Clinical Background

Anomia, or word-finding difficulty, is one of the cardinal features of aphasia 1 2, an acquired neurogenic language disorder that affects 2.5-4 million people in the US alone 3. The primary cause of aphasia is stroke, and 21%–40% of acute stroke patients are diagnosed with anomia by the time they are discharged. Further, advanced age is a major risk factor for aphasia, and given the current aging trend, the incidence of aphasia is expected to increase in the coming decades 4 5 6.

Anomia is believed to be indicative of disruption in accessing a semantic description of the target concept, and/or retrieving a fully phonologically specified representation 7 8. Specifically, paraphasias, which are unintended word production errors, are generally considered to result from reduced or insufficiently persistent activation of target representations relative to competing non-target representations and/or noise in the system 7 9 10. Reduced activation of lexical-semantic representations may result in semantic errors (e.g. “dog” for the target “cat”). Form-related words may also become activated, due to spreading activation and feedback (e.g., “mat” for “cat”). Also, activation of inappropriate phoneme representations may result in phonological errors such as non-word productions known as neologisms (e.g., “tat” for the target “cat”) or real word phonemic errors (e.g., “dog” for the target “log”) 11. Additionally, multiple breakdowns in word retrieval may result in mixed errors that share both a semantic and a phonological relationship with the target (“rat” for “cat”). Finally, unrelated errors, sharing no obvious semantic or phonological features with the target word, can be part of the anomic symptomatology (“chair” for “cat”).

In theoretical and clinical research investigations, professionals typically use confrontation naming tests, during which a person with aphasia is presented with pictures of simple objects and they are asked to name them. In research settings, professionals develop individualized profiles based on the different types of errors elicited through such tests (e.g., phonological, semantic, nonword errors, etc.) and then these profiles can be used to characterize patients’ cognitive-linguistic deficits. Such individualized error profiles have informed theoretical accounts of anomia (e.g., 12 13 14 15) and the cognitive machinery underlying word production 7 8 10 16 17; lesion-symptom mapping fe 18 19 20; personalization of treatments 21; treatment efficacy studies 22 23 24 25; our understanding of cross-linguistic treatment generalization 26; and cortical re-organization investigations after a stroke 27.

Error profiles also have the potential to be highly informative in clinical settings for developing individualized intervention plans 28. However, current approaches to developing a patient’s profile are prohibitively time- and labor-intensive because they require phonemic transcriptions for determining response accuracy and the nature of the errors. For a naming test with dozens or hundreds of items, this is rarely feasible in fast-paced clinical settings including Intensive Care Units and acute rehabilitation units. As such, there is much interest in the aphasiological community in automating this process. Given the notable improvements in the state of the art in ASR in recent years, we believe that existing technology is at a point where this is feasible, and we further believe that the clinical importance and technical challenges of this task would be compelling to many in the mainstream ASR community.

The activities under the Post-Stroke Speech Transcription (PSST) Challenge are a step towards developing an ASR system that can transform current clinical delivery paradigms and accelerate scientific discovery. The PSST Challenge is a collaboration between Oregon Health and Science University (OHSU) and Portland State University (PSU). Our activities are supported via a grant from the National Institute on Deafness and Other Communication Disorders NIH (R01-DC015999-04S1), the explicit purpose of which is to promote the use of clinical datasets of aphasic speech by the mainstream machine learning community towards developing efficient tools for the diagnosis and management of post-stroke language disorders.

Bibliography

  1. A. M. Raymer and L. J. G. Rothi, “Impairments of word comprehension and production,” in Language intervention strategies in aphasia and related neurogenic communication disorders, 4th ed., R. Chapey, Ed. Lippincott Williams & Wilkins, 2001, pp. 606–625. ↩

  2. H. Goodglass and A. Wingfield, Anomia: Neuroanatomical and cognitive correlates. San Diego, CA: Academic Press, 1997. ↩

  3. N. Simmons-Mackie, Aphasia in North America. Moorestown, NJ: Aphasia Access, 2018. ↩

  4. N. I. on Aging, “Health and Aging,” National Institute on Aging, Mar. 21, 2014. http://www.nia.nih.gov/health/publication/aging-hearts-and-arteries/preface (accessed Jan. 26, 2015). ↩

  5. S. T. Engelter et al., “Epidemiology of aphasia attributable to first ischemic stroke incidence, severity, fluency, etiology, and thrombolysis,” Stroke, vol. 37, no. 6, pp. 1379–1384, Jun. 2006, doi: 10.1161/01.STR.0000221815.64093.8c. ↩

  6. U. C. B. P. I. Office, “Aging population,” United States Census Bureau, 2011. https://www.census.gov/newsroom/releases/archives/aging_population/ (accessed Jan. 26, 2015). ↩

  7. G. S. Dell, “A spreading-activation theory of retrieval in sentence production,” Psychol. Rev., vol. 93, no. 3, pp. 283–321, 1986. ↩ ↩2 ↩3

  8. G. S. Dell, M. F. Schwartz, N. Martin, E. M. Saffran, and D. A. Gagnon, “Lexical access in aphasic and nonaphasic speakers,” Psychol. Rev., vol. 104, no. 4, pp. 801–838, Oct. 1997, doi: http://dx.doi.org/10.1037/0033-295X.104.4.801. ↩ ↩2

