Using the GPT-2-simple algorithm google colab notebook https://colab.research.google.com/drive/1VLG8e7YSEwypxU-noRNhsv5dW4NfTGce I trained a bot on 503,547 TFTV posts (I didn't realize when scraping but this probably would have only taken threads that from the forums that are shown to first time visitors, so no dumpster, no LFT/LFP.) I also excluded TFTV CHALLENGE, stream highlights and frag clips thread.
Because I am bad at software design, the actual output from the scraping was formatted very awkward, which translates
into the generated text. The most egregious thing you will see is a lot of repeat text. This is because I thought it would be pretty cool to keep quotes from the original posts included to see what happened. It was not cool. There's still some very readable and entertaining messages, though.
Links: First output, second output .
Everything within one set of equals signs is generated by the algo as one "page" of the entire book.
If I were to retrain, I would run a script to take out all the quotes from the original text.
If I were to rescrape the entire site, I would try and get every possible post, and include all html tags but quote content (I removed that in the last scrape). I don't intend to rescrape though. I might do a tiny scrape+train with griff posts or dumpster posts, though.
Here are some good examples:
In the past few months, I've been having a lot of issues with the game, especially the beta. I'm getting into a lot of issues with the game, including the 'b-mode' bug, which has been bugged in the past, with the new maps, and the changed meta. I've been having these problems for a while now, with the way things work.I've been having issues with the time I've spent playing the game. I'm finding it very frustrating. I just don't know where to start. I'm trying to figure out what's wrong, and what I'm going to do to fix it.<br>I've been working around the issues with the new maps, and the new maps. I've been having fun and getting feedback, but by the end of the week, I'm going to have a lot of work done. And my team has been working around the issues.I guess I'll have to get some extra sleep now that I'm out of the way, but at least I'm feeling good!
<q><q>The fact that you're still raging about this is pathetic.<br><br><br>I think you're right, I'm sorry, I thought that was pretty good. I apologize if I was dumb, it's just just the way things are. Also, as a med, I don't think that's funny.<br><br>I don't think that's funny, I think it's just a bit of a joke. I want to be the first in line to get the meds. I won't be doing it, I'm not of the mindset to do it. I just think that I'm just a bit of a bit of a bit of a bit of a bit of a bit of a bit of a bit of a bit.</q>Sorry, I really don't think you understand, and I apologize if I was dumb. I was thinking that if I was in the right place, I'd be getting the meds, but I was just doing my job and not trying to hide it. I'm not going to tell you that I'm not going to just do it. I'll be making all of my meds publicly available.
(if it looks like it is having a conversation with itself, it is. That is it's recognition of conversations within the actual posts I scraped, both quotes included in singular posts and when posts are layed one after the other. Unfortunately, it was not able to intuitively format its arguments within quote tags, however this next one captures the quotes quite nicely so I've replaced the <q> tags with quote bbs:
====================
<br> best parts of tf2
<br> http://i.imgur.com/uH4nYsX.jpg
<br> http://twitch.tv/esports_bomber
<br> http://twitch.tv/teammate_tftv
<br> I'm glad that this guy is back, he's been a constant stream of shit for the past month. He's one of the best and most consistent TF2 players, and I'm glad that he's finally back.
<br>I'm glad that this guy is back, he's been a constant stream of shit for the past month. He's one of the best and most consistent TF2 players, and I'm glad that he's finally back.You see, he's just a nice guy, one of the few people I've spoken to who gives a shit about tf2 and works really hard to improve it.He's also a pretty good person as well, and you
====================
<br> <q>I think this is just the beginning of the way to the peace process.</q>I'm really happy that there's a group of people who genuinely care about the situation in Syria and want to join the peace process. I'm also pretty disappointed that the Syrian regime is not on the same level as our own parties, but I hope that the more we have to work on the more that we can reach a conclusion. <br> The big thing that should be done is to get the peace process started, because we have to have a political process, or at least one that has a legitimate process in place. If you are ready to join the process, then just pick up a few friends and get out of here. I would really like to see the peace process from a political point of view, but there's also the fact that there are now a lot of people who think that the peace process is a waste of time.The list of people who really want to join the peace process is pretty long. If you are interested in participating, you can check out the U.S. and EU forums. Then you can do your research at home.]
ok im not gonna keep editing this with snippets because then who would need to read the actual output BUT I am having a coughing fit from laughing at this so:
Holy shit. I can't believe I'm wearing a sweatshirt for the first time ever. I feel like a ******* idiot for not wearing a sweatshirt. How can you be stupid when you have a backpack full of tf2 gear? I felt like a second-rate idiot for wearing a sweatshirt.Well, because TF2 is so fucking huge at this point. The only people who are really going to be wearing a sweatshirt are people who have played for a long time.