Sign Texas Work Order Computer

Document type sign work order texas computer

hello and welcome to today's webinar the latest trends in document classification and extraction I'm Teresa Resik director of webinars here at aim and a mature host and producer of today's event and with me today are a team of experts and capture extraction from access technical group now introduce them in just a moment but thank you for taking the time out of your very busy day to join us today like to thank access technical group as the underwriter of today's webinar without the support from our solution providers and wouldn't be able to provide you with these free educational programs like our webinars so access thank you for your participation and now I just want to introduce our panel of speakers today and today we have Kevin ELLs who is the director of marketing for access technical group and for decades kevin has been providing product marketing strategy for leading organizations in the enterprise content management and document imaging industry with the passion for new capture technologies mark Warner is the senior vice president of product management and R&D and he leads the product management and development across various disciplines handling commercialized and enhancing the access AI platform and as in additionally he's supporting the real estate and financial verticals mark has spent close to 25 years in the C and VP level software and product development leadership experience in real estate information and technical industries then we also have with us Vince tonello who's the vice president of sales at access and Vince has over 20 years of solutions selling experience of complex enterprise software solutions to Fortune 500 companies and global 1,000 clients his focus has been in the business process management area with a heavy concentration on automated information classification and extraction and business process improvement at access mr. Tornillo focuses on domestic and international direct and channel sales so right now I am going to turn things over to Kevin to begin the presentation today thank you for the nice introduction my name is Kevin I'll be your first speaker will later be joined by Mark Warner our VP of R&D and Vince Tornillo access VP of Sales let's review the agenda with a full program ahead well first discuss the challenges of complex document data extraction and document classification Marc will explain how access AI offers a new solution that incorporates natural language processing and machine learning to address these challenges I'll provide a complete product overview and demonstration and then Vince will discuss the benefits of our solution offering to service bureaus system integrators and resellers let's get started we'll assume the majority of today's audience is fairly familiar with document imaging and capture solutions therefore we'll get right to the core of our presentation the challenges of classifying and capturing index information or metadata from complex document formats so let's define what we consider complex documents there are really three main document layout structures structured documents or forms these consist of pages that have consistent layout and design the index fields never change therefore it's fairly straightforward to set up software to predict data locations for automatic data extraction semi structured documents follows somewhat predictable index locations like an invoice or healthcare explanation of benefits but additional data and they fall into tables of columns that change based on a number of lines of data reported on the document unstructured documents of the real challenge index data desired can change from document to document and there are many cases and countless numbers of variations of data positioning on these documents due to the embedding of the required data being embedded within a sentence a paragraph or randomly occurring anywhere on the page examples unstructured documents include correspondence contracts deeds loans public land records and even emails today's presentation will focus around how access is developed the solution to take on not only this most challenging but how the technology can also be applied to really any document format we'll also discuss how the access hosted solution can be a great alternative for vendors like yourself who prefer to outsource the capture process of document management this can enable you the service provider to focus on offering an end-to-end solution without the burden of purchasing and learning how to use a software solution even if you could buy one that addresses this most complex data extraction challenge for any document imaging and data capture solution there can be challenges the years software companies have been developing data extraction solutions the majority of these solutions have focused on the structured and semi-structured solutions the ability to capture data from forums is a very mature market and there are dozens of excellent solutions out there however you still have to purchase and train your team how to use the software the semi-structured market has solutions for many departmental needs accounting and invoice processing representing the largest segment of the market and therefore the best target there are at least half a dozen excellent solutions for this sector by developing template or patching matching solutions invoice processing is now a well-established segment of the document capture community and honestly a somewhat saturated and competitive market the most complex and challenging scenario is therefore unstructured content a lot of companies claim to be able to work with these documents but from our experience very few have proven to be capable of successful execution access is developed a solution that fundamentally changes the economics of capturing information the solution access AI is the most advanced document classification and data extraction solution on the market I'd like to share a little history on axis and how this solution was first developed the underlying intellectual property has been in development for over ten years led by one of the brightest minds in the area of natural language processing however because the technology is built on multiple sets of proprietary algorithms that are orchestrated together the processing power required to run the solution was regulated to the arena of super Peters was therefore cost prohibitive as computing technologies advanced and costs came down the viability of using the solution across everyday use cases increased significantly access made a decision to launch the product to market when we first developed the software into production setting we're up against one of the most difficult unstructured document sets out there the software