Comprehension Modeling and AI-Sourced Student Content Recommendations

Abstract

A system generates a training dataset based on historical consumption information and historical comprehension information of historical users, and uses the training dataset to train a machine-learned model to predict a measure of comprehension for a user consuming educational content. The system applies the machine-learned model to behaviors of a target user to determine a target measure of comprehension for target educational content, identifies one or more characteristics of the target educational content, applies a content identification model to identify supplemental educational content, and generates an educational content interface to present the supplemental educational content to the target user. In some examples, the system trains the machine-learned model to predict a collective measure of comprehension for a set of users consuming educational content, identifies a set of supplemental educational content, and generates a teacher interface to present the set of supplemental educational content.

What is claimed is: 
     
         1 . A computer-implemented method, comprising:
 accessing, by a content server, historical consumption information describing historical educational content consumption behaviors of a set of historical users and historical comprehension information comprising a historical measure of comprehension of historical educational content by the set of historical users;   generating, by the content server, a training dataset based on the accessed historical consumption information and the historical comprehension information;   training, by the content server, a machine-learned model configured to predict a measure of comprehension for a user consuming educational content based on behaviors of the user as the user consumes the education content;   applying, by the content server, the machine-learned model to behaviors of a target user consuming target educational content to determine a target measure of comprehension for each of a plurality of portions of the target educational content; and   for a portion of the target educational content corresponding to a below-threshold target measure of comprehension:
 identifying, by the content server, one or more characteristics of the portion of the target educational content; 
 applying, by the content server, a content identification model to the identified characteristics of the portion of the target educational content to identify supplemental educational content related to the portion of the target educational content; and 
 generating, by the content server, an educational content interface to present the supplemental educational content to the target user. 
   
     
     
         2 . The method of  claim 1 , wherein the educational content consumption behaviors include one or more of reading rates, pause time, number of re-read times, delays in content consumption, highlighting, types of highlighting, highlight coverage, time of day, switch outs, and switch duration. 
     
     
         3 . The method of  claim 1 , wherein the measure of comprehension includes one or more of test scores, user evaluations of comprehension, types of highlighting, post-comprehension quiz results. 
     
     
         4 . The method of  claim 1 , wherein the content identification model is trained with a second dataset comprising historical educational content consumed by historical users, and the historical educational content includes one or more target educational content and supplemental educational content associated with each target educational content consumed by the historical users. 
     
     
         5 . The method of  claim 4 , further comprising:
 accessing the second dataset; and   training the content identification model to predict a likelihood of supplemental educational content that improves a user's measure of comprehension for the portion of the target educational content.   
     
     
         6 . The method of  claim 1 , wherein applying the machine-learned model to determine a measure of comprehension comprises:
 determining the measure of comprehension in real time as the target user consumes the portion of the target educational content.   
     
     
         7 . The method of  claim 1 , wherein generating the educational content interface to present the supplemental educational content to the target user comprises:
 displaying the supplemental educational content to the target user in real time as the target user consumes the portion of the target educational content.   
     
     
         8 . The method of  claim 1 , wherein generating the educational content interface to present the supplemental educational content to the target user comprises:
 displaying the supplemental educational content to the target user after the target user consumes the portion of the target educational content.   
     
     
         9 . The method of  claim 1 , further comprising:
 identifying second supplemental educational content related to a second portion of the target educational content, wherein the second portion and the portion of the target educational content are related to a low-comprehension dedicated section; and   modifying the educational content interface to add the second supplemental educational content for display.   
     
