Techniques for generating natural language text customized to linguistic preferences of a user
Described herein are techniques for generating natural language text customized to linguistic preferences of a user. Customizing the generation of natural language text to the linguistic preference of a user can significantly improve the overall user experience. Some embodiments relate to techniques for learning the linguistic preferences of a user, and for assisting NLG systems to generate natural text that reflects more closely the linguistic preferences of the user. A linguistic preference learner can present different natural language options to a user, and can ask the user to select the option that or appears to reflect more closely the user's personal linguistic preferences. The Linguistic preference learner may determine, based on the user selection, information relating to what linguistic characteristics the user appears to prefer.
1. A method for generating natural language text customized to linguistic preferences of a user, the method comprising:
using at least one computer hardware processor to perform:
determining the linguistic preferences of the user at least in part by:
accessing a first plurality of natural language text segments;
accessing current values for a plurality of variables representing the linguistic preferences of the user;
presenting, to the user, a first graphical user interface (GUI) comprising a first plurality of selectable GUI elements, each of at least some of the first plurality of selectable GUI elements being associated with a corresponding natural language text segment in the first plurality of natural language text segments;
obtaining, via the first GUI, a first selection made by the user of at least one of the first plurality of selectable GUI elements in the first GUI;
determining a respective first score for each of at least some of the first plurality of natural language text segments to obtain first scores, the determining performed using:
the first selection, and
at least one of the current values for the plurality of variables representing the linguistic preferences of the user; and
determining updated values for the plurality of variables representing the linguistic preferences of the user using the first scores and at least some of the current values;
generating natural language text customized to the user using the updated values; and
outputting the natural language text customized to the user.
2. The method of claim 1 , wherein generating the natural language text comprises:
receiving, from a natural language generation (NLG) system, a second plurality of natural language text segments;
selecting at least one text segment from the second plurality of natural language segments using the updated values for the plurality of variables representing the linguistic preferences of the user; and
including the at least one text segment in the natural language text.
3. The method of claim 1 , wherein generating the natural language text comprises:
providing the updated values for the plurality of variables representing the linguistic preferences of the user to a natural language generation (NLG) system; and
generating, with the NLG system and using the updated values for the plurality of variables representing the linguistic preferences of the user, the natural language text.
4. The method of claim 1 , wherein determining the plurality of linguistic preferences of the user further comprises:
prior to generating the natural language text customized to the user:
presenting, to the user, a second GUI comprising a second plurality of selectable GUI elements, each of at least some of the second plurality of selectable GUI elements being associated with a corresponding natural language text segment in a second plurality of natural language text segments different from the first plurality of natural language text segments;
obtaining, via the second GUI, a second selection made by the user of at least one of the second plurality of selectable GUI elements in the second GUI;
determining a respective second score for each of at least some of the second plurality of text segments using:
the second selection, and
at least one of the updated values for the plurality of variables representing the linguistic preferences of the user,
wherein determining updated values for the plurality of variables representing the linguistic preferences of the user is performed also using the second scores.
5. The method of claim 4 , wherein:
each of the first plurality of natural language text segments represents a first fact pattern, and
each of the second plurality of natural language text segments represents a second fact pattern different from the first fact pattern.
6. The method of claim 1 , wherein accessing current values for the plurality of variables representing the linguistic preferences of the user comprises accessing current values for the plurality of variables representing at least one linguistic preference selected among the group consisting of a typographic preference, a syntactic preference, and a lexical preference, and
wherein generating the natural language text comprises generating, using the current values for the plurality of variables representing at least one linguistic preference selected among the group, the natural language text.
7. The method of claim 1 , wherein accessing current values for the plurality of variables representing the linguistic preferences of the user comprises accessing current values for the plurality of variables representing at least one linguistic preference selected among the group consisting of:
a number of syllables per word,
a number of characters per word,
a number of words per sentence,
a number of characters per sentence,
a number of words beginning with a vowel per sentence,
a number of lines per sentence,
a number of irregular words,
a number of verbs expressed in past tense,
a number of verbs expressed in present tense, and
a frequency of words appearing at least twice in a same sentence, and
wherein generating the natural language text comprises generating, using the current values for the plurality of variables representing at least one linguistic preference selected among the group, the natural language text.
8. The method of claim 1 , wherein the current values for the plurality of variables representing the linguistic preferences of the user are stored in a data structure associated with the user, the data structure comprising a set of fields storing the current values.
9. The method of claim 1 , wherein determining updated values for the plurality of variables representing the linguistic preferences of the user comprises:
selecting a subset of the first plurality of text segments using the at least some of the first scores; and
updating, using the selected subset of text segments, at least some of the current values to obtain updated values for the plurality of variables representing the linguistic preferences of the user.
10. The method of claim 9 , wherein selecting a subset of the first plurality of text segments using the at least some of the first scores comprises selecting the text segments having the highest N scores, with N>1.
11. The method of claim 9 , wherein determining updated values for the plurality of variables representing the linguistic preferences of the user comprises:
counting frequencies with which each of at least some of the selected subset of text segments satisfies the linguistic preferences of the user.
