How to Perform Supervised Topic Modelling (Supervised Latent Dirichlet Allocation, sLDA) in R

Подписаться 4 тыс.

Просмотров 3,8 тыс.

50% 1

In this video, we learn about supervised topic modelling (supervised Latent Dirichlet Allocation, often abbreviated to sLDA). We first begin with an overview of what topic modelling (LDA) is, and then learn how supervised topic modelling (sLDA) differs. We then walk through an example of performing sLDA in R using the lda package, with two approaches:
1) predicting a continuous response variable using only text and
2) predicting a continuous response variable using text and other explanatory variables.
This video is part 2 of a 2-part video collaboration with fellow RU-vidr ‪@SamerHijjazi‬, who also creates R programming tutorials! You can check out Samer’s channel (and subscribe to him!) here: / samerhijjazi
The dataset used in this video was generated by Samer, who web scraped the “Rate My Professors” website. You can see how this dataset was scraped by watching Samer’s tutorial (part 1 of the 2-part video series) here: • How to Web Scrape Rate...
This video is partially based on section 7.5 of the book “Text Mining for Social and Behavioral Research Using R”, available online for free here: books.psychstat.org/textminin...
All of the code from this tutorial (and from all my RU-vid videos!) is available on my GitHub. You can find the code for this video here: github.com/melissavanbussel/Y...
0:00 - What is supervised topic modelling?
3:32 - Preparing the data
18:18 - Fitting the sLDA model
28:08 - Adding an explanatory variable to the model
34:40 - Performing prediction on unseen data
Want to connect?
Twitter: / melvanbussel
LinkedIn: / melissavanbussel
Website: www.melissavanbussel.com/
GitHub: github.com/melissavanbussel
*Please note: it’s helpful for me if you send me a message on LinkedIn when you add me letting me know you’re adding me because of RU-vid, so that I can keep track of how I know each person in my network😊

Опубликовано:

7 авг 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 21

@KevinVan-pu7cr 8 дней назад

wonderful step-by-step tutorial, thanks!

@Y45HV1N 24 дня назад

I hava a slight attention deficit disorder, I've been trying to learn topic modelling for years with youtube videos, but I'm never able to follow, but I'm able to follow this video perfectly! Thanks so much!!

@ahmed007Jaber Месяц назад

thank you so much Melissa. this is superhelpful. while i wouldn't make models directly it brought me closer to understanding this and learnt couple of tricks for text mining. please keep up the good work and share more videos like this.

@kenandagci 3 месяца назад

thank you for sharing!

@farreledgarrafi1118 2 месяца назад

Thank you so much for the video you made. Your video really helped me understand the sLDA more deeply. But there's something I want to ask, how can the rating predictions that are made from the regression model know, which student and for which professor the predictions are made? And how do I make the prediction of a particular student to a particular professor? Is it possible to do that? Thank you.

@yeokcmark 3 месяца назад

Thanks for sharing. I learnt a lot from your video. I believe the above can now be done using the STM package as well?

@gustavoantoniobrugesmorale1881 Год назад

Thank you!!! Your explanation about topic modeling is clear. I am going to use this in my research

@ggnot2 Год назад

Glad it was clear and helpful for your research!

@AngelFelizF Год назад

Thanks for sharing, this is a great video

@yasamanrajabli Месяц назад

Thank you for this video, Where can I find the CVS file you have used?

@prof.gobindaroy Год назад

Awesome content. Thank you!!!

@ggnot2 Год назад

Glad you enjoyed! Thanks for the kind words.

@tobip7631 Год назад

What a great video, thanks! I do not have a set of documents however, only a csv containing reviews from websites. How would my approach to programming this in R differ?

@ggnot2 Год назад

Hi Tobi, thanks for your question - it's a great one! In this situation, it's just a semantics/wording difference, but you can do the same thing. In this case, you would consider each row of the csv to be a "document". We're just using the word "document" to refer to one piece/one unit of text, which in your case would be one review. They don't necessarily have to be documents!

@tobip7631 Год назад

@@ggnot2 Thanks, that clears it up! I was wondering for supervised topic modeling methods: Supposedly I already knew the 4 topics that I am trying to "categorize" all the reviews into and then do the same analysis with the ratings as you did: Would that approach be different? If yes, do you have any idea what and how to implement this in R? :)

@subtleT Год назад

This video was so informative. It really helped me understand a bit more about what's going on under the hood of some of these package functions. I'm working on a project in which I want to compare topic models across corpora. I was thinking I might be able to do this somehow using sLDA. Do you have any thoughts/recommendations about this?

@dynahmhyte 6 месяцев назад

What can I do with "could not find function slda.em"?

@tobip7631 Год назад

Hi! Suppose I only wanted to use LDA and not sLDA (because I am only interesting in the topics and the visualization of them). How would I need to change the code?

@andrea-mj9ce Год назад

Is it any different to just do LDA then use the topics as variables? I only watch the first 2 minutes of the video for your concern

@ggnot2 Год назад

The basic idea behind sLDA is using the results from LDA to create explanatory variables, yes :) This tutorial walks through how to do that and provides a conceptual understanding of what both LDA and sLDA are.