The TBFY project partners held their first hackathon at the prestigious Alan Turing Institute in London in early April 2019. The hackathon was run by a wide range of leading academics from the University of Southampton, the University of Manchester and the University of Oxford, among others. Spend Network provided access to hundreds of thousands of tender notices from the UK and overseas to give researchers the opportunity to explore ways of matching businesses to government contract opportunities. It is one of TBFY’s key objectives to make the procurement market more accessible and more transparent, which made this a particularly exciting event.
The project partners invited programming wizards with skills in text mining, natural language processing, recommender systems, modelling with structured data records, semi-supervised learning, dashboard visualization and familiarity with EU or UK tenders to this one-week event in central London.
The approach that the team decided to pursue consisted of building vector representations of suppliers and tenders to identify their compatibility with the distance between the respective vectors. In building their representations, our data specialists made use of both their textual descriptions and knowledge of previously awarded contracts.
One of the key outcomes of the experiments over the week is that previous contracts are informative of future procurement decisions. This demonstrates just how important transparency and openness are for the tender and procurement process. The best results of the week-long efforts used correlated topic models for extracting representations of textual descriptions.
Yuchen Zhao from the University of Southampton represented the TBFY community at this hackathon and contributed a full report for the Alan Turing Institute which can be found here.