12. A system for generating natural language text customized to linguistic preferences of a user, the system comprising:
at least one computer hardware processor to perform:
determining the linguistic preferences of the user at least in part by:
accessing a first plurality of natural language text segments;
accessing current values for a plurality of variables representing the linguistic preferences of the user;
presenting, to the user, a first graphical user interface (GUI) comprising a first plurality of selectable GUI elements, each of at least some of the first plurality of selectable GUI elements being associated with a corresponding natural language text segment in the first plurality of natural language text segments;
obtaining, via the first GUI, a first selection made by the user of at least one of the first plurality of selectable GUI elements in the first GUI;
determining a respective first score for each of at least some of the first plurality of natural language text segments to obtain first scores, the determining performed using:
the first selection, and
at least one of the current values for the plurality of variables representing the linguistic preferences of the user; and
determining updated values for the plurality of variables representing the linguistic preferences of the user using the first scores and at least some of the current values;
generating natural language text customized to the user using the updated values; and
outputting the natural language text customized to the user.
13. The system of claim 12 , wherein generating the natural language text comprises:
receiving, from a natural language generation (NLG) system, a second plurality of natural language text segments;
selecting at least one text segment from the second plurality of natural language segments based on the updated values for the plurality of variables representing the linguistic preferences of the user; and
including the at least one text segment in the natural language text.
14. The system of claim 12 , wherein generating the natural language text comprises:
providing the updated values for the plurality of variables representing the linguistic preferences of the user to a natural language generation (NLG) system; and
generating, with the NLG system and based on the updated values for the plurality of variables representing the linguistic preferences of the user, the natural language text.
15. The system of claim 12 , wherein determining the plurality of linguistic preferences of the user further comprises:
prior to generating the natural language text customized to the user:
presenting, to the user, a second GUI comprising a second plurality of selectable GUI elements, each of at least some of the second plurality of selectable GUI elements being associated with a corresponding natural language text segment in a second plurality of natural language text segments different from the first plurality of natural language text segments;
obtaining, via the second GUI, a second selection made by the user of at least one of the second plurality of selectable GUI elements in the second GUI;
determining a respective second score for each of at least some of the second plurality of text segments using:
the second selection, and
at least one of the updated values for the plurality of variables representing the linguistic preferences of the user,
wherein determining updated values for the plurality of variables representing the linguistic preferences of the user is performed also using the second scores.
16. The system of claim 15 , wherein:
each of the first plurality of natural language text segments represents a first fact pattern, and
each of the second plurality of natural language text segments represents a second fact pattern different from the first fact pattern.
17. At least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by at least one processor, cause the at least one processor to perform a method for generating natural language text customized to linguistic preferences of a user, the method comprising:
using at least one computer hardware processor to perform:
determining the linguistic preferences of the user at least in part by:
accessing a first plurality of natural language text segments;
accessing current values for a plurality of variables representing the linguistic preferences of the user;
presenting, to the user, a first graphical user interface (GUI) comprising a first plurality of selectable GUI elements, each of at least some of the first plurality of selectable GUI elements being associated with a corresponding natural language text segment in the first plurality of natural language text segments;
obtaining, via the first GUI, a first selection made by the user of at least one of the first plurality of selectable GUI elements in the first GUI;
determining a respective first score for each of at least some of the first plurality of natural language text segments to obtain first scores, the determining performed using:
the first selection, and
at least one of the current values for the plurality of variables representing the linguistic preferences of the user; and
determining updated values for the plurality of variables representing the linguistic preferences of the user using the first scores and at least some of the current values;
generating natural language text customized to the user using the updated values; and
outputting the natural language text customized to the user.
18. The at least one non-transitory computer-readable storage medium of claim 17 , wherein generating the natural language text comprises:
receiving, from a natural language generation (NLG) system, a second plurality of natural language text segments;
selecting at least one text segment from the second plurality of natural language segments based on the updated values for the plurality of variables representing the linguistic preferences of the user; and
including the at least one text segment in the natural language text.
19. The at least one non-transitory computer-readable storage medium of claim 17 , wherein generating the natural language text comprises:
providing the updated values for the plurality of variables representing the linguistic preferences of the user to a natural language generation (NLG) system; and
generating, with the NLG system and based on the updated values for the plurality of variables representing the linguistic preferences of the user, the natural language text.
20. The at least one non-transitory computer-readable storage medium of claim 17 , wherein determining the plurality of linguistic preferences of the user further comprises:
prior to generating the natural language text customized to the user:
presenting, to the user, a second GUI comprising a second plurality of selectable GUI elements, each of at least some of the second plurality of selectable GUI elements being associated with a corresponding natural language text segment in a second plurality of natural language text segments different from the first plurality of natural language text segments;
obtaining, via the second GUI, a second selection made by the user of at least one of the second plurality of selectable GUI elements in the second GUI;
determining a respective second score for each of at least some of the second plurality of text segments using:
the second selection, and
at least one of the updated values for the plurality of variables representing the linguistic preferences of the user,
wherein determining updated values for the plurality of variables representing the linguistic preferences of the user is performed also using the second scores.