performed exceptionally well the software's ability to understand and accurately capture information far surpassed any other solution available and rivaled the quality one would experience using double data entry approach since those early days we've applied the software to multiple different use cases as will be discussed further today and it continues to prove itself as a way to fundamentally change the economics of capturing information since the original vision and the years of development of access AI for unstructured content access has acquired a variety of clients but not only unstructured document challenges but also scenarios of semi structured and even forum based initiatives some of our clients have reached us to address document challenges like eeob healthcare records or even government tax forms what we have realized is that the technology can also be used for actually almost any style of document some organizations are not even looking for data extraction they just want their paper files organized and identified so their subject matter experts can navigate and manage their documents more productively document classification and separation come to the rescue in this scenario where they send us a mixed batch of documents and we sort them out provide index table information and deliver back a well formatted and classified document portfolio subject matter experts or smees at the client location only want to read the documents but they don't want to spend hours trying to find the right page in a portfolio mortgage loan packets with literally dozens of document types are an excellent candidate for this application patient records and charts or other examples and even oil and gas well files have been ideal applications for these offerings let's get back to the main focus of this presentation the challenges of an unstructured document data extraction here you can see a great example of an unstructured document scenario this example we're showing what's called a grant deed and a release feed they're part of a mortgage package and they are used to describe who's the legal owner of a property the document has critical data used in some real estate transactions including mortgage loan processing in title insurance but the challenges in numerous the documents change from city to city county to county it can not only be challenging in the layout but also in the language used to describe the various key fields that describe the document here you can see three individual designs that are literary saying the same thing about the property the format changes dramatically and the indexes can be presented in tables or buried in sentences when we emphasize the actual key index information you can see they are not only complex to extract but it also take an expert to even decipher this page to find the pertinent index fields it's really a complex layout not only in some cases the data is buried in the sentence it's also in a language not suitable for database queries for align a business system many of our clients want this data normalized and reformat it to be consistent in their system for example dates may need to be all number like day day month month year year or more complex challenges like taking the descriptions and turn them into alphanumeric patterns like a lot 15-block G phase the section to converted to 15 - G - D - - traditionally due to the complexity of these documents clients have had to rely on internal data entry staff to process these pages or they have outsourced the requirements to a third-party vendor one of the pros about sourcing is that can help your business efficiency by not having to hire manage and support your own staff on these projects but it's expensive to outsource key-in to a third party sure it's cheaper than doing it in-house but the costs do add up if you're only doing a few hundred documents this may not be a significant issue but most of our clients are dealing with thousands in some case millions of documents for some of our clients the my cost of king also made it cost prohibitive to capture all the key data elements they wanted in some cases they had to settle to only what they could afford we have customers who were capturing only 12 data elements using data entry shop when they actually wanted to capture upward of 200 it was just too expensive when relying on a third party to hand key your data you can run to some other issues as well one of the issues we see come up time and time again is a drop in quality many of our clients have come to us because of the quality from their existing outsourcer was suffering sometimes the poll quality stems from an inability to retain staff members that have true understanding of the documents since most of the unstructured documents are relatively complex operators require considerable training from experience we see a high turnover rate in the data entry shops when the staff leaves so as the knowledge and the quality suffers the beauty of using access AI it never quits as I mentioned before document imaging companies have been around for decades in fact many of the current access team work for these companies they make great products for many of the imaging challenges most of which are forms or highly predictable semi-structured use cases they use great OCR ICR image recognition barcode extraction and patent matching but they are not using some of the greatest new advances in language recognition technologies they aren't truly designed to read the document one other drawback from other software vendors is you actually have to buy the product take extensive training to master the software now it's not unusual for a software solution but it is a huge investment if you're only using it a few times not to mention you need hardware to host these applications and if you're processing tens or even hundreds of thousands images that's not a cheap investment at access we're creating a new alternative now like to pass the stage over the mark our VP of R&D and product development tell us more about it thanks Kevin at access we have an innovative new solution to extracting data from unstructured content our new solution is different as it does not use template or forms based data extraction template informs Bay data extraction require known patterns and actually coding to those known patterns to extract data from unstructured content that template informed based extraction is a problem because patterns change frequently between documents so you'd be constantly coding to account for every pattern this could be a problem if you're dealing with thousands of documents or tens of thousands of documents companies have tried to solve this problem by using existing technologies to extract unstructured data their technologies work well creating searchable search indexes like