Intelligently summarizing and presenting textual responses with machine learning
This disclosure relates to methods, non-transitory computer readable media, and systems apply machine-learning techniques and computational sentiment analysis to summarize sentences extracted from a group of textual responses or to select representative-textual responses from the group of textual responses. By using a response-extraction-neural network to extract (and sometimes paraphrase) sentences from textual responses, the disclosed methods, non-transitory computer readable media, and systems can generate a response summary of textual responses based on sentiment indicators corresponding to the textual responses. By applying a machine-learning classifier to generate textual quality scores for textual responses, the disclosed methods, non-transitory computer readable media, and systems select representative-textual responses from a group of textual responses based on relevancy parameters and sentiment indicators corresponding to the textual responses. Such computational techniques generate response summaries and representative responses that provide an efficient snapshot of a group of textual responses analyzed by machine learners.
1. A system comprising:
at least one processor; and
at least one non-transitory computer readable storage medium storing instructions that, when executed by the at least one processor, cause the system to:
provide a set of textual responses to a text-quality classifier to generate a textual quality score for each textual response from the set of textual responses;
determine a relevancy parameter, a sentiment indicator, and a topic for each textual response from the set of textual responses, wherein a given relevancy parameter indicates a relevance of a given textual response to a user query, a given sentiment indicator indicates a linguistic sentiment of the given textual response, and a given topic indicates a subject matter of the given textual response;
generate a first response group of textual responses and a second response group of textual responses from the set of textual responses based on the sentiment indicator and the topic for each textual response from the set of textual responses; and
select, for display on a client device, a first representative-textual response from the first response group and a second representative-textual response from the second response group based on the relevancy parameter for each textual response within the first response group and the relevancy parameter for each textual response within the second response group.
2. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to:
sort the set of textual responses by:
selecting, from a response database, textual responses satisfying a textual-quality-score threshold; and
sorting the selected textual responses based on the relevancy parameter for each textual response from the selected textual responses to create a sorted set of textual responses; and
generate, from the sorted set of textual responses, the first response group of textual responses corresponding to a first range of sentiment indicators and the second response group of textual responses corresponding to a second range of sentiment indicators based on the sentiment indicator and the topic for each textual response from the sorted set of textual responses.
3. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to:
determine the sentiment indicator for each textual response by determining, for each textual response, a sentiment score indicating a positive sentiment or a negative sentiment;
determine, for each textual response, a sentiment-polarity indicator based on the sentiment score for each textual response satisfying or not satisfying a positive-sentiment-score threshold or a negative-sentiment-score threshold; and
generate the first response group of textual responses and the second response group of textual responses based on the sentiment score, the sentiment-polarity indicator, and the topic for each textual response from the set of textual responses.
4. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to:
determine the relevancy parameter for each textual response from the set of textual responses by determining, for each textual response, a topic-similarity score indicating a relevance of the textual response to a queried topic and a time identifier indicating a time associated with the textual response;
select, for display on the client device, the first representative-textual response from the first response group based on the textual quality score, the topic-similarity score, and the time identifier for each textual response within the first response group; and
select, for display on the client device, the second representative-textual response from the second response group based on the textual quality score, the topic-similarity score, and the time identifier for each textual response within the second response group.
5. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to:
provide the set of textual responses to the text-quality classifier by:
receiving, from a client device, a user query searching for textual responses corresponding to a time period; and
selecting the set of textual responses based on a time identifier for each textual response corresponding to the time period indicated by the user query.
6. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to determine the sentiment indicator for each textual response from the set of textual responses by determining, for each textual response, a sentiment score indicating a positive sentiment corresponding to the topic or a negative sentiment corresponding to the topic.
7. The system of claim 1 , further comprising instructions that, when executed by the at least one processor, cause the system to generate a response summary for the set of textual responses based on the first representative-textual response and the second representative-textual response by generating the response summary for the set of textual responses comprising both the first representative-textual response and the second representative-textual response.
8. A non-transitory computer readable medium storing instructions thereon that, when executed by at least one processor, cause a computer device to:
provide a set of textual responses to a text-quality classifier to generate a textual quality score for each textual response from the set of textual responses;
determine a relevancy parameter, a sentiment indicator, and a topic for each textual response from the set of textual responses, wherein a given relevancy parameter indicates a relevance of a given textual response to a user query, a given sentiment indicator indicates a linguistic sentiment of the given textual response, and a given topic indicates a subject matter of the given textual response;
generate a first response group of textual responses and a second response group of textual responses from the set of textual responses based on the sentiment indicator and the topic for each textual response from the set of textual responses; and
select, for display on a client device, a first representative-textual response from the first response group and a second representative-textual response from the second response group based on the relevancy parameter for each textual response within the first response group and the relevancy parameter for each textual response within the second response group.
