I'm new to using Notion and am trying to wrap my head around how to best implement it for this use case. My wife and I attend a couple of music festivals plus a handful of concerts a year. A lot of the big festivals have a staggered line-up release schedule that typically goes Overall Lineup, then weeks later a Day by Day Lineup, then much closer to the date of the festival an hour by hour schedule that shows what stage the artist will be at. The majority of the time, we don't know a lot of the artists and use the lineups to discover new (to us) music. But when you are talking about a large number of new to us artists, we have to take some notes to keep track of them all! A lot of artists are on "the festival circuit", where they pop up at multiple festivals during their touring. Since we travel to a couple of festivals, we frequently run into situations where the artist will be at the festival as well as touring in our city within a month or two.
In the past, we've used Trello to kind of organize all of this. In trello, we've made a board for the specific festival that has columns per day (Friday, Saturday, Sunday) with each card being the artist. Then we used some labels for which stage they'll be at and a custom field for the timeslot. In the description field, I'll usually put in Spotify data like number of followers and genres. The comments section of the cards is just us putting in our thoughts about the artists or their songs. Finally, we used a custom field for our 1-5 star "rating" of the artist that we use for scheduling conflicts. This system works well enough for each festival as a standalone, but it isn't ideal for using across multiple festivals or multiple years.
I have some general ideas on how to do all this AND expand on it with Notion but being so inexperienced with it, I think I'll need some help with some of the concepts. It seems like I should have an "artist" database that has Name, Genre, Spotify Followers, our comments/thoughts about the band, and possibly a way to track which festival or venue we've seen them at (would this part be a related database?). Then have a "festival" database that has Stages, Dates, Timeslots. How would I tie those together? I have this idea of a dashboard that would show a calendar view similar to the festival schedule poster I've attached, but I could click each artist to pull up the artist database entry for them. At some later point when I'm more skilled in Notion, I figure I'd incorporate stuff like local to the festival weather and a trip packing list into the dashboard.
Please forgive me for any terminology mistakes and thanks in advance for the help!
https://preview.redd.it/b87qseth1t0d1.png?width=970&format=png&auto=webp&s=41d701ac7a4e4baadea0e782e2d026dbf11296d6

Hi everyone,
I was recently tasked with finding a solution to perform OCR on handwritten forms (timesheets in my case, but this could also be applied to other handwritten forms and handwritten surveys, for example). Since I didn't find a comprehensive guide prior to doing my own research, I thought it could be useful if I shared my results here.
This builds on a useful post here from last year that looked at the state of handwriting OCR in 2023, but had more of an emphasis on pure handwriting transcription rather than handwritten forms and documents in this case.
I hope this summary is helpful for anybody else on a similar search. YMMV, and all of these services offer free trials to test your own documents against. If I have missed any, please add them to the comments, or let me know and I can add them.

The test.

I took a sample image showing a basic timesheet with handwritten text. I ran this image through as many OCR services as I could find that claimed to offer handwriting to text OCR, and compared the results. I was looking not only for accuracy in transcribing the handwriting to text, but also the ability to extract the data in a structured form, either as JSON, or as a spreadsheet (CSV or Excel).

Quick summary.

The two best results came from Microsoft Azure Document AI, and Handwriting OCR. Both provided near-perfect transcriptions of the test subject, and extracted structured data too.
Overall, I preferred Handwriting OCR because of the better UI and direct export to Excel. Azure is a developer-focussed product so needs additional development to use it in production, but for a larger project I would also not hesitate to use it.

Notes.

Here's a list of the services I tried, and my notes as I went along. Most are online OCR services offering handwriting to text conversion. Also some large language models like GPT-4. I have also attached screenshots from some of these services to highlight what was good and bad.

Transkribus

Tested at: https://www.transkribus.org
https://preview.redd.it/mfwdkpblxs0d1.png?width=1444&format=png&auto=webp&s=688665be8959c16083f48b52a7202f5f7e0379fe
This is one of the most well-known services offering handwriting recognition, with a focus on historical documents. As tested, the handwriting recognition was surprisingly poor, with lots of mistakes and non-words. Transkribus does offer the possibility to train on a particular style of handwriting (maybe useful if you have a lots of documents from one source), but for processing lots of documents where each document has handwriting from a new source, it looked like the handwriting to text conversion was too error-prone. This made it a non-starter for me.