google word bags to be able to go in and search content but for unstructured content they do not work well we're actually trying to extract data from the content itself not just a general sense of what is inside of the document we aren't we are specifically looking for a word but we're looking for a word that has meaning like a person as the borrower Valon or looking at a rate of a mortgage on a mortgage document or actually trying to look at services rendered by a doctor on an explanation of benefits form so what were using is new state-of-the-art technology is called natural language processing natural language processing we start bisecting the document into its lexical and grammatical components for example we look at a sentence and we can mark the nouns and verbs the nouns would be the person places or things and the verbs would be the actions between those entities and we can have context of how they relate to each other so in the document we can identify all of the animals all of the places and then we can look for a specific data extraction point based on its context when we're looking for a person or we're looking for a place so we actually look at the the lexicon of the language or the English language in the unstructured content and we break the component of words up into components based on grammar just like the human brain and we look at the relationships between those components and create patterns mathematically create patterns for the structure of the sentence when you're targeting a key data element that you want to extract from the unstructured content mathematical models and the relationships between the grammar the lexicon of the language and the entities themselves are utilized to create that model it's a repeatable pattern that when we encounter a similar pattern in another document we can automatically extract that data element the advantage of using software to do the unstructured data extraction is it learns over time and become smarter as you train it further while while working with the software the quality and the accuracy of the data extraction increases over time the software itself never forgets never calls in sick and never tires speaking of never forgetting software the confidence of the data extraction is critical and we don't want the software extracting data that is incorrect because at that point it will make all data being extracted extracted subject to scrutiny the software itself will prompt the user to help identify areas where it cannot extract data automatically when the user identifies the data that's was supposed to be extracted the mathematical models using the natural language processing may be modified it's critical at that point that the soft are understands what the user identified as data being extracted such that we don't update the model erroneously based on operator input so we won't just utilize what the operator has done we look at patterns of the operator extracting data using that scenario multiple times so that it's a repeatable process so this product sounds complex this technology super advanced you're probably asking yourself how am I gonna learn all this and who's going to learn it the beauty is you don't so currently access as host is as a service we host the hardware we host the software we train the system on your documents and return you the data you're looking for it's as easy as that I'll pass it back to Kevin now to explain how it works thanks mark I'd now like to review a sampling of our customer use cases and the benefits they received title plants is an industry term for database used in title searches for insurance and they are built up from key data from publicly recorded real estate documents government offices this may include deeds mortgages default or foreclosures liens judgments assignments releases and other documents typically to build a plant vendors would have to go to public offices and either scan the paper documents or acquire images from the government office then due to the complexity and the variety of the document data types legal descriptions best things etc the images are traditionally sent offshore to vendors who will manually key the information this may not seem like a huge task but consider that some title plans contain records on millions of properties that's a lot of manual data entry not easy to key and requires specially-trained data entry personnel by using access ai data extraction automation these documents can be processed in a fraction of the time required by offshore vendors our data entry team would still be used when automation is unable to extract the required fields because of poor quality or illegible documents this could be the case when the deed of a property was handwritten which is often the case on a hundred-year-old land deeds used in the gas and oil industry the advantage of access our software will determine what documents can be automated and what needs human intervention the process is much faster and more importantly cheaper as I mentioned previously mortgage loan packets contain tons of paperwork in many cases mortgages can run into the hundreds of pages and dozens of key index fields needing to be captured from a variety of document types and formats while some Docs are similar many buried by provider for one of our clients it was taking over three hours to sort manually through the loan to find and extract up to 100 to 300 fields of data by implementing a software alternative this could be done in minutes quality control assures the data accuracy and human intervention could be provided when the software is not reaching the levels of competence set by the service level agreement what we have discovered is the more pages in the document or package the more valuable of software solution the software does not lose its place on a document it can process and read the pages exponentially faster and boy does it do data entry quickly in the area of health care we have a number of clients using our extraction solution for claims processing extraction has been set up the process CMS 1500s and 1450 forms also known as hicfa not only is the data extracted but the business rules are applied to return the data in a standardized format called 18:37 now 18:37 is a standardized format designed to expedite the goal of achieving total EDI or electronic data interchange for claims processing and payments while claims processing has been around for years the challenge has always been extracting data this is an excellent example of a situation where pure extraction solution just didn't cut it the client needed to capture information not just returned as a table of data but in a format capable of EDI submission of a healthcare claim we have another client who needed tax forms processed although this is really a form based