Video: Moving Past the “Acrobatics”: Automating PII Detection, Redaction & QA | Duration: 1868s | Summary: Moving Past the “Acrobatics”: Automating PII Detection, Redaction & QA | Chapters: Series Introduction (6.51s), Session Recap (63.22s), Technical Setup (167.715s), Sensitive Data Review (220.425s), Redaction Challenges (293.22s), Redaction Demonstration (518.13s), Document Production Setup (1269.145s), Q&A and Closing (1673.72s)
Transcript for "Moving Past the “Acrobatics”: Automating PII Detection, Redaction & QA": Hello, everyone, and welcome to the third session of our four part webinar series on Logical for Public Records. I'm Laura from the Reveal team, and I'll be your host for today. As a quick recap, so far in this series, we've covered how to streamline request intake and management, how to collect, process, and search through large volumes of records to find what's responsive. And this is the third episode. So if you miss those sessions, you can catch up on demand using the registration link that we have included in the docs section. And today, we are dedicating a full episode to one of the biggest roadblocks for many teams managing records requests, and those are reductions. We will cover how to identify sensitive information, apply reductions fast and efficiently, and make sure nothing gets overlooked before the records are disclosed. And to walk us through that process, we are joined again by, Jason Christian, our principal sales engineer. We should be joining soon by Daniel Shuring, who is our principal product manager, but he's trapped in one of those, you know, laptop reboots that are mandatory for companies, like, right... I don't know. It was, like, ten minutes before this session that started. So, hopefully, we get Danielle in very soon. But for now, Jason will kick off the presentation. Good morning, Jason, by the way. Good morning, Laura. Thank you. And, yeah, Daniel should be joining us shortly. He actually is the man of the hour, but I might become the man of the hour today for that, which is gonna be quite alright. But I will definitely get us started on the slide presentation today, and then I'm sure he'll slide in and be able to take over the demo at some point. So today is session three of four sessions. And so if you were with us before, we have already done two sessions before. So just to kinda recap, where... What we've done, what we've shown if you have not been on the past two sessions. On our first session, we talked about our public records request intake portal where citizens, citizens of people can come in and actually be the requester and request any type of document. And that starts inside of our portal that integrates with the LogiQL platform. And so, basically, we go all the way through for... All the way through being submitted, approving it, acknowledging that we're gonna go through it is within scope, and then we create a project and logical to actually start working on that particular project. Session, Jason. two. Yes. Not not meant... Not meaning to interrupt, but, we are not seeing your screen if you're sharing. So. you may need to send, the thank you. so much. I I thought I did start the screen share. I really appreciate that. Yeah. No worries. Alright. There you go. Now we. are seeing too much of your screen. Oh, now we're... I apologize. I've got the wrong screen on there. Give me a quick second, everybody, and I will fix this. No worries at all. I have only two monitors, and some... Somehow, I always manage to share the wrong one That's okay. in every single meeting. That's exactly what it... I shared the wrong screen, so this should be coming up on here really quick. Okay. There you go. Just to quickly start over really quick. So session one, a couple weeks ago, we talked about submitting a request, putting that through the portal, going through approving, acknowledge that it's within scope and that we are going to go through and we're going to work on the actual records request that was put in. Session two, which was last week, is create the project through the portal, go into logical. So we talked about connecting and putting data in there. Right? So we we can either do the old fashioned drag and drop of data, and we can also go in and we can connect into some of the cloud collaborations tools. A lot of you are using Microsoft three sixty five, Google Workstation, Google Vault, all of that are, other cloud connectors that we can get into and automatically bring all of that in. We process the data. We organize. We search. We can tag, create, save searches, all of that good stuff. And now that we've gone through, today, we're going to talk about going through, looking for sensitive data. We call that PII, but going through and looking at exemptions. Maybe put on exemption codes, going over sensitive data, redacting out sensitive data, and that's what we're gonna be focused on today. So what are the challenges? Right? So number one, what's been happening today is a lot of our friends out there are still doing this process manually. Right? A lot of people are still using Adobe. A lot of people are still using Excel to track requests and all that type of stuff. And so they get the PDFs. They get the data. They put that into Adobe, and they're going one by one by one going all the way through. Some people are doing manual redactions, although Adobe can do some redactions as well too, but they're going through. And sometimes these requests can be large. Sometimes it's not just as simple as, you know, 10, 15, 20 documents. Sometimes email is requested, and email can get voluminous based off the scope and how it was collected and put into that. And then once we get the data into there, we have to process we have to process the data, but we also wanna make sure we are not missing anything. If there are sensitive data that needs to be redacted out, it needs to be identified, and it needs to be identified easily. And, again, the theme here, there's always a deadline. You're always you're always racing against the clock. Right? Some people have twenty four hour deadlines. Some people have twenty, thirty day deadlines. That's a little bit nicer. But people who have