Company and program review
Way With Words review for speech transcription and annotation
Evaluate Way With Words for speech transcription and annotation, with transcript conventions and source rights in view.
- Commercial basis
- Transcription and annotation deliverables
- Review updated
- Provider website
- waywithwords.net ↗
The verdict
Consider it when the speech task depends on a particular transcript convention, timestamp grain or speaker label.
What to check: Verify which rights extend to transcripts, recordings and derivative uses.
Review basis. Fiund reviewed the cited public materials. Fiund also operates an AI data licensing business. This review includes no paid engagement or product rating.
Way With Words markets machine-learning transcription and annotation and publishes a speech dataset license document. A buyer who owns recordings should compare a processing service with the separate option to license speech from a dataset. The price and rights depend on the actual deliverable. Way With Words ML transcription and annotation · Published speech dataset license
Define the transcript by the model task
For diarization, require speaker identifiers, timestamps and an explicit convention for overlaps. For recognition of specialized language, provide a vocabulary and ask how reviewers flag uncertain terms or inaudible sections. Test agreement on a small sample and verify the export format and whether corrections can be traced.
Determine which rights move with the transcript
For buyer-supplied audio, confirm instructions, permitted processing, storage and return of files and labels. For a licensed speech dataset, verify permitted use, term and restrictions on sharing the underlying assets. Use the service terms and specific written quote for the final decision. Way With Words terms
An original transcript acceptance example
Use a fictional recording with two speakers, an interruption and a product code that can be confused with an ordinary word. Define how overlap, uncertain words, non-speech sounds and speaker changes appear in the reference. Ask the service to return a sample under those conventions and preserve review notes for unresolved terms.
Compare a model transcript with the checked reference only after applying the same normalization rule. Inspect the difficult code and interruption separately from the surrounding fluent speech. A lower aggregate error score can conceal the exact task failure the collection is intended to address.
Service processing and speech licensing need separate scope
For buyer-owned audio, identify the permitted processing, reviewer access, retention and return of files and labels. For a licensed corpus, identify the selected assets and the allowed model and recipient scope. A transcript derived from a recording does not erase restrictions attached to its source.
The published dataset licence is useful context with its own version and date. Obtain the current agreement and quote offered for the particular engagement. Ask which timestamps, speaker labels, human review and correction history are included and how rejected work is replaced. This review does not establish a universal rate or a tested private delivery.
Sources and further reading
Related reviews and guides
Related reading
Speech Evaluation Workbench
Related reading
Podcast archive inventory
Data licensing and catalogs
Defined.ai
A useful shortlist entry for catalog speech and language data. Compare its commissioned collection separately and verify permissions on the specific asset.
Managed collection and evaluation
LXT review for speech data and multilingual model evaluation
Useful to compare for speech collection when accents, acoustics and label conventions will affect the model.
Managed collection and evaluation
Nexdata review for speech recognition datasets and custom collection
Compare catalog speech datasets with commissioned collection by speaker mix, recording conditions and transcript detail.
Managed collection and evaluation
Pangeanic review for multilingual datasets and anonymization workflows
A shortlist option for multilingual data or anonymization when you can evaluate the language-specific transformations.