o Si @sdZddlZddlmZddlmZmZmZddlm Z ddl m Z m Z ddl mZddlmZmZdd lmZ  dd ed eed eedeeee efffddZdS)aT Data preparation recipe for CMU Kids corpus (https://catalog.ldc.upenn.edu/LDC97S63): Summary of corpus from LDC webpage: This database is comprised of sentences read aloud by children. It was originally designed in order to create a training set of children's speech for the SPHINX II automatic speech recognizer for its use in the LISTEN project at Carnegie Mellon University. The children range in age from six to eleven (see details below) and were in first through third grades (the 11-year-old was in 6th grade) at the time of recording. There were 24 male and 52 female speakers. There are 5,180 utterances in all. The speakers come from two separate populations: 1. SIM95: They were recorded in the summer of 1995 and were enrolled in either the Chatham College Summer Camp or the Mount Lebanon Extended Day Summer Fun program in Pittsburgh. They were recorded on-site. There are 44 speakers and 3,333 utterances in this set. They "good" reading examples. 2. FP: These are examples of errorful reading and dialectic variants. The readers come from Fort Pitt School in Pittsburgh and were recorded in April 1996. There are 32 speakers and 1,847 utterances in this set. The user should be aware that the speakers' dialect partly reflects what is locally called "Pittsburghese." The corpus does not come with a train/dev/test split, and the Kaldi recipe splits it randomly into 70%/30% train-test. We do not perform any splits, and just return the complete recording and supervision manifests. This data is not available for free - your institution needs to have an LDC subscription. N)Path)DictOptionalUnion)$validate_recordings_and_supervisions) Recording RecordingSet) fix_manifests)SupervisionSegmentSupervisionSet)PathlikeT corpus_dir output_dirabsolute_pathsreturncCst|tr t|n|}|jdkr|jn|}g}g}i}t|dddd}|D]}|d\}} } | ||<q)Wdn1sDwYi} t|dddd2}tdD]} t |q[|D]}|d\} }}} } |d \}}|||f| | <qdWdn1swYt|ddd d}|D]t}|jd d \}}|d d} |dd}t |d}| | \}}}|dd| d|d}t j ||rdndd}| || t||d |j| | d dkrdndd|||dkr|nd|dkrt |nd|||dd qWdn 1swYt|}t|}t||\}}t||||d}|duretdt|}|jddd|d |d!|d"|d#|S)$a Prepare manifests for CMU Kids corpus. The prepared supervisions contain the prompt text as the `text`. Additionally, in the `custom` tag, we provide the following data: speaker grade/age, population where the speaker came from (SIM95/FP), spoken transcript, and transcription bin (1/2). Here, bin `1` means utterances where the speaker followed the prompt and no noise/mispronunciation is present, and `2` refers to noisy utterances. The tag `spoken_transcript` is the transcription that was actually spoken. It contains noise tags and phone transcription in case the pronunciation differed from that in CMU Dict. :param corpus_dir: Path to downloaded LDC corpus. :param output_dir: Directory where the manifests should be written. Can be omitted to avoid writing. :param absolute_paths: Wheter to write absolute paths to audio sources (default = False) :return: A dict with manifests. The keys are: ``{'recordings', 'supervisions'}``. cmu_kidstablesz sentence.tblr Nz speaker.tbl/z transcrp.tbl)maxsplitrkidssignalz.sph)relative_path_depthmMaleFemaleEnglishNA) speaker_grade speaker_agespeaker_populationbinspoken_transcript) id recording_idstartdurationspeakergenderlanguagetextcustom) recordings supervisionsz Writing manifests to JSONL filesT)parentsexist_okr2z cmu-kids_recordings_all.jsonl.gzr3z"cmu-kids_supervisions_all.jsonl.gz) isinstancestrrstemparentopenstripsplitrangenextintr from_fileappendr r,rfrom_recordingsr from_segmentsr rlogginginfomkdirto_file)r rrr2r3 utterancesflineuttcountr0 speaker_info_spkpopgr_agegradeagetrn_id transcriptr' audio_path recording manifestsrYK/home/ubuntu/.local/lib/python3.10/site-packages/lhotse/recipes/cmu_kids.pyprepare_cmu_kids,s          $    r[)NT)__doc__rDpathlibrtypingrrrlhotser lhotse.audiorr lhotse.qar lhotse.supervisionr r lhotse.utilsr boolr7r[rYrYrYrZs(