System and method for automatically generating concepts related to a target concept

1. A method for generating a set of concepts related to a target concept, the method comprising:

accessing a set of candidate concepts;

embedding the target concept and the set of candidate concepts in a semantic vector space; and

selecting one or more intermediate concepts from the set of candidate concepts in response to determining whether each embedded candidate concept in the set of embedded candidate concepts satisfies a predetermined relationship with the embedded target concept; wherein

the determining whether each embedded candidate concept in the set of embedded candidate concepts satisfies the predetermined relationship includes determining whether a displacement vector between each embedded candidate concept and the embedded target concept is within a predetermined range of distances and a predetermined range of directions; and

the predetermined range of distances and the predetermined range of directions are determined based on user information.

2. The method of claim 1 , further comprising filtering the one or more intermediate concepts to yield the set of concepts related to the target concept.

3. The method of claim 1 , wherein

the user information identifies capabilities of a user that receives the set of concepts related to the target concept; and

the user information is based on a history of user interactions with one or more previously generated sets of concepts.

4. The method of claim 1 , further comprising generating a multiple-choice question in which the target concept corresponds to a correct answer choice and the set of concepts related to the target concept correspond to distractors.

5. The method of claim 1 , further comprising generating a lesson plan in which the target concept corresponds to a main topic of the lesson plan and the set of concepts related to the target concept correspond to subtopics of the lesson plan.

6. The method of claim 1 , wherein the predetermined range of distances and the predetermined range of directions are determined based on relationship information that identifies a desired relationship between the target concept and the set of concepts related to the target concept.

7. The method of claim 1 , wherein the one or more intermediate concepts are selected using a neural network model that predicts whether each candidate concept in the set of candidate concepts satisfies the predetermined relationship based on the embedded target concept and the set of embedded candidate concepts.

8. The method of claim 1 , further comprising selecting, based on a subject domain of the target concept, a domain-specific concept library among a plurality of domain-specific concept libraries to provide the set of candidate concepts.

9. The method of claim 1 , further comprising selecting, based on a subject domain of the target concept, a domain-specific embedding model among a plurality of domain-specific embedding models to be used to embed the target concept and the set of candidate concepts in the semantic vector space.

10. The method of claim 1 , wherein the one or more intermediate concepts are filtered based on non-semantic features of the one or more intermediate concepts.

11. The method of claim 1 , further comprising providing the set of concepts related to the target concept to a user to curate the set of concepts.

12. A learning system comprising:

a non-transitory memory; and

one or more hardware processors coupled to the non-transitory memory and configured or programmed to execute instructions to cause the learning system to perform operations including:

receiving a target concept;

accessing a set of candidate concepts;

embedding the target concept and the set of candidate concepts in a semantic vector space; and

selecting one or more intermediate concepts from the set of candidate concepts in response to determining whether each embedded candidate concept in the set of embedded candidate concepts satisfies a predetermined relationship with the embedded target concept; wherein

the determining whether each embedded candidate concept in the set of embedded candidate concepts satisfies the predetermined relationship includes determining whether a displacement vector between each embedded candidate concept and the embedded target concept is within a predetermined range of distances and a predetermined range of directions; and

the predetermined range of distances and the predetermined range of directions are determined based on user information.

13. The learning system of claim 12 , wherein the operations further include filtering the one or more intermediate concepts to yield a set of related concepts that are related to the target concept.

14. The learning system of claim 12 , wherein

the user information identifies capabilities of a user that receives the set of concepts related to the target concept; and

the user information is based on a history of user interactions with one or more previously generated sets of concepts.

15. The learning system of claim 12 , wherein the operations further include generating a multiple-choice question in which the target concept corresponds to a correct answer choice and the set of related concepts correspond to distractors.

16. The learning system of claim 12 , wherein the operations further include generating a lesson plan in which the target concept corresponds to a main topic of the lesson plan and the set of related concepts correspond to subtopics of the lesson plan.

17. The learning system of claim 12 , wherein the one or more intermediate concepts are selected using a neural network model that predicts whether each candidate concept in the set of candidate concepts satisfies the predetermined relationship based on the embedded target concept and the set of embedded candidate concepts.

18. The learning system of claim 12 , wherein the predetermined relationship is determined based on one or more of relationship information that identifies a desired relationship between the target concept and the set of related concepts and the user information that identifies capabilities of the user that receives the set of related concepts.

19. The learning system of claim 12 , wherein the operations further include selecting, based on a subject domain of the target concept, a domain-specific concept library among a plurality of domain-specific concept libraries to provide the set of candidate concepts.

20. The learning system of claim 12 , wherein the operations further include selecting, based on a subject domain of the target concept, a domain-specific embedding model among a plurality of domain-specific embedding models to be used to embed the target concept and the set of candidate concepts in the semantic vector space.

21. The learning system of claim 12 , wherein the one or more intermediate concepts are filtered based on non-semantic features of the one or more intermediate concepts.

22. The learning system of claim 12 , wherein the operations further include providing the set of concepts related to the target concept to a user to curate the set of concepts.

23. A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a learning system to perform operations including:

identifying a target concept corresponding to a correct answer to a multiple-choice question;

accessing a set of candidate concepts;

embedding the target concept and the set of candidate concepts in a semantic vector space; and

selecting one or more intermediate concepts from the set of candidate concepts, the one or more intermediate concepts corresponding to one or more of the set of embedded candidate concepts that are usable as distractors for the multiple-choice question; wherein

the selecting of the one or more intermediate concepts from the set of candidate concepts includes determining whether a displacement vector between each embedded candidate concept and the embedded target concept is within a predetermined range of distances and a predetermined range of directions;

the predetermined range of distances and the predetermined range of directions are determined based on user information.

24. The non-transitory machine-readable medium of claim 23 , wherein the operations further include:

filtering the one or more intermediate concepts to yield a set of distractors for the multiple-choice question; and

providing the multiple-choice question with the set of distractors to a user of the learning system.

25. The non-transitory machine-readable medium of claim 23 , wherein

the user information identifies capabilities of a user that receives the set of concepts related to the target concept; and

the user information is based on a history of user interactions with one or more previously generated sets of concepts.

26. The non-transitory machine-readable medium of claim 23 , wherein the one or more intermediate concepts are filtered based on at least one of a part of speech of the target concept, a number of words of the target concept, or a capitalization of the target concept.

27. The non-transitory machine-readable medium of claim 23 , wherein the operations further include:

receiving a user response to the multiple-choice question; and

generating one or more second sets of distractors for one or more second multiple-choice questions, the one or more second sets of distractors being customized to the user based on the user response.