Artificial Researcher

"Information wants to be free"
Stewart Brand, 1985

Artificial Researcher-IT GmbH (AR-IT) helps to do just that - freeing information in the world of scientific publishing, which is too often hoarded behind paywalls, allowing only the richest communities to access potentially groundbreaking knowledge. We are creating an innovative scientific knowledge management system based on Linda Andersson's (our CEO), PhD work developing several state-of-the-art patent text mining end-applications based upon natural language processing technologies. The cloud-based solution allows libraries of lesser means to provide researcher with a tool to: 

A) Find and retrieve relevant open access papers and 
B) Allow a much greater understanding of the content of any paper (and therefore assess price/value) 

AR-IT sets out to provide the growing, but financially strained, global scientific communities with one of the key ingredients for excellent research and development work locally.      

Today, researchers and students search for open access publications as well as restricted scientific publication via general search engines such as Google and Bing. Google scholar and google patent provide users results in a specific text genre. However, there is no safeguard that the users can find the data they need, and yet no advanced features to customize the search functionalities to a specific scientific field. The generally available search engines as well as the scientific search engines provided by the publishing houses still only offer a simple keyword search function with Boolean combinations. 

Moreover, from an abstract alone, one cannot know if the article provides   the information needed, and thereby be worth the price they ask for (often ranging between 39$ to 100$; too much for many communities in developing countries). 

Globally, 58,000 academic libraries are operating, with the vast majority's allotted budgets below USD 1 Mio. annually for purchasing access to scientific publishing and are therefore simply locked out of important parts of the scientific discourse. 

Researchers and students at academic institutions are affected by limited access to scientific knowledge. Our solution aims to eventually target this entire segment. Principally affected though, are groups that live and work in communities where access to education is already limited and where the financial situation of libraries allows even less possibilities to pay for the expensive access to scientific publications. 

We are in contact with library organizations in countries where the acceptance and usage of open science publication among students and researchers are high. For example, among Brazilian universities publishing in open science journals has been a policy for several years. Also, the African Library Association promote publication in open science and is therefore a prime candidate for further development of the solution to solve their needs for accessing state-of-the-art scientific research in fields such as medicine, agri-tech, etc. 

For now - to be able to quickly build and test our product, we have been collaborating with the Technical University of Vienna's library in developing an advanced prototype - AR-Science. Here we are in continuous conversation with researchers, students and the library to adapt to their needs.

The AR-IT solution contributes to developing well-adapted scientific knowledge management systems. The systems will support both the search and the reviewer process, which are an indispensable part of the researchers (and students) work. The key value of the product is that we build a platform that connects the information needs and requirements of libraries (and the universities behind them) and end-users.

Key values:

• In-depth citation summarization in order to assess if a closed scientific publication is worth purchasing 
• Retrieve the paragraphs that match the information need, i.e. avoid sifting through pages of non-relevant information

Possibly: 

• Recommendation lists of other scientific papers that could be relevant for the user, where the retrieved list is divided by closed or open access and cluster by thematic topics. 
• Request comparative summarization of several scientific publications based upon argumentative summarization techniques
• In-depth scientific summarization of with a focus on hypothesis, data, method and outcome.

The Artificial Researcher-IT GmbH, (AR-IT), start-up emphasizes on developing a novel and innovative platform, which automatically handles scientific information needs and presents the information to the users according to their requests. Within our product and services, we explore a new text mining technology to retrieve scientific publications better fitted to the information needs of students and researchers.

Part of the text mining technology was develop during Linda Andersson's PhD work at Technical University of Vienna (TU Wien), the paragraph retrieval, the automatic query formulation, and a merging method of document index and paragraph index. 

The traditional search technology requires keywords, however, keywords are a limited representation of information needs. The future scientific knowledge management system ought to be a dialogue between the user and the system, requiring integration of language, scientific domain knowledge, and understanding of user information needs. Instead of requiring the user to convert her information need into a set of English keywords, the system should aid the user to represent the information need on a conceptual level by means of user contexts. The text mining research must go towards categorizing the information needs and sub discipline of scientific fields. We argue that, by integrating an eLearning system and combining the benefits of supervised learning with reinforcement learning, we obtain partly self-learned annotated data that will boost the efficiency of customized solutions for each specific information need, and, at the same time, will provide users with added knowledge values

