Voglio sapere a riguardo di…

Federal Register / Vol. 86, No. 19 / Monday, February 1, 2021 / Proposed Rules delay the establishment of service quality standards? Should the Commission adopt default standards in these areas, pending completion of consensus standards?
Adding CTS/IP CTS Metrics to TRS
Minimum Standards 5. Caption Delay. The Commission proposes to adopt a minimum standard for caption delay and proposes the following definition:
Caption delay is the difference in time in seconds between when a word can be heard in the audio and when that word appears in the stream of captions on the caption users primary display.

The Commission seeks comment on how to specify more precisely what is meant by when a captioned word appears in the transcript or stream of captions. Should such appearance be defined as the initial appearance of the word i.e., prior to any correction that may be provided subsequently or its final displayed appearance i.e., so that the caption delay includes any time involved in providing a corrected version of the word? Would measuring caption delay based on the initial appearance of a word provide an undesirable incentive for providers to prematurely deliver inaccurate captions? Conversely, would measuring caption delay based on the final displayed appearance provide an undesirable disincentive to correct mistakes in previously delivered captions?
6. Caption delay may vary over the course of a call. The Commission proposes that testing procedures should ensure that caption delay measurements for any service include measurements taken from various segments in the duration of captioned calls. The Commission seeks comment on the above proposals and their costs and benefits. Should caption delays during a single test call be averaged together, with each test call given a score, and the score for each test call given equal weight in the overall average? Or should caption delay be averaged on some other basis, e.g., total delays divided by the total number of minutes tested? Should seconds be measured to the nearest tenth of a second or some other measure? Also, for IP CTS, what internet speeds should be used to measure caption delay? Should delay be measured at more than one internet speed?
7. The Commission seeks comment on setting the applicable metric, i.e., the maximum average caption delay that should be allowed by the FCCs minimum TRS standards. Testing of
VerDate Sep<11>2014

16:45 Jan 29, 2021

Jkt 253001

fully automatic telephone captioning indicates that such services are capable of delivering captions within one or two seconds, on average. How many seconds of delay should be considered the maximum acceptable delay for any form of captioning, in light of the capabilities of current technology, the expectations of caption consumers, and the impact of delay on a users ability to carry on a natural telephone conversation? Should the FCCs minimum standards specify other limits on caption delay, in addition to the maximum average delay?
8. Accuracy. The Commission proposes to amend its rules to provide more specific standards and metrics for the accuracy of telephone captioning, including fully automatic IP CTS with captions created by an automatic speech recognition ASR program. The Commission proposes to combine accuracy with completeness in a single metric, Word Error Rate, which is likely to be easier to administer. Word Error Rate is comprised of individual counts of words that are incorrectly inserted, deleted, or substituted in the captions delivered to the caller.
9. For purposes of measuring compliance with the standard, the Commission seeks comment on the following definition of Word Error Rate:
The Word Error Rate for a captioned telephone conversation is i the number of word substitutions, omissions, and insertions in the captions divided by ii the total number of words in the voice communications being captioned. Accuracy shall be assessed for a caption as delivered to the caption users device within the minimum TRS standard for caption delay. A
substitution error occurs when a spoken word is replaced with another word, an omission error involves the omission of a spoken word, and an insertion error consists of the addition of a word that has not been spoken.

10. The Commission seeks comment on this proposal and its costs and benefits. To implement this definition of Word Error Rate, should the Commission define what constitutes a word? For example, should interjected sounds such as umm and ah or garbled speech count as words?
If a speaker uses a regional dialect or foreign phrase that has no standard English spelling, can there be an error in transcription? The Commission also seeks comment on whether to insert a qualifier in the above definition to limit the word errors that are counted to major errors, which a group of IP CTS
providers define as errors that significantly alter, obscure, or reverse the meaning of the original speech. Does this definition provide a consistent, repeatable determination of what
PO 00000

Frm 00028

Fmt 4702

Sfmt 4702

7683

constitutes a major error, and if not, can it be modified to do so? Would limiting counted errors to major errors produce materially different results in the overall assessment of CTS and IP
CTS providers? More specifically, would any improvements from counting only major errors be sufficient to justify 1 the additional costs and burdens involved in classifying errors as major or minor and 2 the greater likelihood of disputes over which errors count as major errors?
11. Alternatively, if a distinction is needed between major and minor errors, should minor errors i.e., word substitutions such as misspellings, deletions, or insertions that do not alter or obscure the meaning of the original speech still be counted but given less weight than major errors? For example, even though minor errors may not prevent a user from understanding the gist of a conversation, they still may be a distraction and force the CTS or IP
CTS user to work harder to decipher the captions. Or should the standard the Commission adopts be based on a combination of two measurements, one that is limited to major errors and one that takes all errorsincluding substitutions, deletions, and insertions whether major or minorinto account?
12. Should readability a concept that includes correct capitalization and punctuation be included in the Word Error Rate standard, and if so, how should it be measured?
13. The Commission also seeks comment on the maximum Word Error Rate that should be specified for caption service in the FCCs minimum TRS
standards, and how this standard should apply to variable call conditions.
Should the Commission set the accuracy standard based on the expectations of users and the impact of inaccuracies on a users ability to carry on a natural telephone conversation, and if so, how should these be determined?
Alternatively, in order to set an initial standard as expeditiously as possible, should the Commission initially set the maximum permitted Word Error Rate based on the current performance of IP
CTS providers, and subsequently reset the standard based on measures of user expectations and understanding? If a current-performance-based approach is initially used, should the maximum level be set based on the Word Error Rate achieved by the average provider, or at some other defined value on the spectrum of baseline accuracy measurements? Should a different standard be applied to calls with poor audio quality? How would such a determination be made?

E:FRFM01FEP1.SGM

01FEP1