The information about spoken content that's based on purely the media resource contents (and not the embedding page or context, etc).

Attributes

• audioDuration (type: integer(), default: nil) - Duration of audio in processed fragment (including non-speech), in seconds.
• audioOnly (type: boolean(), default: nil) - If true, the media file is audio-only. If false, also has video track(s).
• estimatedAudioDuration (type: integer(), default: nil) - Estimated duration of audio in the whole file (including non-speech), in seconds. If this is greater than 0, then it will either be equal to audio_duration (when truncated_file is false), or to the length of the content (audio or video) according to the file header (when truncated_file is true).
• estimatedAudioDurationConfidence (type: number(), default: nil) - Our confidence in the duration estimate, on a scale from 0 (not confident) to 1 (very confident). An estimate should have a confidence of at least 0.5 if it is to be shown to users.
• languageCode (type: String.t, default: nil) - The spoken language, see i18n/identifiers/languagecode.h and go/gl2014. This may or may not match the language of the written page. (Examples: "en", "sv", "zh-CN").
• numWords (type: integer(), default: nil) - Total number of recognized words in processed fragment.
• recognizerAccuracy (type: number(), default: nil) - Measure of the estimated output accuracy from the speech recognition code, from 0 to 1. Based on word-level confidence and possibly other factors.
• speechDuration (type: integer(), default: nil) - Duration of speech in processed fragment, in seconds.
• truncatedFile (type: boolean(), default: nil) - If true, we may have processed a truncated file (most likely due to a size-cutoff when crawling). As a result, the audio duration is a lower bound and the other fields reflect only the processed prefix of the file.