Our teams create individual bots for themselves and now we want to integrate them into the Enterprise Assistant.
What I need help with is Multi Bot Orchestration.
Has anyone been able to do this efficiently?
Basically, the user asks a question in the Enterprise Assistant and then the bot gets the answer from the respective child/skill QnA bot.
I am able to add skills like calendar, people, SAP, etc. but dealing with QnA bots is proving to be an impossible challenge.
To create a multi bot orchestration, we need to include multiple levels of dispatchers, pipelines for exception handling and fallbacks. It needs an expert system type agent who can handle multiple NLP models. A chat bot is designed based on NLP modelling. The NLP implementation will be depending on number of operations that enterprise application will handle. The following are the different NLP models required to handle a perfect enterprise application.
General QnA
Billing information
Calendar
Leave Planning
Employee storage disk
On demand service to the clients
Fault tolerance
Saving the QnA pattern of chat between bot and human
Calculating the accuracy
Taking users own request and performing NLP
like this we have so many requirements to implement a Bot. All the models will be trained and connected to the dispatcher.
The following are the things followed in bot orchestration.
It consists of 5 different implementations:
Ingress Pipeline
Egress Pipeline
Dispatcher
Exception pipeline
Fallback pipeline
Ingress Pipeline:
The ingress pipeline will be detection few things like
Language detect and translate
Entity recognition
User input redaction
Egress Pipeline:
the following are requirement and operations in egress pipeline
3rd party tools for analytics
Managed responses (default responses)
Data service redaction
Dispatcher:
The dispatcher will be having two different implementations.
Policies and rules,
NLU (Natural Language Understanding) strategy.
Policies and Rules
The policies and rules of the organization must be trained to the model. Because, if the user is asking for the financial status of the orgzanition, the bot must understand the policies and rules it must follow in order to answer for the question.
Natual Language Understanding
Natural Language Understanding (NLU) — The computer’s ability to understand what we say. The context what user is typing in his/her own words is the procedure of NLU.
The dispatcher will be communicating with the egress pipeline and ingress pipeline as it is the parent handler of all the operations and model training results.
To address the exceptions and rollbacks, we need to have exception handlers who are also expert system agents. An agent is having some expertise in handling the exceptions.
Example:
Question: Give the details of pithon projects of this monthh.
Handled Exceptions: In the above question "python" spelt like "pithon" and "month" was spelt like "monthh". NLP and NLU have to understand these cluases and have to handle in exception of mis-spelt words.
The following link can explain in detailed architecture of Multi-bot orchestration.
https://servisbot.com/multi-bot-orchestration-architecture/
Related
I am working on a chatbot, I have implemented it with Dialogflow (Dialogflow ES). I found that Dialogflow has the following Pros
Easy to use
Good at Intent classification
Good at extracting Entities (prebuilt/custom)
Conversations can be chained to a certain extent using input/output contexts and lifespan
But in my use case, there are certain situations where human level judgment is required and it cannot be done using Dialogflow. Can we add our custom logic to process certain user requests in Dialogflow or any other chatbot framework which provide more flexibility?
You're a bit vague what you mean by "custom logic", but this sounds like fulfillment is what you're looking for.
With this, you can enable Intents so they send JSON to code that you run (either by a webhook you run or via some deployed through an inline editor which manages the deployment for you). Your code can apply your business logic to determine what the response might be, including what replies to send, what Output Contexts are set, and any parameters that are in those Contexts.
I am trying to understand the different model in building a bot using dialogflow and came across this 2 methods.
Fulfillment model (with webhook enabled) documentation here
API Interactions documentation here
I understand that both of this models have their own pros and cons, and I understand how they both work. Most online examples are showing the fulfillment method (I guess that's more common?)
However, I would still like to ask what reason will it be to choose one or the other? If anyone had used either model before, what limitations are there?
p/s: I've look through quite a number of tutorials, and read through the dialogflow documentation.
the integration by fulfillment is indeed the default approach because you use DialogFlow to design your conversation flow and (big bonus) manage the integration with the various channels (ie Telegram, Facebook).
It is the easiest way to design a fully fledge conversation, you only need to worry about the post hooks that are sent to your backend to either save the data or alter the conversation (add contexts or trigger events).