The AR-Science software is integrated into the libraries existing system as an add- on and hosted on the libraries infrastructure. AR-IT provides the services of indexing in-house, open access, as well as closed access if the library has the text mining permission in the contract with the information providers. Our solution does not require full-text storage, only a digital fingerprint is stored, and the in-house and the closed publication indices, as well as the usage information is owned and hosted by the library. AR-IT provide services in terms of maintenance, index update, functionalities update and optimization - i.e. we make the searching different type of resources more time efficient for students and researchers, but the libraries have the control of the data resources storage and the usage information. AR-Science provide more precise usage information in comparison with COUNTER standard. For example, the libraries have access to search statistics such as open access versus closed publications for each specific scientific field

Since part of our technology is consider to have Intellectual Property interest we cannot at time demo our technology and prototype. Mrs. Linda Andersson have presented her research work within the intellectual property and medical search industry at 

OAI 11 – The CERN-UNIGE Workshop on Innovations in Scholarly Communication in  her presentation Rethinking Scholarly Communication with AI: Keywords Are Not the Answer to Improve Your Search Result she discusses different types of information needs. 

For further information see: https://indico.cern.ch/event/786048/contributions/3360869/

When we started ideating for a solution for better academic search, we had several indications that the problem was a critical one. For example, the movie “Paywall” (https://paywallthemovie.com/) had shown us the relevance and urgency of it. Besides that, we had our own experience working on our PhD theses in different areas of research and recognized a pattern that did not just affect one subject or geography.

We modeled the solution on the needs of both researchers seeking knowledge behind paywalls and libraries operating under financially strained conditions not knowing the needs of their researcher stakeholders. Our algorithm significantly enhances the way researchers search and find relevant information – beyond verbal approval we were able to obtain first very positive experiential test feedback.

Libraries have been particularly intrigued by the opportunity to understand which paid-for journals are requested and where open source might be a full substitute for such. Beyond several verbal approvals we have started working together with 2 and expect further letters of intent in the near future.  

Currently we are serving 2 major Research Institutions in Central Europe, helping us develop the full product.  

Within one year, we will be working with at least one more - but have been gathering interest especially from institutions in Brazil (3+ Research Institutions)

In 5 years, our very conservative forecast is to serve 20+ Research Institutions, more ambitiously we are hoping to attract 100+ interested institutions for our cloud-based solution. 

Each institution serves approx. 5.000-10.000 students and researchers 

Within the next year, we are well on track to obtain a fully functional prototype, going the beyond singular functional readiness of the current situation. Furthermore, we will have finished developing the user interface allowing the technical roll-out to be starting and gained 300+ freemium users to beta-test the front-end (the add-in that is included in the library portal) Within the next 5 years we are planning a roll-out of the solution into developing countries starting with promising partner institutions in Brazil, followed suite by several central African institutions. At this stage, the medical and health fields seem to offer the greatest potential for change and journals are especially costly to access for libraries. We therefore plan to focus on medical use cases and medical institutions foremost. At the end of this period, we aim to have 100+ institutions operating our algorithm both as a cloud and on-premise solution 

Financial: We understand that the scale-up, that would allow the aimed for impact especially in the developing world needs funds – which we might not be able to receive in central Europe where impact investing is in its infancy and are therefore already looking across and beyond Europe in these matters.     

All our tests so far have been particularly positive in the technical realm – still we are aware that scaled cloud solutions will need expertise – especially supporting libraries’ operating and maintenance needs.

The most critical barrier we see are market entry barriers, with the established players in the industry having locked-in researchers and institutions in reputation-based publishing cycles effectively acting as market makers, controlling all knowledge exchanges. 

Financial: We understand the need for fundraising to be a constant in our endeavors, with a tightly managed cost plan and a long-term funding strategy that includes plans for lower-growth (and a faster route to profitability) we aim to overcome possible roadblocks in generating enough funding. MIT’s SOLVE for example serves not only great opportunities for encountering ideas and possible future partners, but also serve as a powerful publicity tool for such long-term ambitions for additional funds / growth capital.

Technical barriers that we face are in fact solvable with funding and by using the great opportunities that the location – Vienna – provides, by hiring talent from not too distant Eastern European tech hubs such as Romania, Ukraine or Serbia. For maintenance and consulting purposes we plan a dedicated team, and training librarians in utilizing the tool.

For the identified market-based barrier, we are in a favorable position towards the incumbents, since their solutions often shut out customers both in  many developed and especially often in developing countries. Similar to Christensen’s theory of disruptive technology, we plan to offer a cheaper, easier to use version to undercut extant market barriers. At the same time, we target the customer interface (libraries’ database interface) to get direct access to researchers. 

