# Andreas Stolcke

> Ph.D. University of California, Berkeley 1994

**Wikidata**: [Q29051346](https://www.wikidata.org/wiki/Q29051346)  
**Source**: https://4ort.xyz/entity/andreas-stolcke

## Summary

Andreas Stolcke is a computer scientist . He received his education at the University of California, Berkeley [1].Stolcke has been employed by Microsoft and Amazon .

## Summary
Andreas Stolcke is a German-American computer scientist and IEEE Fellow best known for his pioneering work in automatic speech recognition and language modeling. After earning his Ph.D. from UC Berkeley in 1994, he spent over two decades at SRI International and Microsoft, where he co-created the SRILM toolkit that became the standard for statistical language modeling research worldwide.

## Biography
- Born: 1953
- Nationality: German-American
- Education: Ph.D. Computer Science, University of California, Berkeley (1994)
- Doctoral advisor: Jerome A. Feldman
- Known for: Statistical language modeling and speech recognition
- Employer(s): Microsoft, Amazon, SRI International
- Field(s): Computer science, speech technology, machine learning

## Contributions
Andreas Stolcke's most significant contribution is the creation of the SRILM (SRI Language Modeling) toolkit in the mid-1990s, which became the de facto standard for building and evaluating statistical language models. This open-source software enabled researchers worldwide to implement n-gram models, smoothing techniques, and perplexity measurements that were fundamental to advancing speech recognition accuracy. His 1998 paper "Entropy-based Pruning of Backoff Language Models" introduced novel algorithms for compressing language models without significant accuracy loss, solving a critical memory constraint problem in early speech recognition systems.

At SRI International from 1994-2012, Stolcke led teams that developed key components of the DARPA-sponsored speech recognition evaluations, consistently achieving top-tier performance. His work on speaker diarization—the automatic answering of "who spoke when"—produced the influential LIUM speaker diarization system that became standard in broadcast news transcription. After joining Microsoft in 2012, he contributed to the speech recognition capabilities in Cortana, Skype Translator, and Microsoft Cognitive Services, helping transition research advances into commercial products used by millions.

## FAQs
### Q: What is Andreas Stolcke's most famous technical contribution?
A: Stolcke is best known for creating the SRILM toolkit, which became the standard open-source software for statistical language modeling from the mid-1990s through the 2010s. It enabled researchers to build n-gram language models with advanced smoothing techniques.

### Q: Where did Andreas Stolcke work before Microsoft?
A: Stolcke spent 18 years at SRI International's Speech Technology and Research (STAR) Laboratory from 1994 to 2012, where he led speech recognition and language modeling research projects.

### Q: What recognition has Andreas Stolcke received for his work?
A: He was named an IEEE Fellow, one of the highest honors in electrical engineering and computer science, recognizing his contributions to speech recognition and language modeling.

## Why They Matter
Andreas Stolcke's work fundamentally shaped how computers understand human speech during the critical period when speech recognition transitioned from laboratory curiosity to practical technology. His SRILM toolkit provided the foundation that enabled thousands of researchers to build better language models, directly contributing to the speech-to-text systems that became ubiquitous in smartphones, call centers, and accessibility tools. The algorithms he developed for model compression and smoothing are still used in modern deep learning-based systems.

His speaker diarization research solved the "who spoke when" problem that made meeting transcription and broadcast news indexing practical. Without his contributions, the automatic transcription systems we rely on today—from Zoom captions to voice assistants—would be significantly less accurate and require far more computational resources. His role in bridging academic research and commercial deployment at both SRI and Microsoft helped ensure that speech recognition advances moved from conference papers to products used by millions.

## Notable For
- IEEE Fellow recognition for contributions to speech recognition and language modeling
- Creator of SRILM toolkit, the standard for statistical language modeling for over 20 years
- 18-year tenure at SRI International leading speech technology research
- Key contributor to Microsoft Cortana and Skype Translator speech recognition
- Co-author of influential papers on entropy-based language model pruning and speaker diarization

## Body
### Early Career and Education
Andreas Stolcke began his academic journey in Germany before pursuing graduate studies at the University of California, Berkeley. Under the guidance of computer scientist Jerome A. Feldman, he completed his Ph.D. dissertation in 1994, focusing on probabilistic parsing and language modeling techniques that would later prove foundational to his career.

### SRI International Years (1994-2012)
Stolcke joined SRI International's Speech Technology and Research Laboratory immediately after completing his doctorate. During his 18-year tenure, he became one of the most cited researchers in speech recognition, with his work on the SRILM toolkit enabling thousands of researchers to implement state-of-the-art language models. His team's participation in DARPA evaluations consistently ranked among the top performers, particularly in broadcast news transcription and conversational telephone speech recognition tasks.

### Transition to Industry
In 2012, Stolcke moved to Microsoft, where he applied his expertise to commercial speech recognition products. He contributed to the speech recognition backend for Cortana, Microsoft's virtual assistant, and helped develop the real-time translation capabilities in Skype Translator. His work focused on scaling speech recognition to handle millions of users while maintaining the accuracy levels required for consumer applications.

### Recent Work at Amazon
Following his time at Microsoft, Stolcke joined Amazon, continuing his work in speech and language technology. His recent contributions focus on improving the accuracy and efficiency of speech recognition systems at cloud scale, though specific details of his current projects remain proprietary.

## References

1. Mathematics Genealogy Project