solution and there are other solutions for accomplishing this the client wasn't interested in maintaining the software licensing or setup in configuration they realized that the tax forms changed every year and they really didn't want to deal with setting up and training their software every tax season they had tried using another outsourced provider but they found the quality of the data returned was inaccurate they absolutely had to have high-quality results not only did access provide accurate high-quality results but by using access as a solution provider they knew we were available on demand and all they would need to do is send us samples of the new tax forms each year in the oil and gas industry a client of one of our partners wanted to extract data from public land records in order to search for mineral rights and capture related information at the outset of the project our partner was using a hundred percent hand keying approach however because of the unstructured nature of these land records even the hand carriers had a hard time capturing the information correctly so the quality was suffering and our partner was falling behind when they came to us they were about to lose the contract within 15 days of engaging with us we had the system up and running fully trained and extracting data and within 60 days the project was back on track in addition the partner was able to reduce their internal costs by 50% as compared to their hand keying operations it was wildly successful and fundamentally changed the way this partner approached oil and gas projects from that moment on that's just a small sample of our client installations we've also been developing proof of concepts for clients for a number of scenarios banks searching for client contracts for regulatory impact data companies pulling information from disparen sources healthcare companies trying to ascertain information in freeform text like clinical trials law firms searching for correspondence in a discovery process we look forward to hearing from you regarding your other challenges that you've encountered extracting information from other unstructured document scenarios as we have mentioned access is a hosted solution we do most of the work I'd like to run a couple of poll questions regarding your use of outsourcing for capture you have a couple of questions but great thanks now we are take this first poll question so let me go ahead and ask the first question out there to the audience members and it will pop up for you in just a moment here and the question is what types of content are you processing today and you can select multiple answers here our our your content types structure such as forms semi structured such as a OBS invoice P ODS get tables and things like that or unstructured content such as correspondence email contracts freeform text mortgage things like that so we're going to go ahead and I'm going to ask Vince to come in here and chat with me about this and the results that we're seeing and and Vince I'm starting to see a wide variety of different content types that people are processing here and I want to comment on what you're seeing here yeah looks like it's pretty evenly distributed a you know across all all the content types which is not uncommon we see that pretty regularly throughout the industry with the structured and semi-structured probably being the most mature this will kind of lead us into our next question of looking at that unstructured content and how they're currently processing that today okay so let me go ahead and push out this second poll question out to the audience members and this question is for those who are processing unstructured documents how are you capturing that information and again welcome to select multiple answers here are you doing that through manual data entry on premise out of box software or using on-premise software with custom scripting and feel free to go ahead and click those answers there and and it looks like we're seeing a bit of a mix with the manual data entry or the on-prem software with the custom scripting are you seeing that also in your own customer base as well there Vince yes you know this is split exactly down the middle 50/50 so we see a lot of solutions that are out there that require extensive amount of custom scripting to get them to work and then also manual data entry is probably the most prevalent that we've seen out there but it looks like the audience it's a it's a pretty good next 50-50 okay I'm actually refreshing my my to get results button here and I'm actually seeing that mix oh it's still I'm still seeing a high number here at manual data entry as well right yeah it looks like I just refreshed as well and there's a little bit of out of the out of the box software being used as well so that's a probably below 10% it looks like okay well let's go ahead yes I think you'll be I think they'll be excited to see kind of what we have to offer here as far as adjusting this unstructured content okay and let's go ahead and look at our third poll question that we're going to ask the audience members and it's if you're processing unstructured documents how how well is that working and this is select the one here you know is it great is it good is it fair not so well at all but let us know how how it is working for you within your organization so we're going to go ahead and push that question out to you there and go ahead and click on those results they're being okay here the answers coming in here and I see that it's working just like good to fair for the people in the audience members no one's answering that it's you know great fantastic for them or not quite well at all so definitely seeing just a middle-of-the-road responses in here yeah just refreshed and it looks like it looks like it's 50/50 there as well and again we see that because of the nature of the unstructured content being highly variable being embedded in paragraphs it's very difficult to get good results across a wide variety of document types well um well thanks everyone for answering these poll questions and I think we're gonna go back to Kevin for some more of the presentation thank you for sharing your information on the poll questions I'd now like to go over a product overview and demonstration the software for those familiar with document imaging product follows basically the same process of capturing classifying extracting data validating and delivering it to a back-end system here's a high-level overview of the product model images are brought in ocr'd converted to XML we extract the information apply rules on the information and check for confidence if there's an e validation or error correction required we can do that and then we export the information for import to your line of business system when we capture