HyperScience Reviews

This is an OCR solution. We were testing it for one of our existing processes. We were checking its capabilities on how it can provide streamlined services for data entry processes and automate the step that was done manually. 

It was on the cloud, and we had a test instance given by HyperScience to perform all these things.

For our use case, it provided a complete workflow. It reduced our existing touch time by 80%.

From the point of view of technology, it is one of the leading platforms right now in the OCR world. It provides the best accuracy for handwritten forms, which is a struggle in the industry. You can take processes with a lot of manual work and streamline them through this tool. So, there are multiple benefits attached to HyperScience being inducted into any organization.

HyperScience provides a lot of features. It can auto-transcribe the handwritten information on the forms. You can create as many forms as you want in order to provide HyperScience information about the input coming up for similar forms. It has an AI that helps to improve the accuracy day by day as you work on it. It gives you a better output every time you work on it. There is a lot of flexibility around it. 

There is a very robust reporting mechanism inside it, which can give you all sorts of data that you need in order to improve the process and provide feedback to HyperScience.

We tested HyperScience on structured formats of the forms, and it worked really well. We tested it with semi-structured formats, and it worked well. We also tested creating a new form that does not have any format. HyperScience has less capability while working on unstructured forms. Unstructured forms are those where there is no standard structure and the information can be anywhere on the form. They need to develop this capability. Such a capability would make it a perfect system for processes like data entry, accounts tables, etc.

Secondly, before putting the actual volume in the tool, I believe that we have to create a manual form every time. If the procedure of creating the forms can be shortened or automated, it would help. It would reduce the manual effort of creating a form every time you start working on it. There should be an automated form creation system that does not require any manual intervention. Whenever a new form comes, the system should be able to create a dummy form for that. That will help.

I worked on it for almost 1.5 years.

It was stable. I didn't see any issues with its stability. It was also good performance-wise. It has been delivering what was required. 

It was scalable. There wasn't any issue. The company I was working for was pitching it for other projects as well, and it was receiving a lot of attention there.

The number of users was changing. In terms of the regular users, there were a maximum of three people, but during the different phases, the number of users ranged from 20 to 50.

They were really good. They had good product knowledge. They were able to provide the solution and customize it the way we wanted.

I don't have any previous experience with a similar product. 

It was quite simple. It was laid out very easily, and there was, as such, no complexity in terms of the deployment. Any complexity would be more from the point of view of the HyperScience developers. They have to integrate it with the existing system, but that should not be very difficult. I believe its integration can be done with any system. So, that's not an issue.

In the deployment, including the people who were testing it, five to seven people were involved. It took us about a month. The HyperScience support team was available to help us with deployment.

There was no maintenance from our side. The HyperScience team was doing that. They had the developers who were working on it.

I don't have any idea about the licensing costs, but I know that there was a per-page charge. They had different models of pricing. One of them was based on the minimum number of pages, and we had a fixed price. So, the price model depends on the volume that is coming through HyperScience.

If it fits your process requirements, you must try it. It would certainly give you a new dimension and better results.

I would rate it an 8 out of 10. 

My projects are from different PDFs. I use a KFI8. First, we create an email using email extraction. After that, we email or scan the document and print a standby so we have their input in different formats: email, scan, and print. We then put it into HyperScience, which provides the JSON output. Now we can automate using any application, such as RPA Automation Anywhere, which gives you the RPA UI Path. 

Valuable features include tools like IQ Bot and the ability to extract handwritten documents with 93-95 per cent accuracy, which can be improved using ML technology. 

Extracting tables from certain documents could be improved. 

The next product release should include a utility that provides HyperScience JSON output. It would be related to any application, like other platforms and include Power Apps. We need to impact their REST API, but we would need a connection bridge application. And the second feature would be HyperScience providing different outputs. At present, HyperScience only provides JSON output. If a business needs a different output, we must create the utility ourselves and provide that data to XML or something similar. But if HyperScience could do that, we would no longer need third-party apps. 

We've been using the solution for one and a half years.

The solution is stable. 

We have not directly connected with the HyperScience team or technical support. 

The initial setup was easy for me.

Our company has several areas of business, including health and dental insurance, that require handwritten documents. By extracting these documents, we can fully automate the process. We evaluate the HyperScience needed after extraction and build IS or web applications, depending on the extraction. Once we have the output for all fields, we introduce better applications like ISL Java.