Important remark: all user traffic (who says what) goes via Dialogflow cloud
The API interaction becomes a good option when you have already an existing frontend (say an existing application or web sites) and you want to plug in DialogFlow NLP capabilities.
I have done something like that to create a FAQ chatbot that called DialogFlow to identify which intent would match a certain phrase while the BOT was deployed in MS Teams.
The architecture would indeed look like the one in the documentation: MS Team ecosystem is the "End-User" part, then my Java app ("Your System") would use the API to call DialogFlow.
Important remark: only given statements (the ones you send) go to Dialogflow cloud
Is it true Azure Luis only support up to 500 intents per application? https://learn.microsoft.com/en-us/azure/cognitive-services/luis/luis-limits
My requirement is more than 1000 intents. How can I use Luis to do that?
You should consider using Dispatch. It is a tool that was designed specifically for managing multiple LUIS models and/or QnA Maker knowledge bases that a bot needs to access.
You can find C#, Javascript, and Python samples on the BotBuilder-Samples repo, for reference, titled "14.nlp-with-dispatch".
In your case, this tool is provides a means for overcoming LUIS intent limitations by allowing you to create multiple models to draw from. Dispatch negotiates these models by creating a single LUIS app that then routes the requests to the appropriate model.
Hope of help!
To answer your question, yes it is true. But it is plenty. We have some of the biggest company using our platfomr to test their training data on LUIS/WATSON/DF... and it is extremely rare to pass the 500 and still get top performance.
We typically advice anyway to fine tune your training data for 200 or 300 intents max and if you have more, look into a Model controller architecture with several slave (specific) models
So you might be sure you have 1000 intents, can you reduce it with Entity?
I tried to find some Azure service similar to IBM Watson Knowledge Studio but I failed. I'm looking for something I can train to analyze texts and retrieving entities, relations between entities and entities related sentiments.
Do you know if there is anything in Azure I could use to do that?
Yes, there is remote similarity between Azure Language Understanding (LUIS) and IBM Watson Knowledge Studio (WKS), but they are not substitutes.
The notable differences:
LUIS is for building chatbots - conversations with utterance data. WKS is for general unstructured text, usually much larger in volume than utterances. In this respect, LUIS is competing with IBM Conversation service, not with WKS and the IBM Watson services that run the WKS custom models - Watson Natural Language Understanding and Watson Discovery.
Because LUIS is built for processing utterances, it has much lower limits, compared to WKS. For example, LUIS limits the input text to 500 characters, while WKS processes text of up to 40,000 characters. LUIS also limits the Simple entities to 30, which may be ok for processing targeted utterances, but not for building high quality model for processing large documents and complex domains.
LUIS supports only customization of Entity Type mentions (in various forms, like Simple, Hierarchical, Composite, RegEx, List,.. - very similar to the WKS Entity types). WKS (and the runtime services that use the WKS custom models), on the other hand, supports Entity Relations - an important feature that helps you extract insights from the client-specific corpus that you cannot do with Entity mentions alone.
LUIS supports only a fraction of the languages that WKS supports. And the LUIS language support is partial - see https://learn.microsoft.com/en-us/azure/cognitive-services/luis/luis-supported-languages
LUIS, similarly to IBM Conversation service, is a runtime NLP service that allows customization in its tooling. WKS, on the other hand, is a stand alone customization SaaS offering, that was specifically designed to organize a team of domain subject matter experts (SME) and cognitive solution developers to transfer the SME domain knowledge into the custom model that is then deployed to and used by IBM Watson runtime services, like Natural Language Understanding and Discovery. In other words, while LUIS and IBM Conversion provide tooling for customizing the solution directly, WKS provides a separate environment with built-in methodology for managing customization projects and for annotation skill building.
LUIS, to my understanding, is offered as a multi-tenant public service. WKS is offered in both multi-tenant and isolated configurations. In that respect, WKS is suitable not only for the general public, but also for projects with sensitive client data.
In conclusion, there's no WKS equivalent (substitute) that I'm aware of. LUIS may be considered as the Azure alternative to IBM Conversation service, if your solution is built in Azure, but LUIS is not a substitute for IBM Watson Knowledge Studio.
So, it's very important to consider your use case (application domain), when choosing on which platform to build your solution.
Hope this helps.
Did you look at the Azure AI gallery? One common approach is to customise the solutions there for your particular requirements. Here's a search of all text-related items, which you can for example refine to be Microsoft content only.