9. The non-transitory computer readable medium of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computer device to:
sort the set of textual responses by:
selecting, from a response database, textual responses satisfying a textual-quality-score threshold; and
sorting the selected textual responses based on the relevancy parameter for each textual response from the selected textual responses to create a sorted set of textual responses; and
generate, from the sorted set of textual responses, the first response group of textual responses corresponding to a first range of sentiment indicators and the second response group of textual responses corresponding to a second range of sentiment indicators based on the sentiment indicator and the topic for each textual response from the sorted set of textual responses.
10. The non-transitory computer readable medium of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computer device to:
determine the sentiment indicator for each textual response by determining, for each textual response, a sentiment score indicating a positive sentiment or a negative sentiment;
determine, for each textual response, a sentiment-polarity indicator based on the sentiment score for each textual response satisfying or not satisfying a positive-sentiment-score threshold or a negative-sentiment-score threshold; and
generate the first response group of textual responses and the second response group of textual responses based on the sentiment score, the sentiment-polarity indicator, and the topic for each textual response from the set of textual responses.
11. The non-transitory computer readable medium of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computer device to:
determine the relevancy parameter for each textual response from the set of textual responses by determining, for each textual response, a topic-similarity score indicating a relevance of the textual response to a queried topic and a time identifier indicating a time associated with the textual response;
select, for display on the client device, the first representative-textual response from the first response group based on the textual quality score, the topic-similarity score, and the time identifier for each textual response within the first response group; and
select, for display on the client device, the second representative-textual response from the second response group based on the textual quality score, the topic-similarity score, and the time identifier for each textual response within the second response group.
12. The non-transitory computer readable medium of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computer device to:
provide the set of textual responses to the text-quality classifier by:
receiving, from a client device, a user query searching for textual responses corresponding to a time period; and
selecting the set of textual responses based on a time identifier for each textual response corresponding to the time period indicated by the user query.
13. The non-transitory computer readable medium of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computer device to determine the sentiment indicator for each textual response from the set of textual responses by determining, for each textual response, a sentiment score indicating a positive sentiment corresponding to the topic or a negative sentiment corresponding to the topic.
14. The non-transitory computer readable medium of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computer device to generate a response summary for the set of textual responses based on the first representative-textual response and the second representative-textual response by generating the response summary for the set of textual responses comprising both the first representative-textual response and the second representative-textual response.
15. A computer-implemented method comprising:
providing a set of textual responses to a text-quality classifier to generate a textual quality score for each textual response from the set of textual responses;
determining a relevancy parameter, a sentiment indicator, and a topic for each textual response from the set of textual responses, wherein a given relevancy parameter indicates a relevance of a given textual response to a user query, a given sentiment indicator indicates a linguistic sentiment of the given textual response, and a given topic indicates a subject matter of the given textual response;
generating a first response group of textual responses and a second response group of textual responses from the set of textual responses based on the sentiment indicator and the topic for each textual response from the set of textual responses; and
selecting, for display on a client device, a first representative-textual response from the first response group and a second representative-textual response from the second response group based on the relevancy parameter for each textual response within the first response group and the relevancy parameter for each textual response within the second response group.
16. The computer-implemented method of claim 15 , further comprising:
sorting the set of textual responses by:
selecting, from a response database, textual responses satisfying a textual-quality-score threshold; and
sorting the selected textual responses based on the relevancy parameter for each textual response from the selected textual responses to create a sorted set of textual responses; and
generating, from the sorted set of textual responses, the first response group of textual responses corresponding to a first range of sentiment indicators and the second response group of textual responses corresponding to a second range of sentiment indicators based on the sentiment indicator and the topic for each textual response from the sorted set of textual responses.
17. The computer-implemented method of claim 15 , further comprising:
determining the sentiment indicator for each textual response by determining, for each textual response, a sentiment score indicating a positive sentiment or a negative sentiment;
determining, for each textual response, a sentiment-polarity indicator based on the sentiment score for each textual response satisfying or not satisfying a positive-sentiment-score threshold or a negative-sentiment-score threshold; and
generating the first response group of textual responses and the second response group of textual responses based on the sentiment score, the sentiment-polarity indicator, and the topic for each textual response from the set of textual responses.
18. The computer-implemented method of claim 15 , further comprising:
determining the relevancy parameter for each textual response from the set of textual responses by determining, for each textual response, a topic-similarity score indicating a relevance of the textual response to a queried topic and a time identifier indicating a time associated with the textual response;
selecting, for display on the client device, the first representative-textual response from the first response group based on the textual quality score, the topic-similarity score, and the time identifier for each textual response within the first response group; and
selecting, for display on the client device, the second representative-textual response from the second response group based on the textual quality score, the topic-similarity score, and the time identifier for each textual response within the second response group.
19. The computer-implemented method of claim 15 , further comprising:
providing the set of textual responses to the text-quality classifier by:
receiving, from a client device, a user query searching for textual responses corresponding to a time period; and
selecting the set of textual responses based on a time identifier for each textual response corresponding to the time period indicated by the user query.
20. The computer-implemented method of claim 15 , further comprising determining the sentiment indicator for each textual response from the set of textual responses by determining, for each textual response, a sentiment score indicating a positive sentiment corresponding to the topic or a negative sentiment corresponding to the topic.