I would give the product an overall rating of seven and a half out of 10 because it has a lot of functionality, whatever we use. The extraction process for still and noisy documents and some data is good. The need for manual intervention is minimal. A few things could be improved. Not having to do QA would increase accuracy. We also must create different user data types if a particular data type is missing. 

We currently have15 people using the product all the time.

It can only be a cloud-based program because it requires more GV whenever we install HyperScience pieces of more than around 250. So we have to install the relevant cloud servers and gain access using their URL. 

If you are a first-time user, you must show what kind of documents you have. You also need to estimate how many documents you will require each month. Next, you must decide whether structured or non-structured documents would better meet your needs. Based on your decision, you create the appropriate document, considering t which fields are required and which are not. You can then estimate the number of manual tasks that might be required.

It's a good extraction solution if you're looking for an IQ Bot and ARN documents. So. if it's a typed document, it's fine, but whenever you're looking at something handwritten or a still or noisy document, you have to go to HyperScience.

We use this product to help our internal teams automate in areas such as claims automation using intelligent document processing. It deals with handwritten documentation. We are partners with HyperScience and I'm a data science manager. 

The product performs well on handwritten text and on checkboxes with regard to data extraction. It also has some algorithms that can detect a document template even if it is rotated in different directions and if the image has a lot of distortions. It has a great feature called visual template matching, which identifies the template and extracts the data.

The solution lacks support for a variety of languages. It supports three or four languages, whereas other vendors support 60 languages. I'd also like to see some improvement in the various dictionaries, including address validation dictionaries, or phone number validation dictionaries, those things need to be improved. It provides a machine confidence score but doesn't give accuracy scores.

I've been using this solution for two years. 

This is a generally stable solution. 

Scalability is not relevant to our situation. 

HyperScience provided training and we had a lot of technical support from the HyperScience team.

Positive

We didn't have any problems with deployment. We had a HyperScience team assisting us. Deployment time will depend on your requirements. We spent four weeks doing a POC. New template training took a couple of days because we had around 1000 documents.

I'm not sure about licensing costs but for one of the vendors, I was told the cost was $1.50 per page.

HyperScience is good for structured and semi-structured document extraction and it's easy to learn. The solution is built in a way that even if you don't have a lot of knowledge, you can still get the results.

This is an evolving company that hasn't been on the market for a long time but it's doing a good job and I rate this solution seven out of 10. 

We use it for all of our structured, unstructured, and handwritten documents.

As far as the product goes, I really like the product. It’s clear why it is the market leader.

We have seen pretty good accuracy.

There are all different kinds of use cases that we get out of it - especially for its core capabilities. We are experiencing pretty good results with the core capabilities overall.

They could work on the price and make it a bit more reasonable.

HyperScience is already working on some classification models and text classification models in their documents. That's a good roadmap for them. Roadmap-wise, they're on the right track.

For three years, the company has been using the product. However, I've only used it in the last six months.

The stability is good. It offers good reliability and performance. There are no bugs or glitches. It doesn’t crash or freeze.

It's absolutely scalable. You just make the queues, and you scale your queues. You scale the product very quickly.

There are no users who use HyperScience. It's all embedded into our automation pipeline.

We're still scaling up.

Everything is handled in-house. We don’t use outside consultants or integrators.

Technical support is helpful and responsive. We are happy with them.

Positive

We didn't use anything previously. We started with HyperScience. We wanted the best in the industry.

I did not handle the initial setup. It was set up way before I took up the role. I can’t speak to how easy or hard it was.

I don't know if it's that the teams are now experienced with the solution now. However, I have found that the upgrades are pretty smooth. We don't see any issues out there. All of our upgrades have been smooth so far.

My understanding is that functional deployments are pretty quick. I haven't heard of it being a lengthy deployment process. I don't know the exact time it takes to deploy. However, I haven't heard of any deployment concerns on the HyperScience side.

Only a few people are needed for deployment and maintenance tasks. Wherever there is a kick-out, we do have a team that looks at those cases manually, which is understandable. However, it does have good accuracy. We only really need four or five people to just be around and watch.

We have seen an ROI since using the product. I’d rate it a three out of five in terms of realizing ROI benefits.

The solution is expensive. I can’t speak to the exact licensing costs.

