o }oi @sLddlZddlmZddlmZd dedefdd Zd dedefd d ZdS) N)AbstractCTCDecoding) Hypothesis{Gz?subsampling_factor window_stridecsddtdtfdddddtdtffd d |d ur |St|tr,|gSt|trCt|d trCfd d|DSt|trct|d trct|d d trcfdd|DStdt|)aA Processes AED timestamp outputs and extracts word-level timestamps. Args: outputs (Hypothesis, list of Hypotesis or list of list of Hypotesis): The hypothesis outputs to process. Can be a single Hypothesis object or a list of Hypothesis objects. subsampling_factor (int, optional): The subsampling factor used in the model. Default is 1. window_stride (float, optional): The window stride used in the model. Default is 0.01. Returns: list of list of Hypotesis: The processed hypothesis outputs with word-level timestamps added. rrrrc Ss|}td|sd|fSd}g}g}t||D];}t|d}|d}t|d} |||} | ||} |rU||| | | d} || ||qd|}||fS)Nz<\|\d+\|>.*?<\|\d+\|>z<\|(\d+)\|>(.*?)<\|(\d+)\|>r)word start_offset end_offsetstartend )stripresearchfinditerintgroupappendjoin) textrrpatternmatchestext_without_timestampsmatchr contentr start_timeend_timesampler!d/home/ubuntu/.local/lib/python3.10/site-packages/nemo/collections/asr/parts/utils/timestamp_utils.pyextract_words_with_timestamps s0      zDprocess_aed_timestamp_outputs..extract_words_with_timestampscSs6|D]}|d|||d<|d|||d<q|S)Nr r r rr!)segmentsrrsegmentr!r!r"segments_offset_to_timeAsz>process_aed_timestamp_outputs..segments_offset_to_timecs|j||\}}||_|dur8t|jdkri|_|j|ggdtj|gdd}||||jd<|Sgggd|_|S)zM Processes a single Hypothesis object to extract timestamps. Nr)r r%char).?!)segment_delimiter_tokensr%)rlen timestampupdater_get_segment_offsets)hyprrr-rr$)r#r&r!r"process_hypothesisGs&z9process_aed_timestamp_outputs..process_hypothesisNrcg|]}|qSr!r!.0r0r1rrr!r" kz1process_aed_timestamp_outputs..cs"g|] }fdd|DqS)cr2r!r!r3r5r!r"r6or7z..r!)r4 hyps_listr5r!r"r6nszRExpected Hypothesis, list of Hypothesis or list of list of Hypothesis object, got rr)rfloat isinstancerlist ValueErrortype)outputsrrr!)r#r1r&rrr"process_aed_timestamp_outputss  ! * r@cCs|dur|St|tr|g}t|dts tdt|ddd}t|D]F\}}t|ds8td||j}d|vrL||d||||jd<d |vr]||d ||||jd <d |vrn||d ||||jd <q(|S) a Process the timestamps from list of hypothesis to user friendly format. Converts the start and end duration from frames to seconds. Args: outputs: List of Hypothesis objects. subsampling_factor: int, Subsampling factor used in the model. window_stride: float, Window stride used in the model. (sometimes referred to as hop length/shift) Returns: List of Hypothesis objects with processed timestamps Nrz Expected Hypothesis object, got c SsNt|D] \}}|d}|d}|||}|||}||d<||d<q|S)zv Process the timestamp for a single hypothesis. return the start and end duration in seconds. r r r r) enumerate) r-rridxvalr r r rr!r!r"process_timestamps   z4process_timestamp_outputs..process_timestampr-zyExpected Hypothesis object to have 'timestamp' attribute, when compute_timestamps is enabled but got r r'r%)r;rr=r>rAhasattrr-)r?rrrDrBr0r-r!r!r"process_timestamp_outputsws2    rFr9) r2nemo.collections.asr.parts.submodules.ctc_decodingr+nemo.collections.asr.parts.utils.rnnt_utilsrrr:r@rFr!r!r!r"s   b