information it's generally a paper format usually a tiff or a PDF file but we can re electronic documents data sources II documents and even email what we'll do once we capture them as well convert them to a standardized format to input into the system next the documents are classified by they access AI engine to determine what kind of document type in the case of a mortgage it could be a deep release HUD any type of document that's been predefined by the system if we receive a document that's really a combination of multiple documents we'll go through and classify the documents by looking through the document looking for starting and end points in the for each individual document will separate the documents and then create unique files or images for each unique document type next the extraction process will occur and access AI engine will look through each document and extract the unique fields specific to that document type our quality control and validation operators will be notified by the software for character level correction or review and fix any questionable characters keep in mind a system learns as we use it so if a question occurs for example as you see here probably wouldn't really happen but if the letter I was mistaken for the letter 1 the system would prompt the user saying this is a noun it probably shouldn't have a number in it prompt the user to correct it and then the system learns that never prompt the user for that in the future once the data is corrected processed and reviewed the system formats the data for the client we export the information for import into the line of business system now in the interest of time we cannot spend the time required to show you the whole training process of how our sufferers trained how it's tuned and so forth someone spent just a few minutes talking and describing a little bit of demonstration of the training process of how we capture and classify the information train it and then teach it to validate or learn as it goes the first step as I mentioned before is capturing the information and converting it to a standardized format once we have documents in a standardized format we then want to teach the system what these documents look like and define their names we do this by taking samples of truth data documents examples of documents and teaching the system what a document actually looks like so it takes samples here of example of HUD's releases deeds etc and define what a document looks like now in many cases as I mentioned before documents especially in unstructured changed dramatically but they do represent the same document for example here are seeing a couple of release documents and they're dramatically different now the software will see this if we drop a whole bunch of a folder and say these are all release files and the system recognizes there's massive differences in these documents software create what's called the clusters now clustering works in a couple of ways what it could do is group a bunch of documents that are all dramatically different then it could also set up another group of a bunch of documents that are very much the same once you have these different clusters we go into a training process and show the system where the information is on these different document configurations or layouts here we're looking at the access I configuration module for defining document types and field names this will be a very short demonstration and we're owning a define just a couple of documents and fields we're going through here and defining to document types a deed and a release and just one index from each of those documents once we define the documents we're going to take samples and highlight the information that we want to extract off the documents and we tag the information we take as many tags as required to get a good sampling of the document now what's unique about this is we couldn't not only tag paragraphs or sentences we can tag whole areas of the document that spend multiple pages and the software won't capture information that's in between these tagged areas some paragraphs like this one right here can span two pages but you wouldn't want the information that's sitting at the top of the page you just want the context of these paragraphs now we seem to have a good number of captured samples here and even looking at this you can see how much information could be required to type into the system if you're doing manual data entry we're now going to select another batch or another cluster I'm going to go ahead and choose to tag the borrower name now you can see this borrower name is an embedded index that can be found inside a sentence and you'll see from just a small sampling the format of this somewhat similar but still different in the way it's described inside the sentence with different words around it different order and certainly different locations on the page once again we'll take as many sample as required and tag the information on this particular cluster and as we showed earlier there are many different clusters for this type of document format now once the training is completed we'll run a batch of documents through and see how successful it does extracting and finding information from those documents at this point if it fails to find information on certain documents it will separate those documents for us and requi e us to go back and train even more once this trainings been completed we'll then tune the system to find the documents and assist it on maybe defining if it's a now and if it's a verb other things to help the software tune in and find that information we're looking for more accurately once we have a good training model we'll go ahead and apply it against real documents and see how it does against the real documents we'll extract the information and then perform that post-processing conversion we'll take a look at first the extracted data and as you can see this extracted data could be just a single word multiple words are per in this case all upper case could be uppercase and lowercase characters and in some cases there's terms that need to be reformatted using business rules in this example we can see the lot was extracted as a lot 15 the user the client actually wants that information just provided back as 15 and this could be renaming a trustee name to change the order from Jennifer Doe to doe Jennifer many other types of business rule processes once the information has then been converted we'll send it into a validation to make sure not only to be extract the right data but the correct business rules are applied we're now looking at the access AI operation manager and here you can see we're moving through a variety of different documents looking at the extracted values and also looking at the post-processing values you