Tag recommendations using topic modelling (unsupervised machine learning)

[brettkromkamp]

Full description to be provided.

Relevant URLs:

• Topic analysis
• Topic modeling with Scikit Learn
• Latent Dirichlet Allocation for topic modelling explained: Algorithm and Python Scikit-Learn implementation

[navigator8]

Hi Brett, I am not sure if this relevant or note. Two products I use have the following features:

• Mindlib.de - when I create a note / topic/ node - it will try and search for the topic on Google Knowledge Graph / Wikidata and it will return matches and you can select to quickly create the topic.
• Hypernote.io - performs as above but goes a step further - it will return possible nodes /topics that are associated with your topic and if you select - it will automatically create associations.

[ansh-lehri]

Hey Brett, I am not able to understand what issue have you raised in this repository. Please explain it a bit as I am interested in contributing to this.

[brettkromkamp]

Hi @ansh-lehri :)

In summary, I would like to analyse the text of a topic (technically, an occurrence of type "text") using a topic modelling algorithm to determine (statistical) relevant keywords and subsequently generate tag associations (associations between the topic in question and the accompanying tag topics).

To understand associative tagging Contextualise it might make sense to take a quick look at the accompanying source:

• https://github.com/brettkromkamp/topic-db/blob/master/topicdb/core/store/topicstore.py#L794
• https://github.com/brettkromkamp/contextualise/blob/master/contextualise/tag.py#L24

Specifically, if you take at the first code reference you can see that a topic is created of type "tag" followed by two associations of type "categorization" but with different member roles.

That being said, the topic map side-of-things is trivial. The critical and more complicated part is the topic modelling side of things with the algorithms mentioned at the beginning of this issue.

One final note, don't confuse "topic modelling" with "topic map modelling" :) Quite different things. The former is a Natural Language Processing (NLP) technique that allows us to automatically extract meaning from texts by identifying recurrent themes or topics; the latter is semantic modelling based on the topic maps paradigm (https://ontopia.net/topicmaps/materials/tao.html).

[ansh-lehri]

Hey @brettkromkamp, sorry for delayed reply but I read the links you provided and I am interested in solving this issue to the best of my ability. Please assign me this issue.

[brettkromkamp]

@ansh-lehri Alright, I'll assign you the issue. I'm excited to have you work on this.