Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

news.ycombinator.com

Post reply on HN
Print view View original

51–60 of 109 posts

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#51

Post by HeatherJudd »

bambax wrote: post #46

> Most public company data is unstructured and textual. This is surprising to me. Wouldn't it be possible to make a very short list of, say, 12 blunt questions that would help flag fraud? Of course the company could be lying in the answers. (If it was found to be lying in any one answer then it would automatically be flagged as "high risk".) But this may not even be needed, since what you seem to be saying is that th…

An interesting idea. I'm not sure if say 12 or so questions could cover the risks, or it if would just add one more data point for people to (not) read?

There are certainly some sorts of transactions that are much more risky than others - so having an easy source for these transactions would be useful (and we'd built it into our models). But often the signals are more subtle. Companies with overly aggressive accounting policies across the board tend to have completely different undisclosed problems. Since there are estimates and judgement involved in all areas of accounting looking at the aggregate impact of all policies can be important.

On adding more information - The SEC adopted rules to modernize disclosures of risk factors in 2020, requiring a summary risk factor disclosure if the risk factor section exceeds 15 pages (https://www.sec.gov/news/press-release/2020-192).

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#52

Post by andreabee »

Congratulations! I played with that database in 2017 when they opened up the text. I'm so happy to see it's finally been picked up.

I built a thing which recreated the Income statements and then flagged for non-conformance. I found Babcock & Wilcox Enterprises "Good will impairment charges" as an anomalous line in their Nov 8, 2017 filing just prior to when their CEO resigned in 2018 and they had to make so many adjustments. Unfortunately, I'm just a data geek. I did have a financial advisor friend who was VERY interested, but we couldn't get enough interest to validate development, so I moved onto other projects.

I'm no finance whiz, but that database is a trove. So happy to see a company develop around giving it a go!

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#53

Post by xgboosting »

Hey there, I have some experience summarizing 10-K and 10-Q's for a personal project, I clusterd BERT embeddings then selected the clusters that had a high correlation with non standard price deviation in the next 20 days. You can check out the results on https://eclect.us/

I'd love to learn more about the methods you're using and I would also like to know if you're ever looking to hire more developers.

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#54

Post by piesauce »

xgboosting wrote: post #53

Hey there, I have some experience summarizing 10-K and 10-Q's for a personal project, I clusterd BERT embeddings then selected the clusters that had a high correlation with non standard price deviation in the next 20 days. You can check out the results on https://eclect.us/ I'd love to learn more about the methods you're using and I would also like to know if you're ever looking to hire more developers.

I had a look at your website and the related blog post. It looks pretty cool! We are working on some abstractive summarization projects but haven't deployed them yet. We are not hiring right now, but please do send me a note at suhas @ the bedrock ai domain.

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#55

Post by hbcondo714 »

Congrats! Did you by chance work with the YC team that launched MarketBrief? I don't think they are around anymore but they did SEC Filings too, albeit 10 years ago:

https://techcrunch.com/2011/08/15/yc-funded-marketbrief-make...

Disclosure / Shameless Plug: I work on https://last10k.com ...a consumer offering for reading 10K/Q reports more efficiently

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#56

Post by piesauce »

hbcondo714 wrote: post #55

Congrats! Did you by chance work with the YC team that launched MarketBrief? I don't think they are around anymore but they did SEC Filings too, albeit 10 years ago: https://techcrunch.com/2011/08/15/yc-funded-marketbrief-make... Disclosure / Shameless Plug: I work on https://last10k.com ...a consumer offering for reading 10K/Q reports more efficiently

Thanks for the wishes! I have never heard of MarketBrief, this is very interesting.

We most probably couldn't have done what we are able to do today back then because the ML scene was quite nascent.

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#57

Post by ZeroCool2u »

piesauce wrote: post #42
ZeroCool2u wrote: post #40

I've worked with 10-K's and 8-K's extensively for the purposes of using them for NLP. This is extremely arduous work and a clear winner in terms of profitable ideas, so kudos to the team for the launch, this is really impressive. Perhaps this is giving a bit too much away in terms of the secret sauce, but would love if you could talk a bit about how you handle the wild disparities in the structure of the documents. D…

Thanks for the kind words! We don't use XBRL at all. We did try it initially, but it was wildly inconsistent across companies. I think one of the things that worked well for us was that we spent a lot of time at the initial stages of the pipeline (efficient sentence and word tokenization, span detection), that bode well for our models later on.

Thanks! This is similar to where I ended up landing as well. It turns out using a non-standardized standard format is practically worse than dealing with giant blobs of plain text!

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#58

Post by piesauce »

andreabee wrote: post #52

Congratulations! I played with that database in 2017 when they opened up the text. I'm so happy to see it's finally been picked up. I built a thing which recreated the Income statements and then flagged for non-conformance. I found Babcock & Wilcox Enterprises "Good will impairment charges" as an anomalous line in their Nov 8, 2017 filing just prior to when their CEO resigned in 2018 and they had to make so many adju…

Thanks for the wishes. That's really cool that you were able to do that!

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#59

Post by kbennatti »

ZeroCool2u wrote: post #57
piesauce wrote: post #42

Earlier quoted context omitted.

Thanks for the kind words! We don't use XBRL at all. We did try it initially, but it was wildly inconsistent across companies. I think one of the things that worked well for us was that we spent a lot of time at the initial stages of the pipeline (efficient sentence and word tokenization, span detection), that bode well for our models later on.

Thanks! This is similar to where I ended up landing as well. It turns out using a non-standardized standard format is practically worse than dealing with giant blobs of plain text!

So true

Re: Launch HN: Bedrock AI (YC S21) – Using ML to identify red flags in SEC filings

#60

Post by kbennatti »

andreabee wrote: post #52

Congratulations! I played with that database in 2017 when they opened up the text. I'm so happy to see it's finally been picked up. I built a thing which recreated the Income statements and then flagged for non-conformance. I found Babcock & Wilcox Enterprises "Good will impairment charges" as an anomalous line in their Nov 8, 2017 filing just prior to when their CEO resigned in 2018 and they had to make so many adju…

Seconding Suhas here. Super rad
Post reply on HN