can see as we go through the navigation it will highlight the fields that we are reviewing showing both the extracted value and the post-processing value results as you can see here from these examples in the case of the string of section dirty it's been changed to a number format of like 30 - 7 - 16 W we find that many of the systems especially for the land title industry this information needs to be database friendly and it's really changed to a format that's more practical for querying from a database we can see here examples of different strings that have been changed from uppercase lowercase change the order and extracted various fields out or removed certain fields like the example here of the block 7 is now just reduced to a 7 we can also run some rules to run a little reformatting for example here the lot 21 - 26 is now being changed a lot - 21 22 23 24 25 26 much more practical for somebody searching to find information on lot 24 so once this information has been reviewed and checked for quality is then released for our clients to import into their line of business system as I mentioned before this is a hosted service where we're doing everything we're hosting the software and data centers we're doing a configuration setup validation soup-to-nuts in the future there will be ability for the clients to train on their own documents and validate their own data or a hybrid solution where we train the system but the clients validate their information as I mentioned we're currently hosting our solution in data centers and we're moving it to Azure for greater scalability I'd now like to introduce Vince Tornillo our VP of Sales thank you like talking a little bit now about the benefits to service bureaus BPOs and system integrators and even potentially bid entry shots just like many of you I've been in the capture business for a long time providing advanced data capture and abstraction solutions for a variety of large capture providers I've worked with many of you in the past and while these solutions work well for ongoing projects or applications we were always challenged in working with B POS and service providers due to conflicting business models let me explain what I mean let's talk about the ongoing project and where it worked well for an ongoing project might have the time to set up the system to train your employees to acquire the hardware they the licensing on the software because it was an ongoing project in the RO I was there but for typically for B POS or service providers or even system integrators our data entry shops these projects are typically shorter they might be three to six months in length he would spend that time trying to set up the system to get it tested in production so the oral I just wasn't there secondly was licensing challenges you typically charge per document based on the number of pages based on the amount of data you're extracting or typing in or pulling off the page where the software provider would typically charge a volume either monthly or yearly and those two bottles didn't mix due to fluctuations due to projects being finite to typically have to buy more software than you needed and those models just didn't work while access AI offers a solution for your most challenging textual applications one of the things that we wanted to do when we brought this to market was be considered very easy to do business with we really wanted to align your business model with our licensing instead of trying to fit that square peg in the round hole that we banged our heads on the wallet for years talk a little bit more about that later so the first benefit is really speed to production because you don't need to train your people stand up a systems test get into production we can get you up to speed very quickly we use our resources we train the model we're very familiar with it we'll work with you to make sure the data is accurate we can get you up and running very quickly that would address those shorter term projects being able to realize your savings much quicker technically we offer turnkey solution we have the ability to do the validation on our end where we can return back classified separated and extracted information validated back to you or you can leverage your resources or we could do a combination of both additionally we can integrate into your workflow so it's seamless to how your users use it today while we can handle all your document types you may already have a solution workflow that works very well for your structured documents but need a solution for your unstructured documents and send us your unstructured documents and we can send those back to you as well you can brand it as your own there's no training on your part other than if you decide to do the validation it's extremely scalable we can spin-up and spin-down processes and spontaneously we can scale on-demand handle volume fluctuations and meet those accelerated turnaround times so let's talk a little bit about that licensing again we decided to align this the way you do business so today you charge by document by number of pages by the amount of data pulled off the document characters we can align our model with the way you do business today we can provide you a per document pricing based on the number of fields or the number of classification elements so it's an alignment with how you do business additionally we have the flexibility to look at alternative pricing scenarios I have one client that's looking at transactions they have a transaction it's considered a package you may go back and forth five or six times to them and they need to extract data each time we can price it out that way so we have a lot of flexibility on how we do business in the pricing model there's several factors that go into it one you need just classification you need classification separation you need data extraction and then on the data extraction what is the complexity of that extraction is that just a simple field like a date or or an account number or is it more complex like a legal description or some information buried in a paragraph so we take all those factors into account and we price it out and typically we say that we can save you anywhere from twenty to fifty percent over how you're doing it today if you're doing manual data entry or if you look at the total cost of ownership but buying software or Hardware creating people you're getting a system up installed and running we feel very comfortable that we can say that and we're willing to prove that out to you actually give you a high-level overview of how we can work together please feel free to reach out to me personally if you have a project or you're interested in in discussing a partnership back to you Kevin thank you Vince I would now like to review what we've discussed today and talk