>>7544 Some progress, good. I think at the end it needs to be something like the most common words and phrases, which aren't the most common in any random text. But the scope should probably be on the specific posting then. So we might at least get something like [>>1234, 'Windows', 'chatbot', 'AIML', 'run.py'] - Though, looking into the specific postings might be the second step instead, after having a list of keywords.

>>7548 Well, one thing that would be pretty easy to do for me now that I've done most of the basic groundwork for processing /robowaifu/ text is to produce lists of post links for all human-specified keywords. Right now there's no intelligence involved, it's just a mechanical turk to chew through all the words. But make me a list of the top 100 words you're most interested in seeing, and I can give you a list of 100 word's collection of posts containing those words. It would be a good start for now. I'll keep working on better filters and consolidation of terms, you be thinking about keywords to use. Deal?

>>7550 > But make me a list of the top 100 words you're most interested in seeing The whole idea here is, that the software should find that list. That's what I'm talking about. If you take out the most common words in English, then you should have the topic of a posting. That should work as a simple solution. The list is not the input, it's what the software should find... Also, I'm currently working on chatbot stuff and feel like I'm drowning in tasks and todo lists. Though I have to admit this task is something we need exactly for that as well. Finding the topic of a comment or conversation. Looking into it briefly, probably overkill for the current task: https://towardsdatascience.com/nlp-extracting-the-main-topics-from-your-dataset-using-lda-in-minutes-21486f5aa925 https://www.kdnuggets.com/2019/11/understanding-nlp-topic-modeling-part-1.html

>>7555 I see, thanks for helping me out clarifying that. I apologize, I was being inconsiderate of your time. Haha, you just keep doing what you're already doing and we'll take care of this little project Anon! Thank you for everything you're doing here. :^) Thanks also for the links, I'll spend time researching this idea.

>>7201 >servo motor pins >>7575

>>7555 Just a quick update. I've finished working out the data structure requirements for the JSON tool, and have a rudimentary search function working. It's typical systems-type code, but it's pretty quick: once everythings parsed (about 5 seconds on my 2-core potato box) it's currently searching the entire set of words on /robowaifu/ for a given term in ~40-50us. So it's coming along in that way to be efficient. Here's a cap of searching for thingiverse on /robowaifu/ : > I know it's not entirely what we need yet, but at least the fundamental database engineering work is together, roughly speaking.

>>7578 Thanks kindly Anon.

>>7201 >optical speed encoder >>973

>>7202 >Tamiya 4WD tires >>973

>>7183 >basic bots for learning electronics >>969

>>7599 >>7600 >>7601 Index entries made.

>>7601 >>7186

>>7188 >GPT-2 (based chatbot) >>2762

>>7189 >hair actuator (aho-hair) to express emotion >>4560

>>7198 >Product key memory layers >>6618

>>7198 >Python (programming language), NLTK toolkit for NLP >>7837 >>7196 >NLP introduction >>7837 >>7195 >Memory in Neural Networks >>6618 >>7187 >Forgetting in Neural Networks >>6618 >>7204 Value Systems in an AI >>7855

>>7852 >>7853 >>7854 >>7856 I feel that most of these should at the least be cross-linked under the AI index entry, if not proper sub-categories of it there. Opinions on this, Anon? Good idea to just consolidate the requests together btw.

>>7860 Okay, maybe, or both. If one searches for one specific topic he should find it by letter. That was the idea, or not? Also consider how long the entry under AI could get. I thought that even threads like the one on GPT2/3 should also be linked under G like GPT, but could also be mentioned in the entry for AI.

>>7862 >That was the idea, or not? Yes, multi-indexing is both commonplace in these sorts of encyclopedic indexes, and definitely necessary here. The challenge for us as editors is properly choosing the primary/cross-link distinctions. Brevity is a premium here, as you say >Also consider how long the entry under AI could get. If we do our jobs right here, this will likely become quite a comprehensive list IMO. Fortunately we can also just edit things afterwards.

