Search system and method having civility score
A scoring system and method identifies personal attacks in a piece of audio content and generates a civility score for the piece of audio content that can differentiate between personal attacks and vernacular/casual banter. The piece of audio content may be a podcast.
1. A system, comprising:
a computer system having a processor and a plurality of lines of instructions that are executed by the processor that is configured to:
retrieve a plurality of podcasts;
generate, for each podcast using a trained machine learning model, a civility score wherein the civility score for each podcast indicates a quantity of personal attacks in the podcast; and
communicate, via an application programming interface, the civility score for a particular podcast from the computer system to a second computer system.
2. The system of claim 1 , wherein the second computer system is an advertisement system that sends a request for the civility score for the particular podcast.
3. The system of claim 2 , wherein the advertisement system selects an advertisement for the particular podcast based on the civility score.
4. The system of claim 1 , wherein the processor is further configured to generate a user interface that displays an image representing the particular podcast, the civility score for the particular podcast and a change in civility since a prior episode of the particular podcast.
5. The system of claim 1 , wherein the personal attack is directed to a third party and is one of a profanity, derogatory language, a negative evaluation and a negative perspective on a situation.
6. The system of claim 1 , wherein the processor is further configured to separate a transcribed podcast into a plurality of pieces of data and train the trained machine learning model using a training data set formed based on the plurality of pieces of data.
7. The system of claim 6 , wherein each piece of data is a portion of the transcribed podcast when one person is talking during the transcribed podcast.
8. The system of claim 1 , wherein the trained machine learning model is a transformer model.
9. The system of claim 6 , wherein the processor configured to train the trained machine learning model is further configured to invoke a plurality of large language models to generate a label for each piece of data of the transcribed podcast for each large language model and aggregate the labels from each large language model into a final label for each piece of data that form the training data set.
10. A method, comprising:
retrieving, by a computer system having a processor and a plurality of lines of instructions that are executed by the processor, a plurality of podcasts;
generating, by the computer system for each podcast using a trained machine learning model, a civility score wherein the civility score for each podcast indicates a quantity of personal attacks in the podcast; and
communicating, via an application programming interface of the computer system, the civility score for a particular podcast from the computer system to a second computer system.
11. The method of claim 10 further comprising sending, by the second computer system that is an advertisement system, a request for the civility score for the particular podcast.
12. The method of claim 11 further comprising selecting, by the advertisement system, an advertisement for the particular podcast based on the civility score.
13. The method of claim 10 further comprising generating, by the computer system, a user interface that displays an image representing the particular podcast, the civility score for the particular podcast and a change in civility since a prior episode of the particular podcast.
14. The method of claim 10 , wherein the personal attack is directed to a third party and is one of a profanity, derogatory language, a negative evaluation and a negative perspective on a situation.
15. The method of claim 10 further comprising separating, by the computer system, a transcribed podcast into a plurality of pieces of data and training, by the computer system, the trained machine learning model using a training data set formed based on the plurality of pieces of data.
16. The method of claim 15 , wherein each piece of data is a portion of the transcribed podcast when one person is talking during the transcribed podcast.
17. The method of claim 15 , wherein training the trained machine learning model further comprises invoking a plurality of large language models to generate a label for each piece of data of the transcribed podcast for each large language model and aggregating the labels from each large language model into a final label for each piece of data that form the training data set.
18. The system of claim 1 , wherein the particular podcast is one of a podcast episode and an entire podcast.
19. The method of claim 10 , wherein the particular podcast is one of a podcast episode and an entire podcast.