about the next steps and open the floor to question and answer what we have discussed today are the challenges complex document data extraction document classification and how access AI offers a solution that incorporates natural language processing and machine learning to address these challenges you've seen how the solution help solution providers like yourself provide a faster solution to your clients that reduces errors and improves productivity and yet costs less than other alternatives like data entry it requires little training or additional software or hardware investments now we have a lot more to demonstrate on the product but the interest of time we could not show you it all today we would be delighted to set up a one-on-one demonstration and dive into the discussion of your clients and their challenges plus answer any questions that we'd not address today or covered in Q&A after the presentation we'll be sending you a follow-up email with contact information and links to more information on the solution I would now like to open up the session for questions some of your questions may not be appropriate for our general audience and if we don't get to your question we'll follow up with you directly within the next few days but those of you without questions and dropping off the presentation thank you for joining us today we hope you found the presentation stimulating and informative it's been a pleasure presenting to you we look forward to hearing from you soon thank you now let's get to the questions so thank you and yes and let's move on to those questions so Kevin let me have you come back in here once again and go ahead and take over for the rest of your team here thanks Teresa can you hear me all right most certainly excellent so first of all I like to welcome our foreign attendees the reason I bring that up we have a couple of questions coming in from South America and pretty much the same question but a different language the question is do we support other foreign languages besides English since the software is written for natural language processing English and of course they're looking at languages of both Spanish and Portuguese mark you want to field that one sure because we leverage several technologies to deliver a complete solution we can handle different languages for structured and most semi structured documents the natural language processing portion of our solution for the unstructured documents relies on the lexicon or the vocabulary of the language and the grammar that sets the rules for sentence structure and our data extraction engine today has algorithms that are currently only configured around English language with that said well other languages are not out of the question and couldn't use the access AI technology depending on the demand so it is possible to load other lexicons and grammar sets into our engine but there is some minor tuning that has to happen around the algorithms that process that data it is possible today we are not doing it we're only using the English language excellent thanks mark another question and we really didn't cover this of course this is a really a document management document imaging type solution but we didn't really talk about us scanning the pages can you the question is do we scan the pages and mark only to field that one again - okay thanks given currently today we do not scan but we do have partners that can take care of that for you we haven't had any requirements from our clients today to scan documents and many cases the documents are coming from existing content management posit or ease so but we do have partners available to do the skinning for you if needed thank you next question is you know during the case studies we've really talked about big installations big projects what is the smallest size project the access would take on things how about you take on that one yeah sure thanks Kevin yes so we again we handle a variety of different sizes of projects we actually have done some smaller projects with one of our large partners it was for a large US bank with a few thousand documents tax documents that came in a couple of times a year and the challenge they had was being able to handle the changing forms on a regular basis and be able to scale the volume very quickly and while it was pretty small like say a few thousand over a couple of months twice a year what they liked about the solution was our ability to get up and running very quickly and get it get very high quality results back to them in a very complex data form to integrate into their system the other piece of that was is it really the partner liked it because it really opened a lot of doors within this large financial institution I've shown that they could do so they're actually going after a much bigger project so we'd look at you know look at those and be able to work with anybody on the size of the project depending on what their needs are thank you since Marcus one is for you this is always a tough one who is the competition nice they said structured document capture we'll talk about that because there's a lot of structure tools out there in the industry but we see the number of competitors entering the market over the years but with our experience talking to our customers and resellers dissolutions really don't deliver well the forum's looks good and during the devil phase but in production there was a sizable amount of custom development of scripting and they were finding their resumes weren't being met on data quality and things like that so um for us by combining the computer automation human validation resume and quality exceeds the levels we've seen from other vendors also though we do allow the customer to use and do their own validation and correction using our data tool sets so the SLA you know for both the actual automation part versus the data extraction and validation part can be done in-house or find a customer itself said maybe you want maybe you want to touch on that on how we or both of you guys but probably not the same time on how do you handle these custom requirements or custom integrations that required so access his history you know we've been a custom software development shop for 13 years so those those types of custom integrations working with other systems are very familiar with us so we always consider that kind of the easiest part of the solution I've only one off and developed this unstructured solution we decided to tackle that big challenge first but we we feel very comfortable that we can get into virtually any type of system through XML through customs you know through api's or a variety of different custom integrations as needed so you know there's since there is some custom requirements and some integrations how how do you price I mean you cover