• Strengths:
  • May perform well for historical documents.
  • Offers a web UI.
  • Pricing seems reasonable.
• Weaknesses
  • Handwriting recognition was really bad out of the box.
  • Requires the language to be preset - does not appear to detect the language automatically.

Google Document AI

Tested at: https://cloud.google.com/document-ai?hl=en#demo
I expected this to be the best of all, considering Google's massive investment in AI. I tested it through the demo here. Although Google Document AI does offer extraction of table data, the results were not great, with too many transcription errors. Based on these results, I would need to carefully review each extracted table for structure and content - not great. Aside from the demo page, there is no prebuilt UI so this would need a developer to integrate into my workflow.

• Strengths
  • Inexpensive
  • Offers structured data extraction
• Weaknesses
  • API only, requires developer to create a UI to process and download results.
  • Table extraction was inaccurate.
  • Handwriting recognition was not perfect.

Microsoft Azure AI

Tested at: https://azure.microsoft.com/en-us/products/ai-services/ai-document-intelligence
Azure Document AI
Like Google, Azure is one of the leaders in AI and document automation. I expected good results and in this case I was not disappointed. Azure read all handwriting correctly, and provided a correctly-formatted table containing this data. The downside for me, similar to Google AI, is that there is no UI provided (besides a demo portal) so this needs to be built from scratch by a development team.

• Strengths
  • Inexpensive.
  • Accurate results.
  • Offers structured data extraction.
• Weaknesses
  • API only - development costs of building your own interface.
  • Not fast

Pen to Print

Tested at: https://www.pen-to-print.com/handwriting-to-text-online-ocr
Pen to Print - transcription is good but no structured data.
This is a popular iOS and Android app that also has a web app too. The handwriting recognition was good, but the output was not in a structured form so the results are of limited use for my purposes.

• Strengths
  • Simple UI.
  • Handwriting recognition was accurate.
• Weaknesses
  • Feels quite basic.
  • Did not extract structured data.

Handwriting OCR

Tested at https://www.handwritingocr.com
HandwritingOCR - the best results, including all the data I wanted from the form.
Offers handwriting OCR and table data extraction, so this looked promising. Results were outstanding in my test - the transcription was error-free, and a separate tab allowed me to view and download the table data in Excel format. This is all provided inside a web UI with API access available.

• Strengths
  • Excellent result - the best overall.
  • Extracts both key-value pairs and tables together.
  • Export directly to Excel.
  • Web UI was easy to use.
  • Inexpensive.
• Weaknesses
  • No JSON export.

Nanonets OCR

Tested at: https://nanonets.com
Nanonets - incomplete data and transcription errors.
Offers a whole range of OCR services, including handwriting to text. At first glance, it looked impressive and I had high expectations of a good result. In my test, Nanonets managed to extract the table but with numerous handwriting transcription mistakes, so the result was only partially useful.

• Strengths
  • Polished UI
• Weaknesses
  • Transcription mistakes.
  • Expensive ($0.30 per page).

Google Cloud Vision AI

Tested at: https://cloud.google.com/vision/docs/drag-and-drop
Not to be confused with Google Document AI, though the transcription was in fact better. This service does not offer table data extraction though.

• Strengths
  • Handwriting transcription was accurate.
  • Inexpensive.
• Weaknesses
  • Does not extract structured data.
  • API only.

ChatGPT

Tested at: https://chatgpt.com and via API
I tried OpenAI's GPT 4 with Vision through both the API and ChatGPT. I found it can deliver very impressive handwriting to text conversion, but it also suffered from hallucination, and it could not reliably extract structured data. I also had problems with latency and timeouts when calling it through the API.

• Strengths
  • Can be very accurate.
• Weaknesses
  • Quite expensive and slow.
  • Tends to invent (hallucinate) text.
  • Would not reliably extract data in structured form.

Claude AI

Tested at https://claude.ai
Similar to GPT-4, I tried both Claude Sonnet and Opus. The results were ok but suffered from the same problems as GPT-4 - hallucination, and the unreliability of data extraction. Opus is really expensive, too.

Google Gemini Pro

Tested at: https://gemini.google.com/
Much worse than GPT-4, not worthy of further consideration at this point.

https://preview.redd.it/f92a5hfqxs0d1.jpg?width=2148&format=pjpg&auto=webp&s=2fae7abd3afe4fe14058d0250a9a0f949164a307
Hey Reddit! Ever dreamt of turning your love for fantasy sports into a career? Well, buckle up because fantasy sports software development might be the perfect path for you. It's more than just creating apps; it's about building a platform that fuels the competitive spirit and brings the thrill of fantasy sports to life.
So, how do you join the ranks of these developer wizards? Here's a breakdown:
Essential Skills to Master:

• Programming Languages: Being comfortable with languages like Python, Java, or Swift is a must. These are the building blocks for crafting the core functionality of your app.
• Database Management: Fantasy sports rely heavily on player data and statistics. Knowing how to structure and manage databases efficiently is crucial.
• API Integration: Fantasy apps often integrate with external APIs to pull in real-time data on player performance. Understanding APIs will help you connect your app to these valuable resources.
• User Interface (UI) and User Experience (UX) Design: A user-friendly and visually appealing interface is key to keeping players engaged. Familiarity with UI/UX principles will help you design an app that's intuitive and enjoyable to use.
• Problem-Solving and Analytical Skills: Fantasy sports development involves tackling technical challenges and analyzing user behavior. Being a critical thinker and problem solver will come in handy.

Gearing Up for Success:

• Start with the Basics: Don't try to build the next fantasy sports behemoth overnight. Begin by creating a simple app with core functionalities like player drafting and scoring systems. This will help you solidify your development skills and gain valuable experience.
• Stay Updated on the Industry: The fantasy sports landscape is constantly evolving. Keep yourself informed about the latest trends, technologies, and regulations to stay ahead of the curve.
• Build a Portfolio: Showcase your skills by creating sample apps or contributing to open-source fantasy sports projects. This will demonstrate your capabilities to potential employers or investors.
• Network with the Community: Connect with other fantasy sports developers, join online forums, and attend industry events. Building connections will open doors to new opportunities and provide invaluable insights.

Bonus Tip:

• Be a Passionate Fan: A genuine love for the game will fuel your creativity and help you understand the needs and desires of fantasy sports enthusiasts.

What are your thoughts, Reddit fam? Do you have any other tips for aspiring fantasy sports developers? Share your experiences in the comments below!
P.S. If you're already a fantasy sports developer, let us know what you love most about your job!

Pretty wild anecdote about the lab being unwilling to test the raw milk samples for avian flu!
“EISNER: You're right, Ailsa, it would really help to know if and how much virus is out there in the raw milk supply being sold now to people. But testing doesn't seem to be happening regularly, and that data is really hard to come by. That's why we tried to find our own. But when I brought raw milk that I bought from those four Texas farms to one of the few labs authorized by the USDA to test milk for bird flu, the lab insisted on calling each of the four farms first for permission, though the USDA has confirmed the agency doesn't require permission from farms to perform the test.
None of the farms gave the lab permission to run the tests. They told the lab they were aware of what a non-negative result would do for their business. So the lab refused to test our samples. That means we weren't able to find out how much virus was in the raw milk being sold now in Texas, and probably it's very difficult for other members of the public to do that, too.”

image thumbnails are still a pain
This awesome RBY-style Cinccino sprite is made by SharkGuy01--check them out on DeviantArt!
In this series we will analyze how modern Pokémon would function if they existed back in RBY. Note that I will only cover efficiency in RBY OU specifically, as I don't have enough knowledge of RBY lower tier metas to say much with confidence about them. As for the rules when "translating" a mon to RBY, check out the first installment.
For the third entry of the series we will take a step back from tossing high-BST legendaries into the good ol' days and throw in a more innocent little critter. A rather unimpressive mon outside of the kawaii factor, Cinccino doesn't exactly scary many people in modern OU metagames, and when transporting it back to RBY it would lose many important moves like Tidy Up, Triple Axel, Rock Blast and even its signature Tail Slap. Will the coveted Normal type and some hidden gems in that movepool be enough to make it usable? Let's find out.

The Stats

Jumping out is the 115 base speed. This is excellent by RBY standards, outpacing Tauros and speed tying with Starmie. Alakazam and Jolteon are the only relevant things that threaten to outrun you and neither are known to be good at tanking physical hits. You also get a crit rate of roughly 22% which is nice to have.
95 Attack is solidly usable. It's not terribly high, but remember Tauros dominates the meta with only base 100. With STAB Hyper Beam under the belt, rest assured you can threaten some things.
The good parts end there though, as its bulk is... really not good. 75/60 physical bulk is barely better than Jolteon, and low enough that Body Slam into Hyper Beam from Tauros is a KO more often than not, pretty much invalidating your status as a potential switchin to it despite immunity to Body Slam paralysis. Rhydon Earthquake also always 2HKOs, and you often don't even take an Exeggutor Explosion from full HP. It's not as horrid as it is in later gens due to RBY bulk mechanics, but it's definitely far below what the rest of RBY has.
65 Special is also definitely not something to write home about. Its special attacks won't threaten many things unless they're x4 effective or something, and its special bulk isn't too great either. Alakazam's Psychic is a roll to 2HKO without any crits, for example. Granted, this special bulk is comparable to Tauros', so it's not like this is unheard-of levels of low, but it's a definite downside and quite the uphill battle to overcome.

The Typing

There's not much to say here. Normal has a case for best type in RBY. You lack exploitable weaknesses due to Fighting moves being nearly inexistent, so despite being on the frail side, you're almost never going to be OHKOd outright. In addition, you cannot be paralyzed by opposing Body Slam. This greatly improves your matchup into opposing Normals, as while you still take more actual damage from Body Slam than you'd like, you cannot be trolled by paralysis in a 1v1. Offensively, Normal very notably gives you STAB on Hyper Beam, meaning you have a powerful finisher that can threaten moderately chipped stuff to switch out. It also would give you STAB on Body Slam (the 'would' is an omen to the unfortunateness coming later).

The Movepool

Looking at the standard offensive movepool, we notice a big problem: Cinccino does not get Body Slam! This is a huge blow to its general offensive capabilities: in cases where Hyper Beam doesn't kill yet, you are now reliant on Double-Edge for damage, which is admittedly stronger but also has very annoying recoil damage that only exacerbates your subpar bulk. In addition, Double-Edge cannot threaten to paralyze switchins like Starmie and Rhydon, which complicates getting past them in the long run. You do have Hyper Beam, at least.
Your coverage is also pretty pathetic. Only having Thunderbolt and Thunder, coming off 65 Special, is poor. The only real target for these is Cloyster, which you threaten a 2HKO on with Thunder. This is pretty notable since Cloyster otherwise looks like it walls you fairly hard (especially since you cannot even threaten to paralyze it). Other potential targets will be less impressed, like Starmie, who is only 3HKOd by both Thunder and TBolt.
That being said, our little chinchilla rat has three notable moves not mentioned yet that might well elevate it beyond "shitty Tauros with 5 more speed".
The first of these three is TWave. TWave somewhat alleviates the lack of Body Slam paralysis by letting Cinccino threaten to just manually paralyze stuff. Being rather frail and running into 4MSS as we'll see later makes this not the hugest deal, but it's still a very handy tool to have access to and pull out once in a while.
The second hidden gem is Sing. Sing immediately gives Cinccino a massive niche: since it outpaces Gengar, this makes Cinccino the fastest user of a sleep move in RBY. Of course Sing isn't exactly accurate, but the threat of it alone will influence play a lot, and this would definitely give Cinc a spot on at least some of those Gengar Lead teams. It's unlikely to push Gengar out of the meta altogether, since it lacks that notable Normal immunity that Gengar contributes to teams, but people looking for high-risk-high-reward strategies that want the fastest Sleeper out there now have a new option.
The third, and perhaps most notable move, is Super Fang. Super Fang in RBY is... really cracked.
To put things into perspective, Super Fang was enough to put Raticate on the OU VR for a long time. Because everything in RBY is absurdly bulky, even the meta's strongest attackers often only 3HKO their targets. What prevents Tauros from just running through everything on turn 1 is that most things do in fact take Body Slam into Hyper Beam from full HP if no crits are involved. In an environment like this, a move that just cuts the opponent's HP in half is massive. Cinccino is barely strong enough that its Hyper Beams generally do around 50% to most neutral targets that aren't massively bulky, including Starmie, Tauros, Chansey, Zapdos, and Exeggutor. I think it becomes pretty clear what we're going for here. There are very, very little things that can safely switch into Super Fang and not be at risk of dying to a followup Hyper Beam. On a mon that outpaces Tauros and isn't Raticate levels of frail, this is... pretty fucking huge.
Super Fang pushes Cinccino from being down with mons like Dodrio and Kangaskhan as "I wish I was Tauros" to a genuinely huge offensive threat in its own right. It's pretty risky to use because of Super Fang accuracy and needing to get predictions between Super Fang and Hyper Beam right a lot, and truthfully most of the mons mentioned aren't guaranteed to die to Super Fang->Hyper Beam from full HP (unless a crit bails you out) but with a bit of luck the chinchilla will definitely tear massive holes into the opposition. And if you thought it was bad enough as a Tauros replacement, imagine it as a Tauros partner. Even if Cinccino dies in a 1v1 it could never win, the bull in the back will gladly take on the Rhydon or Snorlax it forced to low HP and do Tauros things afterwards.

Sample Sets

Fan of Fang (Cinccino)

• Super Fang
  
• Hyper Beam
  
• Thunder Wave/Thunder
  
• Double-Edge/Thunder
  Do Re Mi Fang (Cinccino)
  
• Sing
  
• Super Fang
  
• Thunder Wave
  
• Hyper Beam/Double-Edge
  

Checks & Counters

The biggest counter to Cinccino is without a doubt Rhydon. Super Fang will massively chip it which is not appreciated in the slightest if Rhydon wants to deal with Cinc's teammates or get Substitutes up later, but beyond that, Cinc isn't doing much to it. A movepool consisting entirely of Normal and Electric moves can obviously not touch Rhydon for relevant damage, with a crit Hyper Beam capping at 38% and a non-crit one not even KOing after two Super Fangs.
Gengar provides similar issues. Interestingly, Ghost isn't immune to Super Fang in RBY, so it still takes hefty damage from it, but Cinc's moves that actually finish a mon off after Super Fang can't really touch it. The most it can do is Thunder, which without a lucky crit caps out at 20%. Gengar meanwhile has a good shot (factoring in Gengar's crit rate) to 3HKO back with Thunderbolt or Psychic of its own or can just Explode for roughly 75% right away.
Very physically bulky things like Lapras, Cloyster (if no Thunder) and Snorlax can win the 1v1 by tanking Super Fang into non-crit Hyper Beam, if they have options to 2HKO Cinccino back. They can't exactly hard switch in though, since a second Super Fang (or just a Double-Edge) will put them in range of Hyper Beam afterwards. This also means stalling Cinccino out with Rest isn't exactly viable, since Super Fang->Super Fang->Hyper Beam will just KO.
Alakazam and Jolteon can serve as checks by virtue of outrunning Cinccino and 2HKOing (in Alakazam's case) or threatening to paralyze, which immediately makes Cinccino's threat level plummet by making it unable to overcome foes with recovery without paralyzing them itself. Jolteon in particular can't even be TWaved back. Needless to say though, neither of them has any business hard switching in: Jolteon will always fall to Super Fang -> Hyper Beam even from full HP (with the tiniest bit of chip, even Double-Edge does the job) and while Alakazam can play mindgames with Recover and potentially Reflect, Cinccino's high crit rate and Alakazam taking at least 80% from even non-crit Hyper Beam turns this into a game it really wouldn't want to play.
While not outrunning it and not at all being a safe answer, Starmie also annoys Cinccino because of the speed tie making a lot of its options unsafe to go for. If it Hyper Beams on a turn that Starmie outruns and Recovers, this gives Starmie a free turn to do whatever it wants, often paralyzing it. Clicking Super Fang on a Starmie that's slower that turn on the other hand wastes Super Fang PP and Starmie will still be out of Hyper Beam range on the following turn. If Cinc spends a turn paralyzing Starmie with Thunder Wave, chances are Starmie will paralyze it back the first chance it gets. And in a raw 1v1 where both mons got a safe entry, they're likely exchanging TWaves as well, which Cinccino hates way more than Starmie does.
Lastly, rather than looking for hardcounters, a good way of dealing with Cinccino would probably be to just outplay and outrpredict it. Super Fang into Hyper Beam gets many KOs, but Hyper Beam on its own gets little. This puts Cinccino in prediction-heavy situations a lot since not getting a KO with Hyper Beam means you're taking a heavy punish, but Super Fanging something that stays in will also have uneasy consequences. You're not Raticate levels of frail, but still take heavy enough damage from many things that you want to be careful with your HP, and if something TWaves you it's as good as game over usually. Couple this with Cinccino not being able to hard switch into most things safely, and Super Fang missing from time to time, and it makes these scary interactions perhaps a bit less likely to happen in practice.

Conclusion

Raticate, whenever it's ranked, usually lingers down in E because even Super Fang has trouble truly popping off on a mon with stats that atrocious. However, we've seen that with Cinccino, which is way faster and just strong enough to pick up many 2HKOs that Raticate lacks, the genuine threat level is there. The mon is pretty cheesey still because of a reliance on good prediction and luck, but teams without a Rhydon are definitely in trouble against a well-played Cinc, and if nothing else the holes it opens up for a teammate like Tauros in the back are easy to exploit.
It's tough to imagine Cinccino being actually reliable and splashable enough to be on par with the A Ranks, but the threat level is likely enough to hang in the lower echelons of B1 still around Cloyster and Jynx's level.
As for how the rest of the meta might adapt to Cinccino's presence, Rhydon might see an increase in usage as probably the hardest answer to it, though being tasked with answering Cinccino does greatly cut into its capability to deal with Chansey later (funnily a Cinccino could kamikaze into an incoming Rhydon with Super Fang -> Super Fang -> Hyper Beam to put it into Zapdos Drill Peck range though, but this isn't likely to be worth it). Gengar is a mixed bag: on one hand it also beats Cinccino decently well, but on the other it faces competition from it as a fast sleeper, so its total usage might not change too much. Snorlax might run Hyper Beam a bit more so that it can properly 2HKO Cinccino with Body Slam -> Hyper Beam (or just Hyper Beam an already chipped one). Lastly a lot of "second Tauros" mons in the meta will likely greatly fall off if not become irrelevant altogether. Kangaskhan has a bit of a case to stay with its good bulk and Rhydon matchup, but Persian has it rougher since it lost its speed tier advantage and Super Fang does more than Slash on average. Raticate, unsurprisingly, would never even be considered for the E Ranks again, its minuscule niche of hitting Rhydon with Bubble Beam not nearly being enough to make up for everything it loses compared to Cinc.
What do you think? Would scarfy have an easier or harder time doing stuff in RBY than I imagined? Thanks for reading!

BITE Data updates, April/May 2024 - new data sets added, and a sampling of data sets that were updated this past month for screening list, commodity controls, BITE List and more.
Access the BITE app for a 14 day free trial to check out the data we have online: https://app.bitedata.io.
https://preview.redd.it/dojqf3hlts0d1.png?width=3300&format=png&auto=webp&s=aa884f43c9652f167cc2d025bd4510de564f8ce6
https://preview.redd.it/h6lpomxlts0d1.png?width=3300&format=png&auto=webp&s=18d3941ba1b907992e5d3b4dd12558600f95b0d6