Systems and methods for automated generation of subtitles

1. A computer implemented method of presenting color coded text generated from an audio track of a video, the color coding denoting respective speakers, comprising:

receiving the audio track of the video divided into a plurality of audio-segments each representing speech spoken by a respective speaker of a plurality of speakers;

for each audio-segment of the plurality of audio-segments:

receiving a text representation of the audio-segment;

extracting a feature vector from the audio-segment;

mapping the feature vector to a color space;

coloring the text representation according to the color space; and

presenting the colored text representation in association with a video-segment corresponding to the audio-segment;

wherein mapping to feature vector to the color space comprises reducing a dimension of the feature vector to a dimension of the color space using a dimension reduction process;

wherein the reducing the dimension is performed by mapping function that maps a plurality of feature vectors of each respective speaker into values of a same color category that have a statistically significant distance from values of another color category of another speaker.

2. The method of claim 1 , wherein feature vectors of different audio-segments of a same speaker are reduced to different shades of a same color category.

3. The method of claim 1 , wherein feature vectors of different audio-segments of different speakers are each reduced to the color space of different color categories.

4. The method of claim 1 , wherein the feature vector is generated by a speaker verification system selected from the group consisting of: i-vector, x-vector, d-vector.

5. The method of claim 1 , wherein the mapping function is trained using a training dataset that includes different speakers and corresponding feature vectors.

6. The method of claim 1 , wherein the text representation of the audio-segment is automatically computed by an automated speech recognition (ASR) process.

7. The method of claim 1 , wherein the audio track is automatically divided into the plurality of audio-segments by a speaker diarization process.

8. The method of claim 1 , further comprising detecting an attribute of speech for each respective audio-segment, mapping the attribute of speech to instructions for adjusting characteristics of the font of the text, and wherein the colored text is presented according to the adjusted characteristics of the font of the colored text.

9. The method of claim 5 , wherein the mapping function is selected from the group consisting of: principal component analysis (PCA), Sammon mapping, t-SNE, and neural network.

10. The method of claim 5 , wherein the training dataset comprises a plurality of feature vectors computed for a plurality of audio-segments for a plurality of videos and a respective color category of a plurality of color categories designated for each of the plurality of speakers of the plurality of videos, wherein the mapping function is trained for maximizing a perceived color difference between the plurality of speakers.

11. The method of claim 5 , wherein the training dataset comprises a plurality of feature vectors computed for the plurality of audio-segments of the plurality of speakers extracted for the entire video, wherein the mapping function is trained for maximizing a perceived color difference between the plurality of speakers.

12. The method of claim 5 , wherein the video is further divided into a plurality of scenes, wherein the training dataset comprises a plurality of feature vectors computed for the plurality of audio-segments of the plurality of speakers extracted for each of the plurality of scenes, wherein the mapping function is trained for maximizing perceived color difference between the plurality of speakers.

13. A computer implemented method of presenting color coded text generated from an audio track of a video, the color coding denoting respective speakers, comprising:

receiving the audio track of the video divided into a plurality of audio-segments each representing speech spoken by a respective speaker of a plurality of speakers;

for each audio-segment of the plurality of audio-segments:

receiving a text representation of the audio-segment;

extracting a feature vector from the audio-segment;

mapping the feature vector to a color space;

coloring the text representation according to the color space;

presenting the colored text representation in association with a video-segment corresponding to the audio-segment; and

detecting an attribute of speech for each respective audio-segment, mapping the attribute of speech to instructions for adjusting characteristics of the background color of the text, and wherein the colored text is presented according to the adjusted characteristics of the background of the colored text.

14. The method of claim 13 , wherein the attributes of speech are selected from the group consisting of: speaking style, whisper, shout, calm, emotions, angry, excited, and sad.

15. The method of claim 13 , further comprising: receiving a respective attribute of each speaker, wherein the mapping the feature vector to a color space is according to the respective attribute of the speaker.

16. A computer implemented method of presenting color coded text generated from an audio track of a video, the color coding denoting respective speakers, comprising:

receiving the audio track of the video divided into a plurality of audio-segments each representing speech spoken by a respective speaker of a plurality of speakers;

for each audio-segment of the plurality of audio-segments:

receiving a text representation of the audio-segment;

extracting a feature vector from the audio-segment;

mapping the feature vector to a color space;

coloring the text representation according to the color space; and

presenting the colored text representation in association with a video-segment corresponding to the audio-segment;

wherein mapping the feature vector to the color space is computed based on a principal component analysis (PCA) process by: calculating a covariance matrix for the feature vector, selecting a number of eigenvectors with maximum eigenvalues, wherein the number of eigenvectors corresponds to a number of dimensions of the color space, wherein a projection of the respective feature vector along the respective eigenvalues is mapped to the color space.

17. A system for presenting color coded text generated from an audio track of a video, the color coding denoting respective speakers, comprising:

at least one hardware processor executing a code for:

receiving the audio track of the video divided into a plurality of audio-segments each representing speech spoken by a respective speaker of a plurality of speakers;

for each audio-segment of the plurality of audio-segments:

receiving a text representation of the audio-segment;

extracting a feature vector from the audio-segment;

mapping the feature vector to a color space;

coloring the text representation according to the color space; and

presenting the colored text representation in association with a video-segment corresponding to the audio-segment;

wherein mapping to feature vector to the color space comprises reducing a dimension of the feature vector to a dimension of the color space using a dimension reduction process;

wherein the reducing the dimension is performed by mapping function that maps a plurality of feature vectors of each respective speaker into values of a same color category that have a statistically significant distance from values of another color category of another speaker.

18. A computer program product for presenting color coded text generated from an audio track of a video, the color coding denoting respective speakers, comprising:

a non-transitory memory storing thereon code for execution by at least one hardware process, the code including instructions for:

receiving the audio track of the video divided into a plurality of audio-segments each representing speech spoken by a respective speaker of a plurality of speakers;

for each audio-segment of the plurality of audio-segments:

receiving a text representation of the audio-segment;

extracting a feature vector from the audio-segment;

mapping the feature vector to a color space;

coloring the text representation according to the color space; and

presenting the colored text representation in association with a video-segment corresponding to the audio-segment;

wherein mapping to feature vector to the color space comprises reducing a dimension of the feature vector to a dimension of the color space using a dimension reduction process;

wherein the reducing the dimension is performed by mapping function that maps a plurality of feature vectors of each respective speaker into values of a same color category that have a statistically significant distance from values of another color category of another speaker.