N/A

1 Full time 

4 Part time

The strength of our endeavor is our team’s competence diversity and our experience of working with industry. We have expertise stretching from marketing & entrepreneurship to solid artificial intelligence knowledge: Mrs Linda Andersson has in-depth knowledge in text mining and Natural Language Processing (NLP), Dr Florina Piroi has solid know-how in artificial intelligence (AI) and mathematics, and Mrs Jenny Andersson has 20 years of experience in quality software management. Mrs Nina Andersson has worked close to the executive management board at Tieto, Tele2 and Coop. Dr. Wolfgang Sachsenhofer has a background in business modeling and entrepreneurship, launching several (corporate) startups. Nina and Wolfgang have an in-depth understanding of the operative requirement to run a successful company. 

The AR-Science prototype is a two-year research project since this is a collaboration with TU Wien Library and department of Information & Software Engineering Group. Recently we also obtained commitments from Intellectual Property industry partners to engage in the development of AR-OpenAccess. The AR-Science solution is a considerable commitment and investment for the academic libraries and industry partners. Therefore they should have the possibility to test and evaluate the technical solution with the AR-OpenAccess before commitment. By first introducing the AR-OpenAccess for demoing and test trial to the research community – we can reach more universities and libraries worldwide. 

Our solution rests on a platform, that connects researchers and libraries with the relevant open science papers, as well as provides detailed information about papers behind paywalls - thereby generating user data for all libraries to utilize in their decision for lower-priced / re-negotiated access deals with publishing firms. Some of the ensuing cost savings are being monetized by AR-IT in form of an annual subscription fee funding the costs of the firm. 

For libraries in low-income communities, the cloud-based AR-OpenAccess solution is rolled out and scaled for free in exchange for usage data allowing AR-IT to build better and on-premise solutions (see above).          

We plan to raise revenue through a freemium model through targeting libraries in financially stronger communities. Model freemium: 

Freemium: e.g. In-depth citation summarization, retrieve the paragraphs that match the information need, 

Paid: e.g. In-depth scientific summarisation of with a focus on hypothesis, data, method and outcome; paragraph cluster retrieval, recommendation lists of other scientific papers that could be relevant for the user

Currently we aim to bootstrap our services with the generous help of two grants that we received in the process (AWS PreSeed; FemPower)

We believe in the power of networks and see SOLVE as a major partnership of like-minded individuals that aim to utilize (digital) innovation to change the world for good. 

Our mission to allowing all people access to critical knowledge rests not only on our product(s) being continuously challenged and thereby advanced. Therefore, we are hoping to create connections that help us advance our approach and spread the word. Specifically, we would see MIT and possibly even more so the SOLVE community itself as the gateway to connect us to interested university-, lab-, corporate- and syndicate-libraries to collaborate with. 

N/A

MIT library

GM - especially in their initiatives on STEM education for marginalized communities 

Innospark Ventures 

We think our AI-based solution would fit the prize's purpose very well. We are currently developing the technology backbone and advance the prototype jointly with university libraries. Our financial runway for the next two years allows us to be fully operational, with the exception of a gap of EUR 50.000.

(1) The prize of USD 50.000 would therefore be very useful in essentially closing this funding gap, allowing a smooth development and roll-out process.

(2) Furthermore, Linda's costs are only partly covered by the grants we have received so far - she therefore needs to do additional (external) consulting work. Any grants exceeding the EUR 50.000 funding gap would allow to accelerate product development and roll-out speed. 

(3) In the medium term, we need more highly trained engineers helping to market and implement the solutions for which we haven't budgeted funds as of now. 

We would like to apply for the GM Prize on community driven innovation, allowing social mobility for underrepresented community members, with an emphasis on STEM education.

We think our AI-based solution would fit the prize's purpose very well. We are currently developing the technology backbone and advance the prototype jointly with TU Vienna. Our financial runway for the next two years allows us to be fully operational, with the exception of a gap of EUR 50.000.

(1) The prize of USD 50.000 would therefore be very useful in essentially closing this funding gap, allowing a smooth development and roll-out process.

(2) Furthermore, Linda's costs are only partly covered by the grants we have received so far - she therefore needs to do additional (external) consulting work. Any grants exceeding the EUR 50.000 funding gap would allow to accelerate product development and roll-out speed.  

 (3) In the medium term, we need more highly trained engineers helping to market and implement the solutions for which we haven't budgeted funds as of now.

N/A

N/A

We understand that since we are based in Europe, Innospark Ventures would not be willing to provide additional funding. 

Since we believe though, that Innospark Ventures would be a tremendous fit to our own mission - since we are also poised to leverage AI solutions that can create disruptive impact in economy and society - notably in the realm of education. 

If selected, we would be honored for consideration for a deepened exchange.  

N/A

N/A