Capturing a subjective viewpoint of a financial market analyst via a machine-learned model
Systems and methods herein provide for establishing a subjective viewpoint in text. In one embodiment, a method includes identifying intents and metrics in each of a plurality of texts, calculating a sentiment score for each text based on the identified intents and metrics of each text, and calculating a disfluency score for each text to weight the sentiment score of each text. The method also includes training the machine learning model with the texts, and processing a subsequent text through the trained machine learning model to determine a sentiment score of the subsequent text.
1. A method of establishing a subjective viewpoint of a text with a machine learning model, the method comprising:
identifying intents and metrics in each text of a plurality of texts;
calculating a sentiment score for each said text based on the identified intents and metrics of each said text;
calculating a disfluency score for each said text;
weighting the sentiment score of each said text based on the disfluency score;
training the machine learning model with the plurality of texts to create a trained machine learning model; and
processing a subsequent text through the trained machine learning model to determine a subsequent text sentiment score of the subsequent text.
2. The method of claim 1 , wherein: each said sentiment score is on a scale of −1 to +1, with −1 being most negative and +1 being most positive.
3. The method of claim 1 , wherein: identifying the intents and the metrics in each said text of the plurality of texts further comprises identifying the intents and the metrics in each said text of the plurality of texts via supervised learning.
4. The method of claim 1 , wherein: the disfluency score is a measure of a degree of fluency within each said text.
5. The method of claim 4 , wherein: repetitive words and filler words are identified by the model as disfluencies.
6. The method of claim 5 , further comprising: computing a ratio of the disfluencies to a total number of words to calculate the disfluency score for each said text.
7. The method of claim 1 , further comprising: transcribing a plurality of speech conversations with a natural language processor to correspondingly generate the plurality of texts.
8. A system for establishing a subjective viewpoint of a text with a machine learning model, the system comprising:
a database operable to store a plurality of texts that have each been labeled with identified intents and metrics;
a natural language processor operable to calculate a sentiment score for each text of the plurality of texts based on the identified intent labels and metrics of each said text, calculate a disfluency score for each said text, weight the sentiment score of each said text based on the disfluency score, train the machine learning model with the plurality of texts, and process a subsequent text through the trained machine learning model to determine a subsequent text sentiment score of the subsequent text.
9. The system of claim 8 , wherein: each said sentiment score is on a scale of −1 to +1, with −1 being most negative and +1 being most positive.
10. The system of claim 8 , wherein: the natural language processor is further operable to identify the intents and the metrics in each said text of the plurality of texts further via supervised learning.
11. The system of claim 8 , wherein: the disfluency score is a measure of a degree of fluency within each said text.
12. The system of claim 11 , wherein: repetitive words and filler words are identified by the model as disfluencies.
13. The system of claim 12 , further comprising: computing a ratio of the disfluencies to a total number of words to calculate the disfluency score for each said text.
14. The system of claim 8 , wherein: the natural language processor is further operable to transcribe a plurality of speech conversations to correspondingly generate the plurality of texts.
15. A non-transitory computer readable medium comprising instructions that, when executed by a processor, direct the processor to implement a machine learning model and to establish a subjective viewpoint of a text with the machine learning model, the instructions further directing the processor to:
identify intents and metrics in each text of a plurality of texts;
calculate a sentiment score for each said text based on the identified intents and metrics of each said text;
calculate a disfluency score for each said text;
weight the sentiment score of each said text based on the disfluency score;
train the machine learning model with the plurality of texts to create a trained machine learning model; and
process a subsequent text through the trained machine learning model to determine a subsequent text sentiment score of the subsequent text.
16. The non-transitory computer readable medium of claim 15 , wherein: each said sentiment score is on a scale of −1 to +1, with −1 being most negative and +1 being most positive.
17. The non-transitory computer readable medium of claim 15 , further comprising instructions that direct the processor to: identify the intents and the metrics in each of the plurality of texts further via supervised learning.
18. The non-transitory computer readable medium of claim 15 , wherein: the disfluency score is a measure of a degree of fluency within each said text.
19. The non-transitory computer readable medium of claim 18 , wherein: repetitive words and filler words are identified by the model as disfluencies.
20. The non-transitory computer readable medium of claim 19 , further comprising instructions that direct the processor to: compute a ratio of the disfluencies to a total number of words to calculate the disfluency score for each said text.