rixvox-alignments / README.md
Lauler's picture
Update README.md
c6947d0 verified

Variable descriptions

Variable Description
speech_id id of speech in SWERIK.
protocol_id id of protocol in SWERIK.
speech_number Speech number within a protocol.
dates Date(s) from SWERIK for the given protocol the speech was in.
name Speaker name
person_id id of the speaker/person in SWERIK.
speaker_id Same as person_id
riksdagen_id id of the person in Riksdagen's own system (if available).
party Party membership(s) of the speaker over their career.
district The district the person represents (if an MOP)
role Role(s) the speaker has had over their career.
gender Gender.
start_segment Start time of the speech within the media file (filename can be found in audio_file variable). Obtained by running speaker diarization on top of fuzzy string matching in order to refine existing timestamps from fuzzy string matching (the values in start_text_time). This is generally the most accurate estimate of when the speech was given. Time in seconds.
end_segment End time of the speech within the media file (according to diarization method).
duration_segment Duration of speech (diarization method).
text Original protocol transcript of speech from SWERIK.
text_normalized Normalized version of original protocol in a format suitable for string matching with wav2vec2 speech to text output (i.e. lowercase, only alphanumerics, convert numbers to spelled out letters, no punctuation).
transcription_w2v The automatically transcribed text in the region between start_segment and end_segment.
start_text_time Estimated start time of speech with fuzzy string matching (generally less accurate than start_segment)
end_text_time Estimated end time of speech with fuzzy string matching.
born Birth date of speaker.
dead Death date of speaker.
bleu_score BLEU score between text_normalized and transcription_w2v as an indication of the overlap.
overall_score Complicated score for fuzzy string matching that needs to be explained in a paper.
nr_speech_segments How many different speakers were identified by diarization within the duration of the speech (often but not always false negatives)
start_segment_same Is the start_segment of this speech a duplicate of some other speech's estimated start_segment? About ~150 speeches have the same identified start time. Generally either one of the two is correct, or there was a failure case of diarization that created too broad a speech duration that covered parts of two speeches.
audio_file File name of the media file in Riksdagen's media archive.