shorter deadlines, we wanna make sure that we can get everything right. So, a, not having to put your, put your eyes on all of the data, but also, two, making sure that we identify everything that needs to be redacted and kinda pull out of that. And then some people need to put exemption codes or reason why things are redacting. And so we're gonna go through we're gonna go through our library. We're gonna bring in some codes, civil codes, and stuff like that. So if that is you out there and you need to actually place a reason or a civil code or for some reason to put a redaction on, put a label on a redaction, we we can do that as well too. But at the end of the day, we have to make sure that it is all it is all redacted. It is all covered before things get sent out. Right? Anything... Even one thing that is missed could be... Cause problems. And what we have up here is just some examples of things that have kind of happened in The United States, things that have happened across, you know, the globe, you know, of bay... Essentially, you know, things that happen, if you miss it, what are the consequences for that? So, today, Jason. I'm here. we're gonna see in today's... Hey. How are you? Welcome, Daniel. Daniel is with us today. Danny... And Daniel is gonna be taking us through this today too, but I'll I'll kinda go through and let you know what, he is gonna show you today. So number one, we're gonna do... He's gonna walk you through the automatic PII detection. So once data is processed, we're going to automatically identify that on a project level, also at the document level. But we have the ability to also customize things. We can customize, you know, regular... We can build regular expressions to find something specific. Right? We build regular expressions for the masses, not the use case, but we can certainly customize and use those. But we also have other custom dictionaries to look up things, like other types of entities. Also, to use keywords, if it's a keyword that needs to be redacted out. Bulk redactions, global redactions, auto redactions, we're gonna go through that as well today. We're gonna look at our exemption library codes and some of the civil codes across the globe, United States and obviously for, GDPR, instances to bring all of those in and then be able to QC. Right? Making sure I've redacted everything. Everything looks the way it does. And then, of course, audit trails. Right? Making sure that we've we've audited everything. We put that we put the redactions on there, who's done it, when it was done it, and was it done correctly. And with all of that, Daniel, I am gonna turn this over to you for the demo presentation for this. So I'm gonna stop sharing my screen, and I will hand this over to you. Jason. So sorry about this, but those were the most stressful minutes in my life. Glad you could make it then. We need to talk to IT about those mandatory laptop reboots so they don't coincide with things like this. Yeah. And, normally, I did one this week already, so I thought I was prepared. But... Okay. You never, see him enough. and, hopefully, everything is up and running. So, yeah, Yep. my part him. here is going to perform the demonstration. So, let's take a few steps steps back because I know Jason did the introduction, which I kind of missed, but this is part three in our webinar series. And in the first session, what we did there was that we, showed you how to do a request. And, as a citizen. And then you get into what we call our dashboard where you have an overview of all the requests that are coming into your agency. And what we did, we created a request back then. And in this request that I've got here, I will open the request. We have here the request details. And also what we have here is the link to the project that we created in logical where we'll do the collection, the review, the redaction part in, for all the documents. Now Jason showed you in the previous webinar already the extensive way of how we can process the data and make sure that you have a proper set of data that you can use for your review and filter down the documents that you really need for the request that is there. So, also, in this case, what I will do is I will open from the records request. I will open our logical, and this is the actual request where we have already uploaded some data. So we already prepare In here, we will start, doing the reduction part. Now if I go into and move into the search where we can do all the filtering, you'll see that we have got about, 1,200 documents, and many of them are duplicates. So we see only here 330. And we have labeled, after doing all the filtering and all the metric that we can do with your documents, the relevant documents, the responsive documents down to 22. So we have only 22 documents that we need to review, and you see the list here. Now in my project setup, I've done a few things, and one of the important parts is that we can set our configuration for the redactions. So you will see here an overview of all the project settings and things that you can, modify and change here and basically manage your projects. And now we'll dive into the redaction part. So first of all, we have the redaction labels. You can add your own redaction labels, but you can also import labels from my, standard library that we have in place for various countries and states. So you can just import a library, and if needed, you can then modify those settings. Next to that, we have... Oh, yeah. One thing not to get. You can also say allow multiple labels per reduction. So you can set it to having only one reduction, only one single reduction label, or you can have multiple labels per reduction. Let me go to the auto reduction configuration. So this is where you find the rules, what you want to have automatically redacted in your documents. And there are various types how you can, ways how you can set this up, because we have, first of all, a option to detect automatically PII, personal identifiable information, like, for example, the name of a person, the phone number, Social Security number, and all kinds of related PI options. What we also can do is that we can include a regular