In recent times, Reciprocal recommender systems (RRSs), which recommends users to each other, have gained immense importance in various online platforms for connecting people in a personified manner, such as online matchmaking, matrimony, recruitment, social connections, online tuition's etc.

Generating successful recommendations in traditional item-item recommendation systems is challenging as the system needs to balance two objectives:

(1) recommended users with whom the requester user is likely to instigate an interaction and

(2) recommended users who are likely to reply affirmatively to the requester user - initiated interaction.

These objectives seem to be partially conflicting in nature since many times it happens that the contacted user do not respond positively to the requester user and vice versa. Furthermore, users on these platforms differ in the way, they regard the other side’s preferences before initiating an interaction. Therefore, an effective recommender system must efficiently model each user and balance these conflicting objectives.

In the Reciprocal recommender system, this challenge is well tackled through two components:

(1) an explanation module, which leverages an estimate of why the recommended user is likely to respond positively to the requester user and

(2) a reciprocal recommendation algorithm, which finds an optimal balance, individually tailored to each user.

Hence, unlike traditional item-to-user recommenders, for e.g., amazon.com, where based on user shopping history, new items are recommended to users for purchase , a fundamental requirement in reciprocal recommendation is that both parties, namely the requester user and the recommended user, must be satisfied with the "user match" recommendation in order to declare it as successful.

A much higher coherent matching is achieved with the use of Reciprocal recommender systems, and this has proven to offer maximum benefits for users of online matchmaking or matrimonial systems.

Formal definition of Reciprocal Recommender Systems (RRSs)

Reciprocal Recommender Systems (RRSs) denote a class of recommender systems which recommend people to people as opposed to traditional recommender systems which recommend items to people.

Success criteria -Traditional Recommender Systems (TRSs) Vs Reciprocal Recommender Systems(RRSs)

In item-to-people recommendations, the success of the recommendation is normally determined by the acceptance of the recommendations (items) by the receiver. For example, a recommendation in an online shopping platform would be considered successful if it gets converted into purchase from the service user/receiver. But, in RRSs, a successful recommendation is one that brings about a successful interaction between the two users, meaning that both the requestor user accepted the recommendation and initiated an interaction with the recommended user, and, most importantly, the recommended user also replied positively. In an online-matchmaking or any matrimonial platform, this means that the requested user has expressed interest in the recommended user (e.g., by sending a message) and the recommended user has expressed interest on their end as well (e.g., by replying with an affirmative message).

Architecture - Traditional Recommendation systems(TRSs) and Reciprocal Recommendation systems(RRSs)

In TRSs, a content filtering method is used to match heterosexual users which can be represented as F x M contact matrix, in which a user is a node. An edge(link between nodes) in the matchmaking network always connects a male and a female. In this method, if user A is interested in user B, it can approach B by sending a message or an initial contact and this initial contact is used to create a recommendation for a user based on one user preference only, not considering the response of the other user.

Further this content- based filtering model was improved to consider both preferences before a recommendation was made. It uses the same contact network as the content-based recommender, but its result produces a different matrix notation showing only reciprocal contacts which are considered for recommendation.

High Level Design of the Reciprocal Recommendation systems(RRSs)

A user interacts with the system, registers a profile, provides user data and a match algorithm is implemented to produce an output of an optimal match. The match algorithm is the algorithm that performs the actual matching after the data set has been processed, classified and the match criteria obtained.

Advantages of Reciprocal Recommendation systems(RRSs)

1. Combining two matching criteria provides more clear judgment for people-people recommendations.

2. Users have a wider range of options to choose from to get their match based on two outcomes of the matches.

3. The proposed system re-assures the users of the matches they receive, as all outputs are displayed and the differences in outputs are made clear convincing the user of the match output.

4. With this new system, users of matching system are given a higher sense of satisfaction as their matches are not based on guesses but scientific algorithms which reduces the chances of errors.

5. This proposed system reveals more possibilities to the users in the application of recommender systems and can be used in sales websites like amazon as well.

Conclusion

There is a wide range of applications for matchmaking in science. Further Study can also be done to introduce matchmaking systems to institutes of learning to match Lecturers with students and courses, in companies to pair trainee workers with their training programs and even in health sector to match specialist doctors to patients.

https://www.sciencedirect.com/science/article/abs/pii/S1566253520304267 and https://link.springer.com/article/10.1007/s11257-020-09279-z