Skip to content

Prior Work

Jump to bottom
Emily Ahn edited this page Mar 30, 2016 · 13 revisions

Accent Detection: Untranscribed

  1. Chouieter, Zweig, and Nguyen, 2008. ICASSP. http://research.microsoft.com/pubs/77909/icassp08c.pdf. GMMs on average PLPs.

  2. Reynolds and Rose, 1995. TASLP. http://www.ece.mcgill.ca/~rrose1/papers/reynolds_rose_sap95.pdf. Speaker recognition, but same general idea. Nice explanations.

  3. Machanavajhala, 2013. MS Thesis from Utah. http://content.lib.utah.edu/utils/getfile/collection/etd3/id/2617/filename/2613.pdf.

  4. Chen et al, 2010. ICASSP. https://www.ll.mit.edu/mission/cybersec/publications/publication-files/full_papers/2010_03_19_Chen_ICASSP10_FP.pdf. Bootstrap transcriptions from a phone recognizer trained on a standard accent.

  5. Lei and Hansen, 2010. TASLP. http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5428854&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5428854. Find the most discriminative GMM mixtures.

  6. Omar and Pelecanos, 2010. ICASSP. http://www.mirlab.org/conference_papers/International_Conference/ICASSP%202010/pdfs/0004398.pdf

  7. Lopez-Moreno et al, 2014. ICASSP. http://www.mirlab.org/conference_papers/International_Conference/ICASSP%202014/papers/p5374-lopez_moreno.pdf

  8. Tang and Ghorbani. 2013. http://www.cs.unb.ca/~ghorbani/ali/papers/abs-sabrina.pdf. Finds that [DAG] SVM method performs as well as HMM for 3-way classifier (American, Turkish, Chinese). Extracts 4 features including F2-F3 contours.

  9. Pederson and Diederich, 2007. ISSPA. http://joachimdiederich.com/assets/isspa07.pdf. Use SVMs for 2-way classification of Arabic & Indian accented English. Best results (up to 97% accuracy) occur when holding speech subject content constant. Comparable to parallel study of human perception of these speech samples.

  10. Stanford CS229 (ML) Project. 2010. http://cs229.stanford.edu/proj2010/WatanaprakornkulEksombatchaiChien-AccentClassification.pdf. Used CSLU corpus and trained SVM & GMM 3-way classifiers between Cantonese, Hindi and Russian. Informal writing. VERY SIMILAR to what I am doing!

Accent Detection: Transcribed

Data-Driven

  1. Arslan and Hansen, 1996. TASLP. http://www.sciencedirect.com/science/article/pii/0167639396000246. HMMs for isolated word and phones for each accent.

  2. Winebarger et al, 2014. AAAI. http://www.aaai.org/ocs/index.php/AIIDE/AIIDE14/paper/viewFile/9080/9037. French-German, German-French language-learning game using ASR.

Linguistically Inspired

  1. Chen et al, 2013. TASLP. https://www.researchgate.net/profile/Nancy_Chen6/publication/260661416_Characterizing_Phonetic_Transformations_and_Acoustic_Differences_Across_English_Dialects/links/5564320d08ae8c0cab37033c.pdf

  2. Sangwan and Hansen. 2012. Speech Communication. http://www.sciencedirect.com/science/article/pii/S0167639311000872. Analyzes Mandarin accented English. Potentially useful if we want to go into analyzing phonological features.

  3. Bartkova and Jouvet. 1999. https://www.internationalphoneticassociation.org/icphs-proceedings/ICPhS1999/papers/p14_1725.pdf. French-accented English.

Crowdsourcing Transcriptions

  1. Parent and Eskenazi. 2010. Interspeech. http://www.cs.cmu.edu/~gparent/papers/amt-transcript.pdf. Good overview of crowdsourcing for transcriptions on Amazon MTurk. Discusses 2-step approach to obtaining more efficient, higher quality results.

Human Perception of Foreign Accents

  1. McCullough. 2013. https://linguistics.osu.edu/files/wpl-vol60-4-McCullough_0.pdf. The following features are indicative of accent detection: VOT, vowel quality, f0, and vowel duration. Accents evaluated: American English, Hindi, Korean, Mandarin.

  2. D'Antilio and Tamati. 2011. http://www.iub.edu/~psyling/papers/Tamati_FASR_L2L.pdf. (Slides format) Conducted 2 experiments on native and non-native English speakers' perceptions of similarity of foreign accents. Used CSLU FAE corpus!

testing folder

testing sidebar

Clone this wiki locally