>>7852 >>7853 >>7854 >>7856 OK.

waifusearch> PWM OR Pulse-width modulation THREAD SUBJECT POST LINK Single board computers & micro-c >>3736 pulse width modulation Micromouse - Minimum Viable Supe >>3574 pwm " >>3589 " " >>3608 " " >>3627 " " >>3640 " " >>3641 " Robowaifu Systems Engineering >>4640 " Embedded Programming Group Learn >>367 pulse width modulation " >>3762 pwm HAPPY BIRTHDAY /ROBOWAIFU/ >>7977 " " >>7984 " Proto1 All Parts >>6729 " ' pwm | pulse width modulation ' = 13 results https://en.wikipedia.org/wiki/Pulse-width_modulation >=== -edit subject -add alternate spelling search -add article link

>>7256 >tags = ['sophie', This is actually a good example case of why we need better refinement of tags, for more specificity. For example, here's the current waifusearch for the term 'sophie': waifusearch> Sophie THREAD SUBJECT POST LINK Robo Face Development >>5042 sophie C++ General >>7616 " R&D General >>5038 " Robot Wife Programming >>7821 " Waifus in society >>4206 " your file is too big onii-chan >>7789 " Robot skeletons and armatures >>7634 " Actuators for waifu movement! >>7086 " " >>7106 " " >>7217 " " >>7220 " Robowaifu Design Software >>5188 " " >>8076 " " >>8095 " HAPPY BIRTHDAY /ROBOWAIFU/ >>7224 " " >>7981 " My Waifu >>2496 " " >>2497 " " >>2498 " " >>2500 " " >>2578 " " >>2580 " " >>2581 " /robowaifu/ Embassy Thread >>8135 " Wheelchair Waifus >>4297 " /robowaifu/meta-2: Electric Boog >>7765 " Sophie Development Continues >>4146 " " >>4154 " " >>4166 " " >>4193 " " >>4197 " " >>4203 " " >>4204 " " >>4216 " Sophie Bot STL Files Uploaded >>4198 " " >>4202 " Hand Development >>4870 " " >>4876 " Work on my Elfdroid Sophie >>4787 " " >>4788 " " >>4858 " " >>4882 " " >>5000 " " >>5159 " " >>5302 " " >>5431 " " >>6049 " " >>6051 " " >>6319 " " >>6322 " " >>6659 " " >>6662 " " >>6713 " " >>6976 " " >>7264 " " >>7502 " " >>7512 " " >>7619 " " >>7621 " " >>7627 " " >>7698 " " >>7945 " " >>7948 " " >>7972 " " >>7993 " " >>8019 " " >>8025 " " >>8032 " " >>8086 " " >>8087 " " >>8094 " " >>8182 " " >>8268 " " >>8269 " " >>8288 " " >>8290 " " >>8308 " " >>8356 " " >>8381 " " >>8392 " " >>8410 " " >>8573 " " >>8598 " Haute Sepplesberry Cuisine TBH >>8027 " Proto1 All Parts >>6326 " " >>6329 " new user and seeing how it works >>7726 " ' sophie ' = 87 results 70 results. While that's comprehensive enough (potentially exhaustive even), it's not bringing a lot to the table for the searcher just trying to quickly find that one important bit of information on Elfdroid Sophie. Her hand designs, for example. And for the newcomer...well, I'd say that result set is rather too large TBH. ATM, I can't really think of a more effective means to manage tagging with sufficient refinement than doing so by hand. Waifusearch is a reality now, so it could help us all along in that process. But as far as doing it otherwise, we still have a ways to go yet in achieving an AI that has enough semantic and contextual awareness to do it all automatically for us. However, this is obviously the goal in the end -- and not just for this specific need either. It strikes me now that this little project of automatically tagging all of /robowaifu/ might just prove to be an effective means to finally achieving the long sought-after goal of a contextually-aware AI. At the least it's a sufficiently narrowly-defined goal to have more than just the slimmest chance of success at it (anytime Real Soon Now™). At least that's my take on the matter. I'd be interested to hear anyone else's views on this important AI topic. >=== -edit prose

>>7997 >Waifusearch is a reality now, so it could help us all along in that process Yes, we'll still need humans for the tags. But when we have more tags, the search could break down huge search results by searching again in those results and break them into sections. >contextually-aware AI Waifusearch is certainly going to be helpful and I like the result so far,. it's quite amazing. Sophie is a big and complex topic, having links to a thread with subjects like "face" or "hand development" creates more context.