I’d rate it a two out of five in terms of affordability.

We also looked into ABBYY and Amazon, among others.

We are a customer.

We're just upgrading to version five. I don't remember the exact version we are on right now.

If accuracy and customer satisfaction are your goals, then you should go for this solution. It's really a fantastic product in the market. We have seen other products doing stuff. They weren't good at all. They all claim to be excellent, yet they weren't.

I’d rate the solution eight out of ten.

We use HyperScience for extractions.

What I liked more about HyperScience was the quality of the OCR it is a lot better compared to Google.

There is a feature called Visual Template Projection which tries to identify the template of the document that is submitted regardless of the orientation of the document, whether it's visually damaged, or in very bad shape. It will try to match the template.

HyperScience could improve the unstructured data extraction feature. 

They need to support more languages. They only support a few languages, such as French, English, and German. Then we had to improvise because other vendors are providing more than 60 languages.

I used HyperScience for more than one year.

HyperScience is a stable solution, but there are some improvements needed in terms of the language support and support of dictionaries. For example, other vendors in the leader space, support many corrections, such as character, address, phone number, open, and validations. The validation dictionaries are very limited in HyperScience.

HyperScience is scalable. We have tested it with a load of 100 documents in a short span of time and it was working well. 

We received training from HyperScience and approximately 67 users are using HyperScience. Most of the roles are consultants.

We were part of the center of excellence team and I was managing the document processing space. Whenever we receive a requirement from vendors, especially for FinTech and healthcare, we are going to use on-premise solutions, including HyperScience, ABBYY, Parascript, and other vendors.

We used a few other solutions, such as ABBYY FlexiCapture, Recognizer, Parascript, Google Document AI, and Azure Form Recognizer service, but HyperScience was giving better results in terms of handwritten text.

We decided to choose HyperScience because the quality is very good. Most of the other vendors do perform very well, but for anything that has a documented quality of the below 300 DPA, you need a HyperScience to extract the content.

We have built a wrapper around the HyperScience based on the SDK available given by the HyperScience using the Python language. This is where we place our files. The action system places the files into a folder, and from there we pick the files and then process it. It sends back the result in the same way, to the output folder. We have built a REST API on top of the HyperScience SDK. 

The setup was completed in a week.

We had some support from the people at HyperScience for the deployment. We have one cloud administration that supports the solution. 

You need to maintain the environment where the HyperScience is installed. If there is a lack of disc volume, we have to clean up some of the data.  HyperScience does not require a lot of maintenance.

Clients have received a return on investment using HyperScience.

I would advise others to choose HyperScience if they're looking for high-quality extractions in case their documents' image quality is below 300 DPI.

I rate HyperScience a seven out of ten. There is a lot of competition in the market at the moment and HyperScience has to evolve a little better.

We are using HyperScience at an enterprise level for all of our document extraction.

We work in the financial services domain. As a result, a lot of our documentation comes through in hard copies as paperwork, such as application or enrollment forms. These documents are consumed by our business team in several different ways and to facilitate this, we use HyperScience. It is used for all of our papers that are coming in, they are scanned in digitally at our scanning facility.

To facilitate further automation or integration with some of the existing systems that we have, we use HyperScience to extract data off of these forms and then pass it through other applications or other tools that we use for automation.

A lot of our interaction with our clients has been through paper documents because of the nature of the business. When we were in process of digitizing our documents, it was not easy to onboard everyone to send in copies electronically. A lot of people were not very tech-savvy. They were still comfortable in sending things through paper. We wanted to keep that option for them without causing much of a disruption in the way we have our customer engagement.

Having a tool like HyperScience which facilitates the process of converting digital copies of the scanned documents and provides us with the extracted data, has helped us to facilitate further automation. Once we have the extracted data, there could be several different ways through which we can consume that data.

One of the most valuable features of HyperScience is the user-training module. Whenever the extraction takes place, based on the way we have trained HyperScience, it would give us some success status or a certain confidence level. If the solution has processed something that it determined was not extracted correctly it will queue those items for manual review.

The user-training module allows users to go in and retrain it by correcting a few configurations that can allow for better extraction. It will enable the extraction to be correct on the second attempt or in the subsequent attempts.

No solution is perfect and there are several different scenarios that could be improved in HyperScience. One area is where there are multiple tables in the same form I have seen HyperScience struggle. There is some issue with supporting the extraction from multiple tables involved on the same form. If this could improve, it would be a big benefit.

The document extraction works well in conjunction with other automation suites. If HyperScience can provide some way to directly integrate with these other RPA automation tools, it would be a good add-on.

My organization has been using HyperScience for approximately three years and I have been using it for a year and a half.

HyperScience is fairly stable. We haven't seen any major performance issues. We spend some time getting the training done right. Once HyperScience is trained and put into production, the performance has been fairly good.

When we have to extract multiple items or multiple processes that are to be automated from a lot of forms, this is where the scalability can improve.

This might not be a limitation of HyperScience but it could be how we have designed and implemented it for our purpose. However, the current cycle time that we have for processing one batch is quite high. That is where we see some challenges in scaling it. The increased time could be how we have configured it at our end.

We have approximately 50 to 75 end-user who are using this solution.

We have a team of approximately 100 that takes care of the extraction operations of the solution.

The implementation is done by a specific team in my organization.

We have a dedicated technical team of approximately 25 that works on HyperScience. They work in collaboration with the development team at HyperScience.They do all the upgrades and maintenance and all related activities.

Whenever someone is looking forward to using HyperScience or any of the IDP tools, I would recommend them to take a look at what their own use cases are, what kind of forms they have, and what documents extractions that they need. They then can map it with the functionality that HyperScience or similar tools have to offer and see what is best compatible for their use case.

HyperScience will be good at a lot of operations, but those operations might not be well-suited for your needs. it is best to look around at the options available on the market.

I rate HyperScience an eight out of ten.

My rating might be a bit biased because of my lack of experience with the solution. I may not be able to gauge the overall performance and capability that the entire solution has to offer. I may have only been exposed to a limited amount of functions or features.

We use HyperScience for testing purposes, document processing, and automation. We tested with both printed and handwritten documents that were related to the medical field. To get better results, we had to test the documents written in cursive and configure the maximum number of documents in HyperScience. If you want better than expected results, you must first configure more documents for that particular vendor.

We're automating invoices and some handwritten filled-out documents. We can automate with accuracy, and even though we used HyperScience initially for selling purposes and as a comparison tool, we're now getting these results with HyperScience. 

I like that compared to other tools, HyperScience works best with handwritten documents.

The product's usability could be better. The first pain point is that we're getting the output in a different format, and we were expecting a different timetable. The second point is that if you want better results, HyperScience says you have to configure a minimal PDF or a maximum of 400 PDFs. 

If you want results with 400 PDFs for what's written by these doctors, then you also configure the maximum of 400 templates for that. So, it's essentially a lack of support from HyperScience. In the next release, it would be better if failure scenarios were reduced. It would also help if they offered different formats, inputs or injections, and added different scenarios.

Whatever versions we have used were stable.

Technical support is responsive, and they will help you, just like other tools.

With Ephesoft Transact, we're getting these results. With ABBYY, we're getting these results. I think all the companies are also developing their own tools for this OCR. The difference between HyperScience and Ephesoft Transact is usability. Right now, I'm using HyperScience tools, and we are expecting basically the print form of the document. For table extraction and all these things, we're expecting the help of Ephesoft Transact. In terms of usability, Ephesoft Transact is one of the best tools for this extraction. But in the 2019.2 version, they were unable to capture the handwritten documents. They released a new version in the last month, 2020.1.05, and now they support handwritten documents.

For example, for an invoice table and table extraction, we have to set the required time to set up this table in ABBYY and other tools. They use new technology like AI/ML, and they have reduced the time to extract what the table contains. Initially, I think that it required about ten to 15 minutes to configure the table. Right now, they have definitely decreased the time with the help of AI/ML technology. 

As a result, in terms of usability, I would say Ephesoft Transact is the best one. We were trained to use HyperScience because the management sanctioned it. They definitely compare other OCR tools for usability or scalability.

The initial setup was straightforward.

HyperScience was set up and installed by our own manager and the team.

I think HyperScience is cheaper when compared to other tools. They also have a lifetime licensing option.

The biggest lesson I have learned is that we have to configure the maximum number of templates for unstructured documents, and they have given the range that goes from a minimum of ten to a maximum of 400. That makes it a time-consuming job.

On a scale from one to ten, I would give HyperScience a seven.