PeopleofSuperintelligence

Researcher

Michael Riley

Michael Riley is a computer scientist whose research developed weighted finite-state methods for speech recognition and language processing. He holds three computer science degrees from MIT and worked with Mehryar Mohri and Fernando Pereira on transducer methods at Bell Labs and AT&T Labs. His research includes a 1996 speech-recognition framework, OpenFst, lattice generation, and neural transducer models. As of October 2026, he is a principal research scientist at Google, with interests in speech and natural language processing, machine learning, and information retrieval.

Works
131
Citations
5,223
h-index
36
Badges

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 1996

    A speech recognition framework

    He holds B.S., M.S., and Ph.D. degrees in computer science from MIT. With Fernando Pereira, he presents a framework for speech recognition based on weighted finite automata and weighted finite-state transducers in 1996.

  2. 2001

    Weight pushing for speech

    At Bell Labs and AT&T Labs, he develops weighted finite-state transducer methods for speech and language with Mehryar Mohri and Fernando Pereira. In 2001, he and Mohri publish a weight-pushing algorithm that redistributes transducer weights to aid pruning and speech recognition.

  3. 2009

    OpenFst for speech and language

    In 2009, he coauthors an OpenFst tutorial with Cyril Allauzen and Martin Jansche. The free, open-source library provides tools for building and using weighted finite-state transducers in speech and language processing.

  4. 2017

    Lattice generation for recognition

    With David Rybach and Johan Schalkwyk, he presents an ASRU paper in 2017 on lattice generation for large-vocabulary speech recognition. It compares phone-pair and N-best histories approaches and proposes pruned determinization followed by minimization to produce deterministic minimal lattices.

  5. 2018

    Algorithms for weighted automata

    In 2018, he coauthors a paper that extends weighted finite-automata algorithms to automata with failure transitions. It presents intersection and shortest-distance algorithms and a method to remove failure transitions.

  6. 2022

    Adaptive HAT weight interpolation

    In 2022, he coauthors a study of rescoring strategies for a two-pass speech-recognition system with a hybrid autoregressive transducer and a neural language model. The paper proposes adaptive combination weights for each acoustic sequence.

  7. October 2025

    Speech-to-Retrieval at Google

    In October 2025, he is a Research Scientist at Google Research and introduces Speech-to-Retrieval (S2R) with Ehsan Variani. The voice-search system retrieves information from spoken queries directly, and Google Voice Search now uses it.

  8. As of October 2026

    Principal research scientist at Google

    He is a principal research scientist at Google, with interests in speech and natural language processing, machine learning, and information retrieval.

Career

  1. 2004–2025Google
  2. 1989–2005AT&T
  3. Google ResearchResearch Scientist

Education

  1. Massachusetts Institute of TechnologyB.S., M.S., and Ph.D. in computer science

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. Weighted finite-state transducers in speech recognition2002 · Computer Speech & Language · 887 citationsWith Mehryar Mohri
  2. OpenFst: A General and Efficient Weighted Finite-State Transducer Library2007 · Springer eBooks · 541 citationsWith Mehryar Mohri
  3. Sample Selection Bias Correction Theory2008 · Lecture notes in computer science · 266 citationsWith Corinna Cortes and Mehryar Mohri
  4. Speech Recognition with Weighted Finite-State Transducers2007 · Springer handbooks · 254 citationsWith Mehryar Mohri
  5. Speech Recognition by Composition of Weighted Finite Automata1997 · The MIT Press eBooks · 188 citationsWith Fernando C. N. Pereira

From OpenAlex, as of 21 October 2026.

  1. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Michael Riley. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  4. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Michael Riley yet. Send us a free-licensed photo.

  1. Michael Riley - Research at Google research.google.com. Checked 21 October 2026.
  2. Speech Recognition by Composition of Weighted Finite Automata arxiv.org. Checked 21 October 2026.
  3. ISCA Archive - A weight pushing algorithm for large vocabulary speech recognition isca-archive.org. Checked 21 October 2026.
  4. OpenFst: An Open-Source, Weighted Finite-State Transducer Library and its Applications to Speech and Language research.google. Checked 21 October 2026.
  5. On Lattice Generation for Large Vocabulary Speech Recognition research.google. Checked 21 October 2026.
  6. Algorithms for Weighted Finite Automata with Failure Transitions research.google. Checked 21 October 2026.
  7. On Adaptive Weight Interpolation of the Hybrid Autoregressive Transducer isca-archive.org. Checked 21 October 2026.
  8. Speech-to-Retrieval (S2R): A new approach to voice search research.google. Checked 21 October 2026.