expression, and I've done one here that detects the first part of an email address. Of course, I get all the pop ups of programs that will start to, to show up because I rebooted my laptop. And besides that, if you select, we can also add keywords. So you can also add the list of keywords that you can redact. So I can add here a keywords. Let's say I want to redact all instances of Laura. I can add here a label, like exemption six and the color for the reduction. And now all documents will be redacted for the keywords Laura. So we can do keywords, regular expressions, or, let's say, any type of pattern that you would like to detect and, of course, the automatically detected PII, which is using an AI model for that. Now besides that, we also offer the option to do redaction exclusions because there are always names of people that you don't want to redact or don't need to redact. And you can list those names, and those names will not be part of the auto redaction. So we will not redact these names in this automated process. Now going back to the documents, so I have my search here and I have my filter. So let me select the documents that are responsive. So I have here now a few documents. And so let's start with one of these. So I will open up here one of these documents like, the summary of today's meeting. And what you will see is that on the right hand side, we have the tags, and you will see that there are redactions here. We don't show them right away, but you can have a view where you will see all the redactions on the document itself. And these are the redactions, like, for example, the email address, the first part of the email address, but also the names that we have found in the document that will be automatically redacted based on the rules that we have created in our system. Now besides these automated rules, which you could, of course, can also can do is that you can create your own redactions. So let's say if you want to create a zonal redaction, you you can just drag and drop here an area and add here a label for a specific exemption. Confirm. And now we have added another reduction. So that's the manual reduction process. And besides drawing these boxes, you can, of course, also select specific parts of the text or paragraphs and perform a redaction there. And if I click on the label, what you can do is you can add here also a specific exemption code or reduction label. The reductions do have a few other options as well. If you click, do the right mouse click, if you wanna quickly add certain labels, you can see here a list of labels that you use a lot. So those you can quickly select without going through the full menu. And we also have the option for the properties where you can add an additional justification for, redacting that part of the document. So that maybe because it's really special reason, you have to, explain that to the opposite party or in court what your justification is for additional justification for redacting that part of the document. Of course, there are a few other options in the redactions. You'll see here all the redactions here, but we also have an audit. So you can see the redactions that have been changed. So you have a nice overview of that. So we also audit that. And, of course, there are options to redact, for example, a full page. So you can redact a full page, or you can redact multiple pages. So you can say that I want to have a certain page range that you would like to redact. Now, of course, I can save here the redactions. If you want to do a little bit more automation, you can also perform a search in the document. And I search, for example, for tomorrow. And then the highlights in the document can be redacted as well and also include there a reduction label. So you will see that we have the automated redaction. You have the manual redaction, and you have a way to perform a search in documents and redact all the keywords that you have found and redact them in one single go. And one thing I would like to show you a little bit more about is the redaction of, pages. And I've got here a larger document. And Jason was already showing you some examples of, news articles, and one of those was the, the full black page. So in this document that I've got here, I have redacted, I think, later on a few pages. I'll show you the redactions. I've done some full page redactions here. And so you see these pages are fully blacked out. Now if we're going to produce this for the clients, for the requester, we will normally end up with a lot of pages in black. It's not always that nice. It doesn't look that great. So what you would like to do in these situations is that you also would like to have the option to basically hide the fully redacted pages. Now we have an option for that, and I will show that later in the, production or in the download process, where we can say that fully redacted pages can be replaced by a blank page that highlights... That explains the reason why that document or those pages were removed from the document. And I already prepared an example. So I've got here... And that's the nice thing about Logical as well. You can also see the documents you already have prepared for the client and you already handed out. So this is the end result of this document once it is being produced. And then you will see that in here... And I will go to the later pages. Page 20. Where is it? Here we go. We see now that page 20 to 25 are with health with a specific exemption. And so instead of handing over a lot of black pages, we can just replace that... Those five pages with a single page, explain the reason why those pages were not present. And that makes it, yeah, much nicer to look at for the, recipient, instead of having all these, black pages there. One thing I also would like to, explain to you is that there's also an option, for adding, or tagging individual keywords, but also removing, items that have been redacted. So if you have redacted everything and, there's a mistake being made. You can have to... You have the option to select all documents, and then with the redact option, you can, add or remove redactions. So, we just, rejected, for example, name. I think I can... Why