About

Schemes SG started as a side project by our product lead. A long-time volunteer with various VWOs, he collated a "help-list" to facilitate referral work and built a quick front-end to share these resources with his friends. The resource gained unexpected traction with social workers and volunteers. Sensing that a consolidated directory could address care workers' pain point of having to navigate the confusing social assistance landscape, he gathered like-minded individuals from friends and the better.sg tech community to improve the tool.

The team engaged social workers, caregivers and friends to understand lived experiences and help-seeking practices. They found that social needs are often intertwined, and if technology could improve the search process by making sense of the entangled issues that people face, it would immensely alleviate mental burden faced by help-seekers and professionals. This inspired our natural language search tool and tagging system.

Schemes SG is ultimately what we hope to build for the community, with the community. Your continued usage, feedback, searches and contribution of crowdsourced assistance listings help this tool get better everyday.

Serious thought was put in before building Schemes SG. Initial landscape scans led to the realisation that:

1. Social assistance listings were piecemeal and information was fragmented across various sites. There were some compilations, but they were often PDF files hidden within the repositories of organisations' websites, so they might not be easy to find. Search engines might also miss them.

2. Even if one could get their hands on a compilation, it would take a million "Ctrl + F"s and painstaking excavation to find schemes, given how complex social assistance is. The volume of information was simply mind-boggling.

3. The listings might not necessarily be updated. New versions were usually held in completely different links. PDF listings also meant that social workers and volunteers had to depend on the original poster to issue a new version should there be changes.

4. There were actually intuitive directories (e.g. SupportGoWhere has done a great job), but they were primarily government portals and might not include NGO or VWO schemes. Again, the power of the crowdsourcing could be useful here, given the size of the non-profit sector.

This portal hopes to address the above issues by tapping on the power of the crowd to make social assistance info 1) comprehensive and 2) updated, and then using technologies like AI/NLP and filters in data visualisation to make this info 3) navigable. 😊

Here are the parameters governing how the Bank was populated:

1. All information is public-domain. Schemes SG only agglomerates public info to help navigate complexity. Where individual schemes are concerned, we use the descriptions from the organisations' writeups wherever possible to let them speak for their own good work :) If we make edits, it is to improve search functionality, and we ensure that they are factually accurate.

2. Currently, Schemes SG only lists schemes that provide benefits in cash (financial assistance, subsidies) or in kind (free food, food vouchers, free clinics, special cards which ascribe certain benefits). We are just starting to include services (e.g. subsidised special education) as our team grows in capacity.

Schemes SG does not include:

1. Auto-inclusion schemes. The purpose of a public aid portal is to help reduce bandwidth tax, so we see no need to put in extra information that social workers and volunteers have no scope to act on.

2. Schemes that do not have a public listing or are not verified. We understand that sometimes organisations may have reasons for keeping their assistance informal. Hence, if there is no public info on it, we will not include it. If the info is crowdsourced, we ask the contributor for a link. If there is none, we do our own research to populate the info.

Schemes Pal's natural language model involves the following transformation: Bag of Words (BoW) -> TF-IDF -> latent semantic indexing (LSI). Some resources used include this, this and this. We are still improving the natural language feature, and if you have engineering expertise or insights to offer, reach out via the "Feedback" form.

Our steady-state vision is that as the user base grows, we get more Schemes Bank contributions and Schemes Pal queries, allowing us to train more robust and accurate semantic matches. Schemes Case, our volunteer service, will cover the blind spots of the model. The three components work in tandem to create an ever-improving, ever more robust Schemes SG.