Hi - as a teacher I have a lot of documents that has students reports, behaviour, interests etc.

Hi!

Below I listed some possibilities:

- Knowledge bases: allows you to find candidate answers looking for passages and paragraphs in a knowledge base that you can create just by uploading a set of documents in common formats like pdf, docx, txt, etc. (students reports, tests, etc). This module is totally unsupervised, that means that there is no effort categorizing/labeling information.

- Text classification: allows you to map any text data (a summary of the student profile for example) that you (with your expertise) consider relevant to correlate with some behavior your want to study. This could be helpful to try to infer any variable you want to measure. In this case, this feature requires a manual effort to tag your students with the ground truth and categorize them based on the variable you want to measure (for instance, if they were able to finish school or not, by observing relevant information only from the first couple of years or the initial student profile).

- In educational and e-learning platforms, Cogniflow can be used for example to moderate channels and conversations, recognize and block potential offensive message or analyze students' sentiment/emotions (you can take a look at the already trained models that we have in the public experiments view), or use our upcoming feature of semantic search to search for potential copy in tests, duplicated contents, etc.

- Other use cases, to push student's experience to the next level could be to combine our transcription model with the question-answering system by creating knowledge bases using the transcriptions you can get from videos, virtual zoom classes, etc, so students can ask questions directly to the content and Cogniflow find probable answers if they appear in the content.

- Other interesting feature is the OCR that you could use to extract text from images in scanned documents, or photos of students' work.

- And I can imagine many possibilities with the upcoming semantic search feature that is planned for next year. By using that functionality you could build a text based profile of your students using well studied relevant data that you can use to correlate some kind of behavior you wish to infer. Later, once the search space is created by Cogniflow with these reference profiles, you can use this search space (formally a vector space model or VSM) to search the most similar profiles for a given new one. Once you have these matches (which you know their characteristics, like if they approved the course, if they were conflictive students, or any important things/variable you want to measure) you can take this info and think the new profile will probably have many things in common, thus you can take action. You can think in this functionality like a recommender system that you can use to find "similar students" to give suggestions, recommend materials for study, or any other thing that already worked fine for your reference profiles in the past.

Hope this can be helpful!