it was in there? I can add that as a term to search on, and now I can say come on to remove the redactions from all of these documents, or I... If I find the term, can lay... I want to add reductions to it. So that is also an additional option to manage your reduction process. So once you have redacted all your documents... So you've reviewed all your documents like the ones we've shown here. And if you want, you can also adapt to your workflow with additional responsiveness items so you could define your own, labels here to guide you in your process. So maybe you need to have a, second reviewer on these documents, then you can list that here. So all these documents are now redacted. So next step in the process is that we want to perform a download. And what I've done is that I've saved these documents as a single search. And now what I can do if I want to produce these and create a set of documents that I want to hand over to the requester, that I can, select here to say spur... Search. So these are these 22 documents that I had that I want to produce with 91 pages. And, of course, I've set everything in a specific template. So everything that you see is... Can be templatized. So if, if you create a project, you don't want to add your redaction rules... Auto redaction rules. It can be part of the project template that we're using. And this is the template that we're using for the download where we have now the 22 documents. We will include a load file, and a load file is a file with additional metadata for the documents, which can be the document number, the name of the document, or any type of information that you define that you want to release to the requester. There are a few options that you can do how you can define your production. First of all, we want to produce everything as PDF documents. So we're going to produce one single format for all documents except, for example, for, of course, audio, video files. There's a specific name that you can do. I'm adding here now a sequential document number so that each document gets its own individual number. And what I will do is that I will also include some endorsements from branding. So I brand a burn in the number of the documents, but also the, the individual number for each document. And we won't produce any native files in this case. Everything is going into PDF files. And then we have some other options, like we want to include redactions. And you can also select if you want to include all redactions or just specific ones with specific, redaction labels. You can make the transparent for review. So if you hand... Have to hand over a copy to the judge or the court, you can also make a version that has transparent redactions if that is required. And then we have the option to withhold fully redacted pages. And we can either remove the pages or, as I showed you just now, we can also, replace them with a slip sheet. For the base numbering, we can keep the original numbers or we can, renumber everything if that's needed. And now the only thing I need to do is give the name. So I will give this a... Call this the webinar production. So the next time, we can actually see this, and we we have here starting number, and we'll start that at 001. And now we can start the actual download. And what's happening now is that all the documents are collected. They are converted into PDF, including the reductions, and then this production, will become visible in the request management system. And that's where we will continue on for the next time where we will show you how we are going to release those documents to the requester. Do we have any questions, Jason? Nothing directly right now through the chat. We've had two questions that came up, and we did answer them directly. But, now is a good time, everybody, at the end of the presentation that if you do have any questions, feel free to put them in the chat so we can answer them for you. Yeah. And quick time check, since we were coming to the 10:30AM my time, we have a few more minutes, as a buffer. So if anyone if anyone is to bounce, obviously, they can bounce. But if anyone has questions, we still have a few minutes to answer them. So don't be shy. Think we're here in the presentation. I'll rise. Yeah. Perfect. Well, while, we wait... Brett, Rebecca had a. question. She said, is there a. limit on the number of exemption reasons we can input? And and the answer to that is no. That was an easy one. So, well, while we see if we get any other questions, I wanted to remind everyone that we have one more session coming up on December 15. It was originally scheduled for December 8, but we had to move it due to some, scheduling conflicts. And on that one, we will bring the entire public records response process full circle. We will be covering production and disclosure. Oh, yeah. That that was still the old date. Yeah. We will be covering production and disclosure, how to prepare and deliver responsive records, manage the final steps of a request, and make all the information that you select available through a public reading room. If you've been following along, this is the session that ties everything together from the moment a request comes in, that's what we covered the first one, to the moment the records are disclosed. You can find the registration link in the docs section again. We really hope that you join us for that one. We've also made some additional resources available in the docs section, including our public record software RFP template. There's also a comparison guide for agencies evaluating different solutions like GovQA or NextRequest. And there are a few customer stories that show how different public agencies are improving the records response process using Logikcull. And, of course, if you would like to explore how Logical for public records could work for your own agency, you can request a personalized demo using the bottom that you see above. And if we don't have any other questions coming in, which I think we don't, thank you, Daniel and Jason, for the awesome demo and presentation, and thanks for everyone who joined us today. We hope to see you back here on December 15. Thank you all for. joining.