harbin, China (Date unknown). [cited by applicant]
Bao, J. et al., “Text Generation From Tables”; IEEE/ACM Transactions on Audio, Speech, and Language Processing (2019); Vo. 27:2, p. 311. [cited by applicant]
Alsuhaibani, M. et al., “Joint Learning of Hierarchical Word Embeddings from a Corpus and a Taxonomy”; Automated Knowledge Base Construction (2019), 18 pgs. [cited by applicant]
Al, Q. et al., “Learning a Hierarchical Embedding Model for Personalized Product Search”; SIGIR (2017); 10 pgs. [cited by applicant]
Lebret, R., et al., “Neural Text Generation from Structured Data with Application to the Biography Domain”; Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing (2016); pp. 1203-1213. [cited by applicant]
Liu, T. et al., “Table-to-text Generation by Structure-aware Seq2seq Learning”, arXiv:1711.09724v1 [cs.CL] (2017); 8 pgs. [cited by applicant]
Govindaraju, V. et al., “Understanding Tables in Context Using Standard NLP Toolkits”; Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (ACL—2013); 7 pgs. [cited by applicant]
Krishnamurthy, J. et al., “Neural Semantic Parsing with Type Constraints for Semi-Structured Tables”; Allen Institute for Artificial Intelligence, 11 pgs. (Date unknown). [cited by applicant]
Bordawekar et al, “Exploiting Latent Information in Relational Databases via Word Embedding and Application to Degrees of Disclosure”, CIDR, 2019, 6 pages. [cited by applicant]
Deng et a., “Table2Vec: NeuralWord and Entity Embeddings for Table Population and Retrieval”, arXiv, May 31, 2019, 4 pages. [cited by applicant]
Rong, “word2vec Parameter Learning Explained”, arXiv, Jun. 5, 2016, 21 pages. [cited by applicant]