The predefined response in each intent support to use mask in order to fill variables saved in dialog status. In each turn the chatbot engine will do the semantic search for the most closed user says in the search scope, and then call the defined function. The entity mask provides additional control for intent search range, and with defined child id, the intent search scope will be limited for next loop dialog. In the rule definition, each intent has multiple user_says, which indicates to some possible query with same meaning and multiple response for randomly return. With a simple designed rule interface, user can easily define their dialog flows. The rule-retrieval based chatbot engine reads predefined rules from an excel speadsheet. Currently ElsaBot includes several engines: rule retrieval based, end2end goal oriented chatbot engine, controllable generative based chatbot engine, knowledge based chatbot, read comprehension QA from raw texts. Chatbot enginesĮach chatbot engine uses the same input (dialog status) and output that can easily be mixed via topic manager. Specifically, the topic manager in our solution can either be a text classifier for deciding whether switch topics, or as simple as sequential decide which engine will be used based on the score returned from each engine. Here topic manager is able to understand users’ inquires and allocate them to different domains. Topic manager is to identify topics via conversion and identify human behavior of changing topics. Topic ManagerĮlsabot aims to process different domains at the same time, and it mixes several chatbot engines for different purpose. The dialog status will then be used as input for further chatbot engines. The dialog status saves any variables generated in a whole dialog session, besides the variables mentioned above, it also saves variables got from function call. After all, the intermediate data, tokens, sentiment, entities, will be saved to a global register: dialog status. A sentiment analyzer is used to get the sentiment of query. When a user query come, the query will be tokenized and spell checked, the related entities will also be extracted. Elsa chatbot can be embedded into several different business scenarios, such like customer service, recommondation, investment assistant, risk alert, system monitor. An easy-to-use dialog-design system can also be used for designing to do a custom job. Elsa chatbot not only allows to do several different types of jobs at a time, but also allows to reply a human-like emotional response. It allows people to interact with each backend service via natural language. Contact me if you are interested ^_^ĮlsaBot is a general purpose chatbot as a human-machine dialogue interface for other services. It allows you to have a smart and customized chatbot in a very short time. I developed a chatbot that can be easily integrated into your system. Virtual agents are most effective in customer service applications in service-heavy industries like financial services, retail, travel, and telecom. ![]() ![]() Chatbots - automated conversation systems - have become increasingly sophisticated. If you’ve ever used a customer support livechat service, you’ve probably experienced that vague, sneaking suspicion that the “person” you’re chatting with might actually be a robot.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |