Most of the chatbots available on the market typically fall under one of the three categories: 1) “Small talk chatbots” pre-trained on massive dialogue data and capable to simulate social kind of pointless conversation; 2) “Personal Assistants” manually hardcoded to perform a variety of pre-programmed activities; 3) “Dialog-flow based” chatbots executing conversational templates manually engineered by people to solve specific problems. Here we offer another possibility — let the bot consume your data and start talking about this data to you and your customers immediately.
The Aigents® Personal Artificial Intelligence platform already provides chatbot capabilities such as conversational intelligence based on dialogue templates, custom Web search, sentiment analysis, monitoring the Web and group chats, and computation of reputation in communities. The Aigents release 3.0.7 couples semantic graph search technology with the concept of controlled language under the chatbot interface. It enables a chatbot user to navigate through the existing data finding objects of interest just using chatting with the Aigents bot.
The use cases are any wherever one needs to expose the data to personnel of a customer base or a wider audience so the elements of the knowledge may be searched and found by the searchers precisely, quickly, and interactively. It could location of a good in a store or a departure gate in the airport or a book in the library or a person in an enterprise or a specific service company office in the office building or a support case in the CRM database — anything with multiple attributed data items like stored in SQL database and can be dumped to a tab-separated or a comma-separated file.
The newest technology involves few steps, each of them quite simple on itself, all coupled under the Aigents® Personal Artificial Intelligence platform architecture.
- A dedicated Aigents server is hosted in the cloud or corporate server or a personal desktop, following the respective documentation pages. Note that even an Android smartphone or tablet can be used for the purpose in case if you want it to be available on Telegram only.
- The schema of the data is defined using a controlled language, like saying “there name restaurant, trust true, has address, city, country, province” in Aigents Language requests the creation of an abstract thing called restaurant with attributes called address, city, country, province. The extra trust true clause makes the statement trusted by the admin user so the thing is not forgotten by garbage collection eventually. Each such abstract thing can be thought of as a class in terms of object-oriented programming or a table in terms of the SQL database. Note that there is a name attribute implied for every “thing”, so there is no need to define it explicitly.
- The data is loaded by the load command such as the data about the fast-food restaurants in the US available from the Kaggle Datafinity project under non-commercial license may be downloaded as “load file http://aigents.com/download/data/us-restraunt-data.csv as restaurant”. The command uploads the data to the internal semantic graph (knowledge graph in modern language) being executed by your Aigents server.
- You ask any questions referring to any attributes of your data to the Aigents bot using either Telegram chatbot, or Facebook Messenger chatbot, or Web chatbot as you have deployed and configured them. The Aigents module called Finder implementing the Intenter interface in the Aigents architecture keeps the context of your interaction context against the vector space of your data items in the dimensions of its attributes and figures out the right answers to ask you, so your chat-based search leads you to the target of your exploration as soon as possible.
Caveat: We don’t recommend to to use https://aigents.com/ to load your own data unless you are doing it under commercial subscription. The data uploaded under free subscription may be removed by garbage collection policy or because of the server administration reasons.
Encouragement: To run your own Aigents but serving your own data, deploy your own Aigents server on your own premises or subscribe to commercial service on https://aigents.com/. In the former case, you will have your own chatbot serving your data under your own branding and your proprietary chatbots across the messengers. In the latter case, your data will be served by @AigentsBot on Telegram, Aigents bot on Facebook Messenger, and Aigents Web Chat.
Important: The functionality above is provided by the extensible Aigents Conversationer/Responser/Intenter architecture, featuring controlled language, conversational intelligence based on dialogue templates, custom Web search, sentiment analysis, monitoring the Web and group chats, computation of reputation in communities, and now the semantic graph search in a chatbot. Future extensions to the architecture are coming, including better support for understandable natural language processing. We welcome partners and contributors to add their own extensions to the platform.
Feedback may be directed to the following channels: