sumitup <ideas>
11 subscribers
2 photos
5 files
20 links
ideas box

Chat group: @sumitupchat

Bot: @sumitupbot
Download Telegram
Forwarded from Telegram Contests
The Telegram Data Clustering Contest starts now.

Based on the input data we provide, you are expected to develop an algorithm that:

1. Identifies content in English and Russian and discards the rest.
2. Identifies news articles from the result of (1) and discards the rest.
3. Classifies each news piece from the result of (2) into one of these 7 categories: Society, Economy, Technology, Entertainment, Science, Sports and Other.
4. Identifies news pieces about the same event and groups them together into news threads.
5. Sorts news threads based on perceived importance.

Below is the sample input data. We will be publishing more sample data sets as the contest progresses. Check out the detailed description of the contest task here (and here in Russian).

Participants have two weeks until December, 2 (the deadline is 23:50 Dubai time) to come up with a solution and upload it to @jobs_bot.

The authors of the best solutions will share a prize fund of $100,000 and will be able to take part in the second stage of the contest, getting a chance to claim another $100,000 in prizes.
DataClusteringSample0107.tar.gz
230.7 MB
Here's a sneak peak of the data format you'll have to parse in order to compete in the Data Clustering Contest that starts in 2 days.
Forwarded from Telegram Contests
DataClusteringSample1821.tar.gz
415.9 MB
More sample data for the participants of the Data Clustering Contest. Have fun!
Forwarded from Markus Ra
contest2225.tar.gz
437.9 MB
Forwarded from Telegram Contests
The second stage of the blockchain contest starts tomorrow and will last 2 weeks. The 41 winners of the first stage will be able to take part. The prize fund is $100,000 + another $100,000 subject to additional conditions.

The results of the Javascript contest will be published within the next few days. Next week we are going to host the solutions submitted for the Data Clustering contest in a way that would allow anyone to test their output. Stay tuned!
Forwarded from Telegram Contests
Submissions for the first round of the Data Clustering Contest are now available for public testing:
https://contest.com/data-clustering

Everyone is welcome to leave their feedback while our judges tally up the results.

For reference, this was the data set we used: raw, en, ru. We will be feeding more test data to the submitted algorithms in the coming days.
Forwarded from Telegram Contests
The second round of the Data Clustering Сontest starts today.

The task is to create a module that could be used to power a news aggregator.

Check out the full description of the task here: https://contest.com/docs/data_clustering2

The prize fund for this round of the contest is €100,000.
The contest ends on May, 25 at 23:50 UTC.

Everyone is welcome to participate, including contestants who didn’t take part in the first round of the Data Clustering Contest.
Forwarded from Telegram Contests
Everyone is welcome to test submissions for the second round of the Data Clustering contest in a new mode that shows today's news:
https://contest.com/data-clustering-2

We have also added comments with results of our preliminary tests. Stay tuned for the final results.
New feature to ask questions from the text you send, use with /add command followed by text or reply directly to a message with /ask question
New Features in Sum It Up bot 🤩

- PDF export of URL
- Screenshot of website
- Technologies used by website
- Full page length screenshot
- Webpage translation instant view
- Question Answers from text message

You can use help command to know more about them, https://t.me/sumitupchat/615
Forwarded from Lulzx
#todo merge alternative to search bot in sumitup
sumitup <ideas>
#todo merge ocr into sumitup https://github.com/lulzx/ocr
🗣 OCR functionality has been added to sumitup bot, simply send a picture to extract it's text :D
what's bad / should be added / changed?

- tags are useless, make the +/- of tags configurable ⚠️
- reading list not implemented yet, should be added asap 💡
- the summary should be better, list key points and dialogue ⚠️
- try out implementation of abstractive summarization for links 💡
- outline is sometimes blank, it should be shown in output result ⚠️
- no conversation analyzer yet, merge message merger in a recorder way 💡
Forwarded from Lulzx
next features:

- add outline of document ✔️
- enumerate heading list
- h1, h2 in bold ✔️
- link to h1, h2 if they have id="something"
- something to do with youtube links
- ??? (your request here)
Forwarded from Somebody (+/%#)::'not j'
#bug Instagram login 🌚
Forwarded from Lulzx
#todo Auto generate YouTube video summary
Forwarded from Somebody (+/%#)::'not j'
#idea YouTube summaries
Two options:

1. Use Chapter timestamps
a. Author provided
b. Third party
c. Extract from comments
2. Summarize subtitles