diphone

Adopted diphone sense

Record whether the application uses diphone for an adjacent-phone transition type or for a bounded acoustic realization, and identify the authority for that usage.

1. What definition does the relevant linguistic reference or synthesis system adopt, and what qualifies as an instance? definition
2. Does this record describe a transition type, an original audio token, or a transformed token? boundary

Phone label resolution

Record ordered labels, inventory version, language variety, and any mapping between phonemic input and realized phones.

1. Which inventory and language variety define the first and second phone labels? provenance
2. Do the labels denote realized phones, phonemes, or synthesis-specific variants, and how are these mapped? definition

Endpoint convention

Distinguish automatic temporal midpoints from deliberately selected cut points. FestVox documents midpoint defaults with explicit boundary overrides: [diphone database construction](https://www.cs.cmu.edu/afs/cs/user/lenzo/html/areas/papers/festvox/festvox_5.html).

1. Were the start and end selected by temporal midpoint, acoustic stability, or another documented rule? measurement
2. How should endpoints be selected when a short or rapidly changing phone has no clear stable region? action

Transition landmarks

Record the internal boundary, its uncertainty, and acoustic evidence that the extracted interval retains the intended transition.

1. Where are the start, internal phone boundary, and end on the source recording's time base? measurement
2. What acoustic evidence supports the phone boundary, and how much annotation uncertainty remains? measurement

Variant selection conditions

Record when allophonic or contextual variants need separate selection. Festival documents augmented diphone identities, including distinctions between light and dark lateral realizations: [diphone selection](https://www.cstr.ed.ac.uk/projects/festival/manual/festival_21.html).

1. Which stress, allophone, syllable-position, or cluster distinctions change the required realization of this pair? boundary
2. What evidence permits one token to substitute for another contextual variant? action

Coverage and exceptions

Separate required, attested, excluded, missing, and convention-dependent pairs, including repeated labels and silence-related units.

1. Which target pronunciations or phonotactic rules establish that this ordered pair is required? provenance
2. Does the application include silence, utterance-edge units, and repeated-phone pairs, and which required cases lack usable tokens? boundary

Recoverable source interval

Link the token to the source utterance, speaker or voice, elicitation context, annotations, and processing history.

1. Which recording, speaker, elicitation text, and source interval produced this token? provenance
2. Which labeling, resampling, trimming, or signal-processing steps changed the token or its timing coordinates? provenance

Token acceptance evidence

Record review of phone identity, retained transition, clipping, noise, and relevant voicing landmarks before accepting a token.

1. Does listening and acoustic inspection support the intended phones and transition rather than merely the expected transcript? measurement
2. Which observed defects require boundary correction, relabeling, exclusion, or rerecording? action

Join suitability

Assess whether adjacent diphones share compatible realizations of their common phone and whether their acoustic endpoints produce an acceptable join.

1. Do the neighboring units agree on the shared phone's inventory identity and required contextual realization? boundary
2. What listening results and declared spectral, amplitude, or voicing measures support accepting the join? measurement

Permitted token adaptation

Record target duration and pitch, synthesis-method requirements, tested modification limits, and explicit fallback conditions.

1. What duration and voiced-region pitch changes have been evaluated for this token with the intended synthesis method? measurement
2. When a token is absent or fails quality criteria, should the agent select another variant, invoke a documented fallback, or request new recording? action

Check these first

Recalled without web access and unsourced; every item is a lead to verify.

1. Which of these check these first hold for the sense of diphone this model covers, and on what evidence? provenance

Real-world use

Recalled without web access and unsourced; every item is a lead to verify.

1. Which of these real-world use hold for the sense of diphone this model covers, and on what evidence? provenance

Typical measurements

Recalled without web access and unsourced; every item is a lead to verify.

1. Which of these typical measurements hold for the sense of diphone this model covers, and on what evidence? provenance

Failure modes and hazards

Recalled without web access and unsourced; every item is a lead to verify.

1. Which of these failure modes and hazards hold for the sense of diphone this model covers, and on what evidence? provenance

Regional variation

Recalled without web access and unsourced; every item is a lead to verify.

1. Which of these regional variation hold for the sense of diphone this model covers, and on what evidence? provenance

Neighbouring kinds and how to tell them apart

Recalled without web access and unsourced; every item is a lead to verify.

1. Which of these neighbouring kinds and how to tell them apart hold for the sense of diphone this model covers, and on what evidence? provenance