>>8000 >But when we have more tags, the search could break down huge search results by searching again in those results and break them into sections. Yes, we can easily do just that Anon. I'll put it on the TODO: listings. >checked digits :^) Happy New Year

>>8000 Just in case you have a RaspberryPi machine Anon, we have waifusearch working there now too. >>8026

waifusearch> SolveSpace THREAD SUBJECT POST LINK Robo Face Development >>4600 solvespace " >>4607 " Robot skeletons and armatures >>8093 " " >>8596 " Actuators for waifu movement! >>8170 " Robowaifu Design Software >>4390 " " >>4423 " " >>4795 " " >>5188 " " >>8076 " " >>8079 " Prototypes and failures >>8034 " " >>8372 " Work on my Elfdroid Sophie >>6342 " " >>6344 " " >>8379 " " >>8380 " ' solvespace ' = 17 results https://github.com/solvespace/solvespace https://en.wikipedia.org/wiki/SolveSpace >=== -add external links

Alright, for the time being (and until we come up with a more sophisticated solution), I think we'll just have to settle for making waifusearch posts ITT, and then we can link them into the index system above. So please get the tool working on your own systems Anons, and post important searches ITT. BTW, you might think about different spellings, etc., and add any valuable external links into your post as well. For example, PWM >>7982 , or RC >>7983 . Cheers. >software building instructions >>7933 >RaspberryPi building instructions >>8026 https://files.catbox.moe/tmu6qr.7z 3e3173220fd46b093771f98c37760da8eb0a69be3b32a4ed9b20388d81a226de waifusearch-0.1f.tar.xz >=== -add crosslinks -add D/L link & sha256sum

>>8090 Heh, not much use atm. I'll redo this search here after the Papercraft waifu thread is repopulated.

waifusearch> linear algebra THREAD SUBJECT POST LINK AI Software >>5813 linear algebra Robowaifu Simulator >>1899 " " >>2017 " " >>2044 " ROBOWAIFU U >>2970 " General Robotics/A.I. news and c >>6873 " /robowaifu/meta-2: Electric Boog >>6934 " " >>6962 " " >>6964 " Sophie Development Continues >>4216 " Beginners guide to AI, ML, DL. >>6567 " " >>6570 " " >>6573 " F = ma >>7787 " ' linear algebra ' = 14 results http://immersivemath.com/ila/tableofcontents.html

>>7300 Actually after a month or two using what's now called waifusearch, I've come to entirely agree with you. I'm no longer including the JSON for this thread or for the old archives thread >>273 in the source files, because of redundant (and useless) results showing up. Heh, I should've taken your advice earlier Anon. :^)

waifusearch> Neuromorphic THREAD SUBJECT POST LINK AI, chatbots, and waifus >>4061 neuromorphic " >>4062 " AI Software >>1227 " Robot Vision General >>4786 " ROBOWAIFU U >>2526 " " >>2558 " " >>3011 " ITT: Anons derail the board into >>2081 " " >>2085 " " >>2086 " " >>2087 " Biological synthetic brains for >>2204 " /robowaifu/meta-2: Electric Boog >>3937 " " >>4045 " Why consider alternative CPU arc >>8321 " ' neuromorphic ' = 15 results https://en.wikipedia.org/wiki/Neuromorphic_engineering

Can this get stickied?

>>11384 Yes, it usually is ATM, unless an 'event' thread is ongoing. R/N that's the 'Delete' thread (the idea is for everyone to weigh in on their view of anon's request to do so). This should be restickied again after Sunday or so.