In matching.py, I set "max_duration": 29, but some segments are longer than 29 seconds, and a few are even up to 200 seconds. Below is my input for matching.py (I only changed has_punctuation and max_duration):
params = AttributeDict( { # parameters for loading source texts # you can find the docs in textsearch/datatypes.py "use_utf8": False, "is_bpe": True, "use_uppercase": True, "has_punctuation": False, # parameters for aligning queries # you can find the docs in textsearch/match.py#align_queries "num_close_matches": 2, "segment_length": 5000, "reference_length_difference": 0.4, "min_matched_query_ratio": 0.33, # parameters for splitting aligned queries # you can find the docs in textsearch/match.py#split_aligned_queries "preceding_context_length": 1000, "timestamp_position": "current", "duration_add_on_left": 0.0, "duration_add_on_right": 0.5, "silence_length_to_break": 0.45, "overlap_ratio": 0.25, "min_duration": 2, "max_duration": 29, "expected_duration": (5, 20), "max_error_rate": 0.20, # output "output_post_texts": False, } )
In matching.py, I set "max_duration": 29, but some segments are longer than 29 seconds, and a few are even up to 200 seconds. Below is my input for matching.py (I only changed has_punctuation and max_duration):
params = AttributeDict( { # parameters for loading source texts # you can find the docs in textsearch/datatypes.py "use_utf8": False, "is_bpe": True, "use_uppercase": True, "has_punctuation": False, # parameters for aligning queries # you can find the docs in textsearch/match.py#align_queries "num_close_matches": 2, "segment_length": 5000, "reference_length_difference": 0.4, "min_matched_query_ratio": 0.33, # parameters for splitting aligned queries # you can find the docs in textsearch/match.py#split_aligned_queries "preceding_context_length": 1000, "timestamp_position": "current", "duration_add_on_left": 0.0, "duration_add_on_right": 0.5, "silence_length_to_break": 0.45, "overlap_ratio": 0.25, "min_duration": 2, "max_duration": 29, "expected_duration": (5, 20), "max_error_rate": 0.20, # output "output_post_texts": False, } )