some of the pricing before but how do you stay competitive and in such a market with so many different options out there yeah well first of all the solution being able to extract that information I don't know if you wanted me to get into the actual pricing or pricing scenarios given or want to talk about just the custom part just the high level is how we how does it stay competitive yeah so typically we can say that you know if you're doing this manually or you're leveraging some out-of-the-box solution that requires you know expensive setup we can stay pretty competitive by leveraging our resources to build and maintain the system so that you don't have to we have people that are very knowledgeable in in our solution so we can set it up very quickly and we typically say that we can save you probably 20 to 50 percent over kind of how you're doing it today very good well I'm looking at the clock here and we have a couple of other questions but they're very specific to that person's needs but there anyone just came in I think it's worth talking about how do you process handprint and both with drop out boxes and cursive and Mar ecology field that one and I think if I make that the last question in the interest of time and I'll allows Teresa to come back and do some closing remarks but mark and writing sure give me this didn't give me the hard question here driving canal come on there are we can't we can process I see are but obviously that's the intelligent character recognition type engines that look at hand print the problem of hand print is its so the concern so much that the quality of hand printed documents unless you're telling the user or the person creating the document that this is going to be recognized by a computer it causes major issues if it's blocked print that is doable using an IVR engine and reading that in and using our technology as we talked about during the presentation drop off boxes and chat boxes and things like that not a problem we have a technology to do that but the hand print itself that always raises rate of questions and we don't we don't spend a lot of time today with our customers in the hand print arena but we have that question asked often and if needed we will use I see our engines to try to accomplish that all right so it did Teresa closing remarks most certainly thank you we have the contact information up here on the screen here for access AI if you go to the access - AI comm you can get some product information or you can reach out to these guys by the email address that's on your screen and just wanted to remind everyone to below the slide area if you would click on that feedback and survey link and offer us the feedback and I'm also clicking a few slides up here with some other resources at a on aims website more information and help you with to learn more about the all the different ways that we can help you in with improving your business and strengthening the connections with your customers and so all of that's all really good information to have there and then now that we are at the end of the of our webinar hour I do want to ask each of the speakers that we've had with us today for their closing thoughts or a key takeaway because there's a lot of really good information that we've shared today and I'm going to start first with Vince turn yellow your closing thoughts today yeah well I one first want to thank everybody for attending and you know what we've seen in especially the poll questions and it's pretty standard in the industry that the unstructured content is is really been a challenge and most people are hands-on in that or a good portion of people are handing that manually and what I'd like to share with you is take a look at what we can do I think we can give you a lot of time and money and being able to offer this to you and kind of a turnkey fashion thanks Vince and Mark Warner your closing thoughts today you know as we watch the industry over the past years and currently we're seeing a lot of changes in the technology sector around natural language process and machine learning so this is coming this is here today for us for data extraction it is successful the software works as described it works almost like a human brain creating mathematical models around sentence structure and and grammar so you know we'll continue to enhance our product over the years that some day the old humans won't be needed for a lot of these processes we see the change is happening year to year and humans are always my dog humans are always treated yes you're correct they're always thank you maybe we should carve manufacturing something is going to be the computer just sitting there processing all the day done yeah definitely thanks mark and Kevin else your closing thoughts today first of all I hope computers don't take over marketing guys and I want hopefully in this market for many more years one of the reasons I joined this organization was because I fascinated by document imaging document capture technologies I've been doing it for 20 years and when I saw this technology I saw the potential for really an untapped market now we're tapping some pretty obvious opportunities here with like correspondents and contracts and some of these documents were seeing in there the legal space in the healthcare space but I think something that's what we're looking forward to is what we don't know is where this technology can take us in the future many cases people are just tapping this information to put it into a database for our line of business process but there's other ideas we've been churning around about using it for big data analysis I love big data analysis lots of power processing power out there to look at information but in many cases that information buried in documents was unaccessible because it was cost prohibitive to go in there and find these different variables and different patterns that were going on in documents so I think it's exciting time as Mark said to be having the power of these machine learning engines and human natural language processing and the only will be exciting to apply the document imaging and big data analysis but as you mentioned before expanding it up to other languages and making this universally acceptable tool for everybody around the world so I'd also like to thank a more hosting us today and also like to thank my co-presenters for doing an excellent job you guys you're you're on the marketing team for this week at least well thank you Kevin and thank you access technical group for sponsoring our webinar today and and certainly for sharing the valuable information that you have presented always enjoy listening to your presentations so that's what we do have time for today for aim this is Theresa Resik and we will see you next time and have a good afternoon