autosistant | A research project aimed at building an automated sales assistant

autosistant is a Ruby library. autosistant has no bugs, it has no vulnerabilities, it has a Strong Copyleft License and it has low support. You can download it from GitHub.

The system attempts to help online shoppers buy items from an online store by having a directed conversation with the user. The user interacts with the system via a web-page where the user can type in a message to be sent to the system. The system gets the message, analyses it, and returns an appropriate message when complete. The system’s message analyses begins by splitting the input message into distinct words. It will then attempt to find any words that indicate actions, also called tasks, that the system must complete. When such a word is found, the system adds the task to a task queue. This queue is always kept sorted based on task priority. If the system has no tasks to complete and the most recent message processed does not exactly tell the system what it needs to do, then the system will attempt to suggest what the user might have meant by measuring edit distance of each input word with each task determining word. If the system has some number of tasks it is managing, it will treat all words not mapped to actions as input for the task it will complete next. Once the system has at least one task to complete, on each conversation turn, the system attempts at least one task on the queue. This may lead to task completion and then task removal and/or the adding of further tasks to complete based on user input. Tasks include product identification, adding identified products to the order, placing an order, suspending a session, and resuming a session. Additionally, there is a system administration page. This allows retailers to modify the system to meet their needs. Specifically, administrators can configure system settings such as the company name, identification questions, answers to identification questions for specific products, and which words map to the various system actions. Additionally, if the system administrator has knowledge of the relational model, the system’s identification path and order process can be modified. Future plans to implement this configuration in a more user-friendly way are a possibility (post-project).