Technology

Natural Language Processing (NLP)

Whether a Data Partner curates job opportunities, mental health supports, or anything in between, the records in their data set contain natural language text describing them. Using a large language model (LLM), CORDS leverages artificial intelligence (AI) to uniformly encode each record’s text.

When someone uses CORDS, they also enter their query as natural language text (e.g., “I’m looking for free food”). The same LLM is used to encode that query.

By producing a common representation, CORDS can find the best matches for a user’s needs from different sets of opportunities and resources curated by different Data Partners, each of which boasts expertise in different domains.

Location awareness

Opportunity and resource records in CORDS include information about the location or region where they are available. When a user’s location is specified, CORDS serves results in and around that location, including in-person and virtual services.

Search location is under the user’s control. CORDS initially determines a user’s location from their IP address. Users may also provide their precise location to CORDS by enabling their device’s location services, or they can specify a search location explicitly by using an interactive map.

Results can be further refined to include any combination of local, regional, provincial, or national options.

When a user clicks to view a particular result, CORDS presents similar options, as well as recommendations for related supports from other sectors that might be useful on their journey.

Recommendation Engine

When a user chooses to see the details of one the results from a search, they are also offered two additional sets of results related to that record:

Similar opportunities or resources.
Things other people who entered similar queries have viewed (recommendations).

Whereas similar records tend to be from the same domain (e.g., if the user is looking at a job posting, similar records are likely to include other job postings like that one), recommendations often contain records from different domains. For instance, a user viewing the details of a record for a local shelter may receive a recommendation for a local “hot meals” service.

Currently, recommendations are implemented with a set of expert selected rules, and from referral information provided by Data Partners.

No matter how a user lands on the website for a given opportunity or resource, CORDS aims to make that the right door to have virtually walked through. By enabling CORDS on those websites, users can receive recommendations from CORDS to continue their service journey through a network of vetted opportunities and resources.

Community Uptake Tools (CUT)

CORDS utilizes a “No Wrong Door” approach to youth services. This means that any organization can implement the CORDS search tool on their website. CORDS enabled websites form the CORDS Community Partner Network. Once a user begins searching for services, CORDS can follow them across any CORDS enabled website within the network. Through their journey, CORDS will continue to offer recommendations based on what the user is currently viewing (the “context” in CORDS).

Community Uptake Tools (CUT) are available to support CORDS integration on any website, enabling the “No Wrong Door” approach envisioned. The CUT is currently available for adoption via API, a TypeScript SDK, and an alpha (for the brave of heart) WordPress plugin.

Users can start their search process from any website or app that integrates CORDS. Throughout each user journey, CORDS will collect anonymous navigation data to train and improve its recommendations.

Realtime needs data

Anonymous user navigation data is collected wherever CORDS is enabled. The CORDS Partner Portal allows those who have built CORDS into a website or app to access the queries, query locations, and query dates and times associated with traffic through their CORDS enabled system.