     
         10 . A computer system comprising:
 one or more computer processors; and   one or more computer-readable mediums storing instructions that, when executed by the one or more computer processors, cause the system to:
 access historical consumption information describing historical educational content consumption behaviors of a set of historical users and historical comprehension information comprising a historical measure of comprehension of historical educational content by the set of historical users; 
 generate a training dataset based on the accessed historical consumption information and the historical comprehension information; 
 train a machine-learned model configured to predict a measure of comprehension for a user consuming educational content based on behaviors of the user as the user consumes the education content; 
 apply the machine-learned model to behaviors of a target user consuming target educational content to determine a target measure of comprehension for each of a plurality of portions of the target educational content; and 
 for a portion of the target educational content corresponding to a below-threshold target measure of comprehension:
 identify one or more characteristics of the portion of the target educational content; 
 apply a content identification model to the identified characteristics of the portion of the target educational content to identify supplemental educational content related to the portion of the target educational content; and 
 generate an educational content interface to present the supplemental educational content to the target user. 
 
   
     
     
         11 . The system of  claim 10 , wherein the educational content consumption behaviors include one or more of reading rates, pause time, number of re-read times, delays in content consumption, highlighting, types of highlighting, highlight coverage, time of day, switch outs, and switch duration. 
     
     
         12 . The system of  claim 10 , wherein the measure of comprehension includes one or more of test scores, user evaluations of comprehension, types of highlighting, post-comprehension quiz results. 
     
     
         13 . The system of  claim 10 , wherein the instructions to apply the machine-learned model to determine a measure of comprehension comprise:
 determining the measure of comprehension in real time as the target user consumes the portion of the target educational content.   
     
     
         14 . The system of  claim 10 , wherein the instructions to generate the educational content interface to present the supplemental educational content to the target user comprise:
 displaying the supplemental educational content to the target user in real time as the target user consumes the portion of the target educational content.   
     
     
         15 . The system of  claim 10 , wherein the instructions, when executed by the one or more computer processors, cause the system to:
 identify second supplemental educational content related to a second portion of the target educational content, wherein the second portion and the portion of the target educational content are related to a low-comprehension dedicated section; and   modify the educational content interface to add the second supplemental educational content for display.   
     
     
         16 . A non-transitory computer-readable medium comprising stored instructions that when executed by one or more processors of one or more computing devices, cause the one or more computing devices to:
 access historical consumption information describing historical educational content consumption behaviors of a set of historical users and historical comprehension information comprising a historical measure of comprehension of historical educational content by the set of historical users;   generate a training dataset based on the accessed historical consumption information and the historical comprehension information;   train a machine-learned model configured to predict a measure of comprehension for a user consuming educational content based on behaviors of the user as the user consumes the education content;   apply the machine-learned model to behaviors of a target user consuming target educational content to determine a target measure of comprehension for each of a plurality of portions of the target educational content; and   for a portion of the target educational content corresponding to a below-threshold target measure of comprehension:
 identify one or more characteristics of the portion of the target educational content; 
 apply a content identification model to the identified characteristics of the portion of the target educational content to identify supplemental educational content related to the portion of the target educational content; and 
 generate an educational content interface to present the supplemental educational content to the target user. 
   
     
     
         17 . The non-transitory computer-readable medium of  claim 16 , wherein the educational content consumption behaviors include one or more of reading rates, pause time, number of re-read times, delays in content consumption, highlighting, types of highlighting, highlight coverage, time of day, switch outs, and switch duration. 
     
     
         18 . The non-transitory computer-readable medium of  claim 16 , wherein the measure of comprehension includes one or more of test scores, user evaluations of comprehension, types of highlighting, post-comprehension quiz results. 
     
     
         19 . The non-transitory computer-readable medium of  claim 16 , wherein the instructions to apply the machine-learned model to determine a measure of comprehension comprise:
 determining the measure of comprehension in real time as the target user consumes the portion of the target educational content.   
     
     
         20 . The non-transitory computer-readable medium of  claim 16 , wherein the instructions that when executed by one or more processors of one or more computing devices, cause the one or more computing devices to:
 identify second supplemental educational content related to a second portion of the target educational content, wherein the second portion and the portion of the target educational content are related to a low-comprehension dedicated section; and   modify the educational content interface to add the second supplemental educational content for display.