The PeARS app
Building low-resource Web search applications from cognitive models
It is widely believed that Web search engines require immense resources to operate, making it impossible for individuals to explore alternatives to the dominant information retrieval paradigms. The PeARS project aims at changing this view by providing search tools that can be used by anyone to index and share Web content on specific topics. The focus is specifically on designing algorithms that will run on entry-level hardware, producing compact but semantically rich representations of Web documents. In this project, we will use a cognitively-inspired algorithm to produce queryable representations of Web pages in a highly efficient and transparent manner. The proposed algorithm is a hashing function inspired by the olfactory system of the fruit fly, which has already been used in other computer science applications and is recognised for its simplicity and high efficiency. We will implement and evaluate the algorithm on the task of document retrieval. It will then be integrated into a Web application aimed at supporting the growing practice of 'digital gardening', allowing users to research and categorise Web content related to their interests, without requiring access to centralised search engines.
- The project's own website: https://pearsproject.org/
Why does this actually matter to end users?
We have come to associate search and discovery of digital content with online search engines. Somewhere on the planet there is an army of all-knowing machines waiting day and night for our inquiries, ready to point us to wherever we need to be - if we ask them nicely. However, this tremendous luxury comes with quite a heavy real-time dependency for internet users: it requires us to have an active connection to the internet whenever we need to find something. As our use of the internet has become more nomadic over the years due to the rise of mobile phones, there are in fact many situations that we find ourselves in where our use of the internet is very restricted or even temporarily cut off. Like when you are on a train, in a busy city centre where the wifi is completely saturated, in a remote area with limited coverage, or when you've ran out of your monthly mobile data plan. Or something more serious, when the network is offline for a longer time due to a cascading network failure or cyberattack.
All of a sudden, we are at a loss. It feels we are thrown back in time. We cannot find anything anymore outside of the files and documents we have stored on our devices. Our on-line search engines are all out of reach and of no use to us. Our many questions will have to wait: there is nothing we can do until we get back online. Such a real-time dependency on a critical resource is not only annoying for users (and sometimes downright disadvantageous when you really need to look up something like a manual or an important reference document). It is also not necessary. There are other, more efficient ways to approach web search, which may even provide you with richer results. This project takes a unique approach of searching and indexing data in documents to allow people to research and categorize in 'digital gardens': instead of having to wade through (irrelevant) data, you search only content relevant to your interests, making search a personal experience instead of a confusing and commercialized pain.
Run by University of Trento
This project was funded through the NGI0 Discovery Fund, a fund established by NLnet with financial support from the European Commission's Next Generation Internet programme, under the aegis of DG Communications Networks, Content and Technology under grant agreement No 825322.
