Research Projects

In the Indian context, one of the crucial requirements is to have an automatic travel assistant, that can assist the commuters and tourists in planning their travel as well as obtaining recommendations. This would include use cases such as: (1) A villager trying to find a way around in a city, local language support becomes essential (2) English-speaking city dweller, but needs to work around civic constructions, road blockages in the city (3) for law-enforcement agents to get real-time info about various traffic situations and bottlenecks. The disruptions can be either pre-announced (known to law-enforcement agency) or it may be sudden (can again be gathered from social networks). Increasingly ride-sharing is emerging as an economical but private way to drive. One of the main impediments behind ride-sharing is the apprehension commuters have about sharing space with strangers. With the support of background social networks which generate real-time information, the agent can provide a pleasurable social environment for ride-sharing. Summing up, the information upon which chatbots will act will come from first a slow-changing database that stores information about specific routes, sites as well as ratings of people. This would be supplemented with more real-time unstructured information – an important task would be to structure them and make them useful. This would be possible if we can build up domain-specific knowledge graphs. It is aimed to build a neural conversational system, which is assisted by domain knowledge, in the form of a knowledge graph. Another important aspect of the Chabot is its ability to interact with the subject in regional languages. In order to have the interface as natural as possible, investigators plan to make it operational in two regional languages (Hindi, Bengali) as well as its corresponding code-mixing version. The entire solution will be deployed within a city and extensive experimentations would be done to understand the various nuances attached