User:Jsalsman

James Salsman is a statistician, software engineer, speech recognition specialist, and World Wide Web Consortium Invited Expert in the Device API, Web Apps, and HTML working groups. Salsman has over 20 years of telephony, signal processing, C, Perl, Javascript, Flash MXML, R, SQL, Tcl/Tk, Java and related experience. Salsman studied computer science and mathematics at Carnegie Mellon University and has worked or corresponded with EnglishCentral.com (Google Ventures), Scientific Learning ("Reading Assistant"), Rosetta Stone, 8DWorld.com ("Wiz World Online"), Transparent.com, DynEd.com, Ordinate.com (Pearson), and Carnegie Speech, Netscape, Silicon Graphics, LeapFrog Enterprises, MobiTV, Cisco, Mindsource, and a variety of Silicon Valley startups. Salsman's contributions to open source software include substantial improvements to the phase vocoder algorithm efficiency, upgrades to TCL and Android, and work to patch and extend Mediawiki. He is currently working on speech recognition for pronunciation evaluation, helping people learn to speak and read well. Other interests include:

• Hack the Future (volunteer mentor)
• Google Summer of Code (volunteer mentor)

Recent highlights

EF Education First, EF Learning Labs, Shanghai, China, 2013–2014: Improved automatic speech recognition (ASR) systems providing pronunciation assessment for English language learning by diagnosing Adobe Flash-based microphone upload channel faults, immediately reversing a 30% accuracy drop prior to my arrival. Architected, validated, and implemented further pronunciation assessment accuracy improvement using Sensory Fluentsoft ASR with phoneme duration and acoustic scores normalized by establishing a leaderboard of exemplar pronunciations from student uploads, achieving a 24% increase in scores’ agreement with a panel of human judges. Prototyped auditory feedback for pronunciation exercises, designed ASR QA systems, and additional word and phrase score improvements on cross-platform mobile and desktop ASR implementations. Several other contributions to processes, internal technical documentation, and online learning functions. Used C, JavaScript, sh, C#, and ObjectiveC on Android, iOS, Linux servers, Windows ASP.NET servers and desktop, and OS X.

Selected publications

J. Salsman (July 2014) “Development challenges in automatic speech recognition for computer assisted pronunciation teaching and language learning” in Proceedings of the Research Challenges in Computer Aided Language Learning Conference (CALL 2014) Antwerp, Belgium: talknicer.com/Salsman-CALL-2014.pdf

S. Ronanki, J. Salsman, and L. Bo (December 2012) “Automatic Pronunciation Evaluation and Mispronunciation Detection using CMU Sphinx.” in Proceedings of the Workshop on Speech and Language Processing Tools in Education, pp. 61–68. 24th International Conference on Computational Linguistics (COLING 2012) Mumbai, India: www.aclweb.org/anthology/W12-5808

K. Roast and J. Salsman (August 2011) “K3D JavaScript Canvas Library.” Software documentation: en.wikibooks.org/wiki/K3D_JavaScript_Canvas_Library

J. Salsman (May 2010) “Asynchronous Microphone Upload – for Pronunciation Assessment, High-Quality, Low-Bandwidth Voice, Speech Transcription, Translation, and Speaker Identification and Verification.” in the Proceedings of the World Wide Web Consortium Workshop on Conversational Applications (W3C CONVAPPS) June 18–19, 2010, Somerset, New Jersey: www.w3.org/2010/02/convapps/Papers/asynchMicUpload.pdf

J. Salsman (October 2010) “Teaching computers to teach people to read and speak.” One Laptop Per Child San Francisco Bay Area Community Summit (OLPC-SF 2010) presentation. San Francisco, California: talknicer.com/olpcsf.pdf

J. Salsman (2005) “ReadSay PROnounce English System.” Self-published commercial software and instructional modules: talknicer.com/pronounce

J. Salsman (August 2004) “Getting Sorted Indices out of lsort.” Tcl Improvement Proposal (TCL TIP) #217. Tcl Developer Xchange: www.tcl.tk/cgi-bin/tct/tip/217.html

 J. P. Salsman (July 1999) “Form-based Device Input and Upload in HTML.” World Wide Web Consortium Note submission from Cisco Systems, San Jose, California:  www.w3.org/TR/device-upload

J. Salsman and H. Alvestrand (May 1999) “The Audio/L16 MIME content type.” Internet Engineering Task Force Request for Comments (IETF RFC 2586) www.ietf.org/rfc/rfc2586.txt

Interested in

• Google Summer of Code 2015
  • Systems for accuracy review, driven primarily by
    • Lua-based templates interacting with a simple housekeeping bot
    • Low stakes instructional assessment content in Moodle's GIFT (for reviewer and QA presentation, from template to bot input; bot reads review questions and copies them to subscribed reviewer users)
    • Vital article lists and their covering categories (helps select articles for review)
    • Keywords indicating facts and figures which are likely to become out of date (helps to select passages for review)
    • Database (from dumps) with age of each word on an (article) page, in days (also helps to select passages for review)
    • The DELPH-IN LOGIN parser (helps to select passages in need of copyediting for review)
    • (Optional stretch goal) Microphone audio upload (Flash and WebRTC) using PocketSphinx-based pronunciation assessment on answers (not sure if this is over-the-top or a legit stretch, since it's already open source from my 2012 GSoC work with CMU Sphinx)
  • TODO
    • write proposal in terms of bot stubs and Lua template prototypes
    • Justify with details about:
      • Community involvement
      • Deployment plan
      • Generic MediaWiki basis
      • Free software APIs
• Simple language wikipedias in languages other than English
• Most popular related articles

...among other things. Jsalsman (talk) 04:02, 10 February 2015 (UTC)