Display apparatus, server, system and information-providing methods thereof

1. An electronic apparatus comprising:

a memory configured to store an instruction; and

a processor configured to execute the instruction to:

extract video features of frames from a plurality of frames of video content,

extract text features from transcript information comprising text of the video content, the transcript information being provided separately from the video content,

identify importance scores of the frames based on the video features and the text features,

select at least one key frame of the video content among the plurality of frames based on the identified importance scores, and

produce a video summary with the selected at least one key frame,

wherein the processor is configured to:

predict the importance scores of the frames corresponding to association with the text features using a model learned by performing machine learning, and

select the at least one key frame to produce the video summary based on the predicted importance scores corresponding to association with the text features, the produced video summary being semantically coherent with respect to the transcript information.

2. The electronic apparatus according to claim 1 , wherein the processor is configured to create the model, a configuration of the model being selected and adjusted through iterative training of the machine learning based on the video features and text features as input values and importance scores of the frames as ground truth to predict the importance scores of the frames corresponding to association with the text features, and predict the importance scores of the frames based on the created learned-model.

3. The electronic apparatus according to claim 2 , wherein the processor is configured to perform the machine learning based on a recurrent neural network for a time-series analysis.

4. The electronic apparatus according to claim 1 , wherein the processor is configured to extract the video features based on at least one of histogram of oriented gradient (HOG), scale invariant feature transform (SIFT), a color histogram, texture extraction, or machine learning.

5. The electronic apparatus according to claim 1 , wherein the processor is configured to extract the text features based on at least one of bag of words (Bow), term frequency-inverse document frequency (TF-IDF), word-embedding, or machine learning based models.

6. The electronic apparatus according to claim 1 , wherein the processor is configured to:

perform cleaning by removing or stemming a stop word from the text comprising the transcript information of the video content, and

extract the text features from the cleaned text.

7. The electronic apparatus according to claim 1 , further comprising a display configured to display the produced video summary.

8. A method of controlling an electronic apparatus, comprising:

extracting, by a processor, video features of frames from a plurality of frames of video content;

extracting, by the processor, text features from transcript information comprising text of the video content, the transcript information being provided separately from the video content;

identifying, by the processor, importance scores of the frames based on the video features and the text features;

selecting, by the processor, at least one key frame of the video content among the plurality of frames based on the identified importance scores; and

producing a video summary with the selected at least one key frame,

wherein the identifying of the importance scores comprises predicting the importance scores of the frames corresponding to association with the text features using a model learned by performing machine leaning, and

wherein the selecting of the at least one key frame comprises selecting the at least one key frame to produce the video summary based on the predicted importance scores corresponding to association with the text features, the produced video summary being semantically coherent with respect to the transcript information.

9. The method according to claim 8 , wherein the identifying the importance scores comprises:

creating the model, a configuration of the model being selected and adjusted through iterative training of the machine learning based on the video features and text features as input values and importance scores of the frames as ground truth to predict the importance scores of the frames corresponding to association with the text features; and

predicting the importance scores of the frames based on the created learned-model.

10. The method according to claim 9 , wherein the creating of the model comprises performing the machine learning based on a recurrent neural network for a time-series analysis.

11. The method according to claim 8 , wherein the extracting of the video features comprises extracting the video features based on at least one of histogram of oriented gradient (HOG), scale invariant feature transform (SIFT), a color histogram, texture extraction, or machine learning.

12. The method according to claim 8 , wherein the extracting the text features comprises extracting the text features based on at least one of bag of words (Bow), term frequency-inverse document frequency (TF-IDF), word-embedding, or machine learning based models.

13. The method according to claim 12 , wherein the extracting of the text features comprises:

performing cleaning by removing or stemming a stop word from the text comprising the transcript information of the video content; and

extracting the text features from the cleaned text.

14. The method according to claim 8 , further comprising:

displaying the produced video summary.