Impressive, thank you for sharing! Initially I struggled with scraping due to cookies configuration, after some research I found Google News API which worked like a charm with your code. Thanks again!
Amazing video man. I watched couple of them from your channel already and you really know how to transfer your knowledge to viewers. Keep up the good work!
What a wonderful video!!! This was exactly what I was looking for. I do wonder if running the sentiment analysis on the summary was the best approach...perhaps running it on the full article may have been better? Regards from Orange County, California USA!!
I'm liking this longer style, tomorrow I'll be able to follow a tutorial like this a lot easier! Thanks!! (Did you do any analysis on whether the positive or negative sentiment actually leads to an increase or decrease in price? It may be an interesting add-on!)
Agreed, literally as I posted it I thought it might be useful to have article timestamps and matching prices at those dates! Might add it onto a future video!
@@guyincognito1985 ohhhh man, I didn't even realise it was a thing until one day I saw I had this random timestamp 🤣. I'm going a little crazy with them now! My OCD has definitely been triggered though because they're not all equal sizes. 😅
Thanks. I enjoyed watching this longer video. It would be interesting to use BeautifulSoup to scrape the web page title (whatever is in h1/title tag) and compare that to what the NLP summarization came back with. Thanks for making these videos!
Wow. Mega interesting material and such an impressive way of communicating. You are a truly gifted teacher Nick. I’m so glad I found your channel! 👍 subscribed and plan to go through all you videos.
@@NicholasRenotte you’re very welcome Nick. Btw. Is there a continuation of this tutorial? We’re summarized the news, now, how to apply it to stock market time series?
Man the work you put in for every single one of your videos is just 🔥🔥🔥 extraordinary man! I love your videos! Thank you so much for making such detailed videos! Your videos stand out💯 More power to you!💫
Hey man...Again this was a great video. In the future can you make a video on how to train object detection models on Colab or something similar/the file structure on Drive....It becomes very time consuming when we don't have GPU enabled systems. A tutorial on that would be very helpful.
Awesome, many thanks Nicholas! I enjoyed your teaching style a lot! However, I do have an issue when running the pipeline ("no model was supplied..."), leading to a RunTimeError. By any chance, do you know how to fix the issue so that the sentiment analysis works?
Hi Nick, firstly I'd like to congratulate you for putting this up. I've started using it and it's great. I'm wondering if can be used as a more multi purpose search engine to allow us to scrape the internet for other topics? I've installed the google app (which works fine finding urls howeverk, it won't pull the data and summarise. I'm thinking its because soe sites are blocking from scraping their content. Any thoughts? Thanks again, Peter
Hey amazing project man!! I was wondering if one could deploy this project on the web, that would make it a complete project right? What's your say on this?
Hi Nicholas, this is such a great video and is almost exactly what I needed. I wanted to do some tweaking by finding the daily sentiment analysis from January until, inserting the new code around the 35 mins mark of the video... unless I automate it, it would make a very tedious task but in order to set a custom date range, I would need to keep the whole of the google news link? Do you think this would be a problem? If so then do you have any suggestions? Thank you !
Heya @Josh, yeah it's a little tricky, you could extract the date from the articles and try that. There are two other options I might suggest: 1. Scrape every day and store your results with an assigned timestamp 2. Use reddit or twitter data and leverage the native timestamps
@@NicholasRenotte Hey Nicholas, so I think I've figured out a way around it (in theory ), I just noticed though that the list of atags I'm getting are just the "policies", "maps", "accounts" type of links and not the actual news links im looking for. Could this be due to some sort of permission issue? i did try the cookies solution you mentioned in another comment but that hasnt worked for me.
Anytime! I used to a while ago but not anymore unfortunately. I'm thinking of getting back into it though if people are interested in seeing how it's done!
Hi, before I start, the video is great. I learned a lot of new concepts and these are great. but when I try to retrieve all the links of the Google page with for example, the articles about BTC, I can't access the links. I only get the footer links and the page link. Would you have a solution? Thanks in advance.
Hi, Great Video! Tell me something: I am building the tennis analyzer system that I told you (it uses object detection and pose estimation) and I would like to know if I should make an web app or a mobile app? What would be more efficient in terms of cost? Thanks
Web app is wayyyyy easier and less effort in terms of deployment. You can always write your code using React-Native which allows you to code native apps but using the same language, Javascript.
given that the sentiment analysis module is not as good as the text summarizer model here, should we use the unsummarized text for generating the sentiment
If I use BeautifulSoup to get all the Yahoo Finance Links, the returned soup only contains the google policy stuff that is later filtered out. "find("a")" does not work for me. Any ideas? If I inspect the google search I can see the href links which are wrapped in ""... however, BeautifulSoup seems to not fetch them.
As soon as I run the source code from GitHub, I get the following message in the output file: It’s an error that’s hard to diagnose because it’s a malformed request.
About that 👀, I didn't really pay attention here as we searched through Google then went to the site from there. Bad I know but it demonstrates what's possible.
scores = {ticker:sentiment(summaries[ticker]) for ticker in monitored_tickers} is it that this command takes long time to get executed? Just wanna confirm. Thanks in advance
Wonderful tutorial. But I am getting error "AttributeError: 'NoneType' object has no attribute 'encode' at "input_ids = tokenizer.encode(ARTICLE, return_tensors='pt')" section when I tried it. How can ı fix it?
Hi. I'm working on a similar project. Your videos are very very very helpful. Just want to say thanks first. I have a question, my project requires a 30-day stock news data. How do I scrape the second page of the google yahoo finance search result? Do i run it again and store the scraped data under (search_url 2) , similar to 4.1 in the Notebook. Hope it make sense to you? Thank you very very much Nicholas
Hi Nicholas! I was trying to follow your steps, but whenever I try to run the following code the kernel dies: input_ids = tokenizer.encode(ARTICLE, return_tensors='pt') output = model.generate(input_ids, max_length=55, num_beams=5, early_stopping=True) summary = tokenizer.decode(output[0], skip_special_tokens=True) Is this because of my PC specs? I'm using an M1 mac with 8GB of ram. How can I solve this issue? Thanks in advance!
load model just before summarize method, otherwise it will take up space in ram at the same time as you load articles. make sure also to close some chrome tabs since chrome is known to use to much ram.
Hi im working on google colab and it indicates me that tokenizer is a NoneType so then when apply "" tokenizer.encode """ it goes to Error, is there something I have to do by working with google Colab??Packages are correctly installed
getting an error right at the beginning when defining tokenizer variable... This is the error 'TypeError Traceback (most recent call last) in 1 model_name = "human-centered-summarization/financial-summarization-pegasus" ----> 2 tokenizer = PegasusTokenizer.from_pretrained(model_name) 3 model = PegasusForConditionalGeneration.from_pretrained(model_name) TypeError: 'NoneType' object is not callable Can you please help?? I cloned your repository with the notebook and tried to run it
