logo

For the Inventor, By the Inventor

Yokohama, Japan

Masafumi Nishimura

This inventor holds 62 USPTO granted patents and 3 published patent applications, plus 1 CIPO patent, primarily in Speech Recognition (CPC class G10L). Top assignees: International Business Machines Corporation, Nuance Communications, Inc., Sinoeast Concept Limited. Active years: 1989-2019.

IDiyas Innovation Intelligence. (2026). Inventor Profile: Masafumi Nishimura. Retrieved from https://idiyas.com/inventor/masafumi-nishimura

Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile

USPTO Granted Patents = 62 

% Patents Active = 61.3


Average Co-Inventor Count = 2.6

ph-index = 13

Forward Citations = 1,707(Granted Patents)

Forward Citations (Not Self Cited) = 1,693(Dec 10, 2025)


Inventors with similar research interests:


Location History:

  • Tokyo, JP (2015)
  • Kanagawa, JP (1991 - 2017)
  • Kanagawa, JE (2017)
  • Yokohama, JP (1989 - 2018)
  • Kanagawa-ken, JP (2009 - 2019)

Company Filing History:

goldMedal40 out of 166,351 
 
International Business Machines Corporation
 patents
silverMedal20 out of 1,422 
 
Nuance Communications, Inc.
 patents
bronzeMedal2 out of 12 
 
Sinoeast Concept Limited
 patents

Years Active: 1989-2019

Loading Chart...
Loading Chart...
Areas of Expertise:
Language Model Training
Pronunciation Accuracy
Speech Synthesis
N-gram Language Model
Speech Retrieval
Voice Activity Detection
Phoneme Annotation
Noise Reduction
Natural Language Processing
Speaker-Adaptive Voice
Feature Extraction
Text-to-Speech
62 patents (USPTO):Explore Patents
1. 10418029 - Method of selecting training text for language model, and method of training language model using the training text, and computer and computer program for executing the methods (2019-09-17),
2. 9978364 - Pronunciation accuracy in speech recognition (2018-05-22),
3. 9934776 - Method of selecting training text for language model, and method of training language model using the training text, and computer and computer program for executing the methods (2018-04-03),
4. 9892727 - Method of selecting training text for language model, and method of training language model using the training text, and computer and computer program for executing the methods (2018-02-13),
5. 9747893 - Unsupervised training method, training apparatus, and training program for an N-gram language model based upon recognition reliability (2017-08-29),
6. 9626958 - Speech retrieval method, speech retrieval apparatus, and program for speech retrieval apparatus (2017-04-18),
7. 9626957 - Speech retrieval method, speech retrieval apparatus, and program for speech retrieval apparatus (2017-04-18),
8. 9601110 - Unsupervised training method for an N-gram language model based upon recognition reliability (2017-03-21),
9. 9536518 - Unsupervised training method, training apparatus, and training program for an N-gram language model based upon recognition reliability (2017-01-03),
10. 9384730 - Pronunciation accuracy in speech recognition (2016-07-05),
11. 9378736 - Speech retrieval method, speech retrieval apparatus, and program for speech retrieval apparatus (2016-06-28),
12. 9373328 - Speech retrieval method, speech retrieval apparatus, and program for speech retrieval apparatus (2016-06-21),
13. 9311291 - Correcting N-gram probabilities by page view information (2016-04-12),
14. 9275631 - Speech synthesis system, speech synthesis program product, and speech synthesis method (2016-03-01),
15. 9269357 - System and method for extracting a specific situation from a conversation (2016-02-23),

Title: **Masafumi Nishimura: A Pioneer in Language Model Innovations**

Introduction

Masafumi Nishimura, based in Yokohama, Japan, is a notable inventor known for his contributions in the field of language models. With an impressive portfolio of 62 patents, Nishimura has made significant advancements in technology that enhance natural language processing and speech recognition systems.

Latest Patents

Among his latest inventions, Nishimura has developed a method for selecting training text for language models and a corresponding training method. These innovations are described in detail in his patents, which introduce a template generation technique for selecting training text from a corpus. The method involves the replacement of certain words with special symbols and selecting relevant text from a different domain to improve the training process. Furthermore, his work on a reading accuracy-improving system analyses speech recognition results to enhance the accuracy of reading conversion, employing a scoring system to select the best candidate word strings.

Career Highlights

Masafumi Nishimura has garnered extensive experience through his roles at renowned companies, including IBM and Nuance Communications, Inc. His work at these organizations has allowed him to delve deep into research and development, leading to the innovation of several patents that are pivotal in the field of AI and language processing.

Collaborations

Nishimura has collaborated with noteworthy colleagues, including Gakuto Kurata and Nobuyasu Itoh. These partnerships have fostered an excellent environment for knowledge sharing and technological advancement, resulting in groundbreaking tools and methods that contribute to the evolution of language technologies.

Conclusion

Masafumi Nishimura stands out as a visionary in the realm of language models and speech recognition. His extensive patent portfolio and collaborations with leading professionals illustrate his commitment to innovation and excellence in technology. As he continues to develop new methodologies, his contributions are poised to have a lasting impact on the industry and enhance how machines understand and process human language.

Profile summary based on public USPTO records.
Data Sources: USPTO Patent Grant XML, Patent Center, EPO & CIPO • Normalized by IDiyas Innovation Graph. Methodology & provenance architecturePlease report any incorrect information to [email protected]
Loading…