Research data
How people type, speak and dictate, in data you can check.
Open datasets on typing speed, speech recognition accuracy, speaking rates, prices and privacy. Every data point is a published figure with the exact words of its source, and every statistic is calculated from those data points. Download any dataset and reuse it with credit.
Measured by us
Benchmarks we run ourselves on public test data, with every transcript, timing and recording checksum published so anyone can recompute the results.
Compiled from published research
- Typing Speed Statistics: Average WPM by GroupHow fast do people type on a computer keyboard, and how does it differ between groups and tasks?14 data points · 6 sources
- Speech vs Typing Speed: Text Entry Rates ComparedHow fast can people enter text by keyboard, touchscreen, handwriting, voice and other methods?57 data points · 25 sources
- Speaking Rate Statistics: Words per MinuteHow many words per minute do people speak in conversation, presentations, reading aloud and recordings?7 data points · 5 sources
- Speech Recognition Accuracy: Word Error RatesHow accurate is speech recognition, by system, test set, language and speaking condition?597 data points · 32 sources
- Speech Recognition Speed: Real-Time Factor DataHow much faster than real time do speech recognition models transcribe audio, and on which hardware?24 data points · 0 sources
- How Many Languages Speech Recognition SupportsHow many languages can each dictation feature, voice assistant, speech API and open model recognize?45 data points · 39 sources
- Speech-to-Text API Pricing: Cost per Hour of AudioWhat do speech-to-text APIs and transcription services charge per hour of audio?34 data points · 15 sources
- Dictation Software Prices: What Each App CostsWhat do dictation and voice typing apps cost, by subscription and to buy outright?37 data points · 13 sources
- How Long Voice Assistants Keep Your RecordingsHow long do voice assistants and dictation services keep recordings and transcripts, and are they reviewed by people or used for training?38 data points · 20 sources
- Repetitive Strain Injury Statistics: How Common Is RSIHow common are repetitive strain injuries and hand, wrist, arm and neck pain among computer users and workers?18 data points · 9 sources
- Disability Statistics: Conditions That Make Typing HardHow many people live with conditions that make typing or handwriting hard, where voice input can help?22 data points · 10 sources
- Voice Assistant and Voice Search Usage StatisticsHow many people use voice assistants, voice search and dictation, by country, age and year?2 data points · 2 sources
Method
From a search to a sentence, every step on the record.
The pages in the library are the last step of a research process, not the first. The numbers on the right are live counts of each stage.
- 1
Search95
Each dataset has a written search protocol, run against OpenAlex, Semantic Scholar and arXiv for research, and against the official pages of every company it covers. 2,342 works found so far.
- 2
Screen1,008
A fixed rule sets aside works whose abstract reports no figure in the dataset's terms. A researcher reads the rest against the dataset's inclusion rule and records why each one is kept or not (86 judged, 65 kept).
- 3
Extract5,684
Every figure is saved as a fact: the number, a plain sentence, and the exact words of its source with where they are.
- 4
Verify5,328
Each quote is checked by code, word for word, against the live source before anything is published. A fact whose source changes is flagged.
- 5
Structure921
Comparable figures become data points: the number in one unit, with the group, device, test or year that makes it comparable.
- 6
Calculate53
Medians, ranges and counts are calculated by code from the data points, never typed by hand, and each one lists the values it summarises.
- 7
Write456
Guides and answers are written from the data: a sentence states a number through its fact or a live statistic, so every figure on a page drills down to the dataset and the source. This many pages already draw on the datasets.
Reuse
Free to reuse, with credit.
The datasets are released under CC BY 4.0: copy, share and adapt them for any purpose, including commercially, if you credit the dataset with a link. Each dataset page has the credit line to copy and CSV and JSON downloads with the full provenance of every row.
Quoted passages remain the words of their authors. When you rely on a single figure, cite its original source, which every fact page names and links.
Found a figure that is wrong, out of date or missing? Tell us and it will be checked against its source and corrected.
5,684
All facts
Every number we use, with the exact words of its source.
1,830
All sources
The studies, standards and official pages behind the facts.