  9. G. S. Dell, F. Chang, and Z. M. Griffin, “Connectionist models of language production: Lexical access and grammatical encoding,” Cogn. Sci., vol. 23, no. 4, pp. 517–542, Oct. 1999, doi: 10.1016/S0364-0213(99)00014-2. ↩

  10. M. F. Schwartz, G. S. Dell, N. Martin, S. Gahl, and P. Sobel, “A case-series test of the interactive two-step model of lexical access: Evidence from picture naming,” J. Mem. Lang., vol. 54, no. 2, pp. 228–264, Feb. 2006, doi: 10.1016/j.jml.2005.10.001. ↩ ↩2

  11. D. Foygel and G. S. Dell, “Models of impaired lexical access in speech production,” J. Mem. Lang., vol. 43, no. 2, pp. 182–216, Aug. 2000, doi: 10.1006/jmla.2000.2716. ↩

  12. S. Freud, On aphasia: A critical study. Madison, CT: International Universities Press, 1953. ↩

  13. V. A. Fromkin, “The non-anomalous nature of anomolous utterances,” Language, vol. 47, no. 1, pp. 27–52, 1971, doi: 10.2307/412187. ↩

  14. M. F. Garrett, “Levels of processing in sentence production.,” in Language production, B. Butterworth, Ed. New York, NY: Academic Press, 1980, pp. 177–220. ↩

  15. M. F. Garrett, “The analysis of sentence production.,” in Psychology of learning and motivation, vol. 9, G. Bower, Ed. New York, NY: Academic Press, 1975, pp. 133–177. ↩

  16. N. Nozari and G. S. Dell, “How damaged brains repeat words: A computational approach,” Brain Lang., vol. 126, no. 3, pp. 327–337, 2013, doi: 10.1016/j.bandl.2013.07.005. ↩

  17. G. S. Dell and P. G. O’Seaghdha, “Stages of lexical access in language production,” Cognition, vol. 42, no. 1–3, pp. 287–314, 1992, doi: 10.1016/0010-0277(92)90046-K. ↩

  18. M. F. Schwartz et al., “Anterior temporal involvement in semantic word retrieval: Voxel-based lesion-symptom mapping evidence from aphasia,” Brain, Nov. 2009, doi: 10.1093/brain/awp284. ↩

  19. G. M. Walker et al., “Support for anterior temporal involvement in semantic error production in aphasia: New evidence from VLSM,” Brain Lang., no. 3, pp. 110–122, 2011, doi: 10.1016/j.bandl.2010.09.008. ↩

  20. M. F. Schwartz, O. Faseyitan, J. Kim, and H. B. Coslett, “The dorsal stream contribution to phonological retrieval in object naming,” Brain, vol. 135, no. 12, pp. 3799–3814, Dec. 2012, doi: 10.1093/brain/aws300. ↩

  21. W. Best, A. Greenwood, J. Grassly, R. Herbert, J. Hickin, and D. Howard, “Aphasia rehabilitation: Does generalisation from anomia therapy occur and is it predictable? A case series study,” Cortex, vol. 49, no. 9, pp. 2345–2357, Oct. 2013, doi: 10.1016/j.cortex.2013.01.005. ↩

  22. D. Kendall, T. Conway, J. Rosenbek, and L. Gonzalez‐Rothi, “Case study: Phonological rehabilitation of acquired phonologic alexia,” Aphasiology, vol. 17, no. 11, pp. 1073–1095, Jan. 2003, doi: 10.1080/02687030344000355. ↩

  23. C. E. Brookshire, T. Conway, R. H. Pompon, M. Oelke, and D. L. Kendall, “Effects of intensive phonomotor treatment on reading in eight individuals with aphasia and phonological alexia,” Am. J. Speech Lang. Pathol., vol. 23, no. 2, pp. S300–S311, May 2014, doi: 10.1044/2014_AJSLP-13-0083. ↩

  24. D. L. Kendall et al., “Phoneme-based rehabilitation of anomia in aphasia,” Brain Lang., vol. 105, no. 1, pp. 1–17, Apr. 2008, doi: 10.1016/j.bandl.2007.11.007. ↩

  25. D. L. Kendall, A. Rodriguez, J. Rosenbek, T. Conway, and L. J. Gonzalez Rothi, “The influence of intensive phonomotor rehabilitation of apraxia of speech,” J. Rehabil. Res. Dev., vol. 43, no. 3, pp. 409–418, 2006. ↩

  26. L. A. Edmonds and S. Kiran, “Effect of semantic naming treatment on crosslinguistic generalization in bilingual aphasia,” J. Speech Lang. Hear. Res., vol. 49, no. 4, p. 729, Aug. 2006, doi: 10.1044/1092-4388(2006/053). ↩

  27. J. Fridriksson, J. D. Richardson, P. Fillmore, and B. Cai, “Left hemisphere plasticity and aphasia recovery,” NeuroImage, vol. 60, no. 2, pp. 854–863, Apr. 2012, doi: 10.1016/j.neuroimage.2011.12.057. ↩

  28. S. Abel, K. Willmes, and W. Huber, “Model‐oriented naming therapy: Testing predictions of a connectionist model,” Aphasiology, vol. 21, no. 5, pp. 411–447, Apr. 2007, doi: 10.1080/02687030701192687. ↩