Speech Recognition, transcription (Ruan et al., 2016)

Measured by Ruan, Wobbrock, Liou, Ng & Landay · arXiv:1608.07323 (Stanford HCI) (2016)

Voice Speech recognition 2016

Inputs

The measured or assumed values behind the calculations, each with its source.

rate = 152.86 wpm: English text-entry rate via speech (Baidu Deep Speech 2) on a smartphone, including time to correct recognition errors. Keyboard baseline was 52.24 wpm; speech was 3.0× faster. This is a transcription (read-aloud) task, at the natural speaking-rate ceiling (~150 wpm); the separate composition (Foley 2020) and early-ASR (Karat 1999) speech entries give other points in the category.
H = 1.0 bits/char: English-text entropy (Shannon).

Strictest ITR

Each scoring method is an upper bound on the channel, so the headline is the strictest (smallest) one for this entry. Use the score selector on the home page to view any single method across entries.

Strictest Shannon (text) Recomputed

Character-entropy throughput

Modern ASR, transcription task (Ruan et al. 2016)

13 bits/s

Characters per minute

152.86 wpm × 5 chars/word = 764 chars/min

Bits per character
```
H(English) ≈ 1.0 bit/char (Shannon)
```
Recognition WER was 4.37% uncorrected (Deep Speech 2, English). The entry rate already includes error-correction time and participants corrected to near-perfect final text, so net throughput ≈ entry rate; the WER is documented here, not re-applied.

Information transfer rate

764 char/min × 1.0 bit/char ÷ 60 s/min = 12.7 bits/s

What counts as a bit depends on the action space. The number of distinguishable actions and how likely each one is are design choices of the task, not the sensing hardware. The same modality can present a fixed set of targets, a set pruned per step by a grammar or language model, or a continuous control space. Each of these changes how many actions are live and how the probability mass is spread, and therefore the information per selection. Read the action space below before comparing headline numbers across entries.

Action space

What the user can produce at each step, and how those options are distributed.

Structure: Context-dependent (the live set changes per step)
Size: Continuous
Prior: Context-conditioned: likelihoods depend on prior actions
Notes: Open-vocabulary automatic speech recognition with a language model in the loop. The bits/s figure uses an English character-entropy proxy on the produced text; the per-utterance action space is effectively unbounded, so this is not a fixed-target selection.

Comparability The strictest bound here is the Shannon entropy of the output text, under one predictor held constant across the whole atlas (≈1 bit per character). That shared predictor makes it directly comparable to every other text entry (keyboards, spellers, silent speech and speech BCIs) regardless of prior or vocabulary size. For most text interfaces it comes out tighter than the raw-selection bounds, but not always. Where a small vocabulary makes Wolpaw tighter, that wins instead. Any Fitts, Wolpaw or log₂(N) figure shown below is another bound on the same channel. Switch the home-page score selector to compare one across entries.

Source