I'm not aware of a single service that maps directly; the Text Analytics API for example just does language and phrase detection and sentiment analysis.
Have a look at Luis.ai...should do what you need.
I want to build a chatbot for a customer service application. I tried SaaS services like Wit.Ai, Motion.Ai, Api.Ai, LUIS.ai etc. These cognitive services find the "intent" and "entities" when trained with the typical interactions model.
I need to build chatbot for on-premise solution, without using any of these SaaS services.
e.g Typical conversation would be as following -
Can you book me a ticket?
Is my ticket booked?
What is the status of my booking BK02?
I want to cancel the booking BK02.
Book the tickets
StandFord NLP toolkit looks promising but there are licensing constraints. Hence I started experimenting with the OpenNLP. I assume, there are two OpenNLP tasks involved -
Use 'Document Categorizer' to find out the intent
Use 'Named Entity Recognition' to find out entities
Once the context is identified, I will call my application APIS to build the response.
Is it a right approach?
How good OpenNLP is in parsing the text?
Can I use Facebook FASTTEXT library for Intent identification?
Is there any other open source library which can be helpful in building the BOT?
Will "SyntaxNet" be useful for my adventure?
I prefer to do this in Java. BUT open to node or python solution too.
PS - I am new to NLP.
Have a look at this. It says it is an Open-source language understanding for bots and a drop-in replacement for popular NLP tools like wit.ai, api.ai or LUIS
https://rasa.ai/
Have a look at my other answer for a plan of attack when using Luis.ai:
Creating an API for LUIS.AI or using .JSON files in order to train the bot for non-technical users
In short use Luis.ai and setup some intents, start with one or two and train it based on your domain. I am using asp.net to call the Cognitive Service API as outlined above. Then customize the response via some JQuery...you could search a list of your rules in a javascript array when each intent or action is raised by the response from Luis.
If your Bot is english based, then I would use OpenNLP's sentence parser to dump the customer input into a database (I do this today). I then use the OpenNLP tokenizer and push the keywords (less the stop words) and Parts of Speech into a database table for keyword analysis. I have a custom Sentiment model built for OpenNLP that will tag each sentence with a Pos, Neg, Neutral sentiment...You can then use this to identify negative customer service feedback. To build your own Sentiment model have a look at SentiWord.net and download their domain agnostic data file to build and train an OpenNLP model or have a look at this Node version...
https://www.npmjs.com/package/sentiword
Hope that helps.
I'd definitely recommend Rasa, it's great for your use case, working on-premise easily, handling intents and entities for you and on top of that it has a friendly community too.
Check out my repo for an example of how to build a chatbot with Rasa that interacts with a simple database: https://github.com/nmstoker/lockebot
I tried RASA, But one glitch I found there was the inability of Rasa to answer unmatched/untrained user texts.
Now, I'm using ChatterBot and I'm totally in love with it.
Use "ChatterBot", and host it locally using - 'flask-chatterbot-master"
Links:
ChatterBot Installation: https://chatterbot.readthedocs.io/en/stable/setup.html
Host Locally using - flask-chatterbot-master: https://github.com/chamkank/flask-chatterbot
Cheers,
Ratnakar
With the help of the RASA and Botkit framework we can build the onpremise chatbot and the NLP engine for any channel. Please follow this link for End to End steps on building the same. An awsome blog that helped me to create a one for my office
https://creospiders.blogspot.com/2018/03/complete-on-premise-and-fully.html
First of all any chatbot is going to be the program that runs along with the NLP, Its the NLP that brings the knowledge to the chatbot. NLP lies on the hands of the Machine learning techniques.
There are few reasons why the on premise chatbots are less.
We need to build the infrastructure
We need to train the model often
But using the cloud based NLP may not provide the data privacy and security and also the flexibility of including my business logic is very less.
All together going to the on premise or on cloud is based on the needs and the use case of the requirements.
How ever please refer this link for end to end knowledge on building the chatbot on premise with very few steps and easily and fully customisable.
Complete On-Premise and Fully Customisable Chat Bot - Part 1 - Overview
Complete On-Premise and Fully Customisable Chat Bot - Part 2 - Agent Building Using Botkit
Complete On-Premise and Fully Customisable Chat Bot - Part 3 - Communicating to the Agent that has been built
Complete On-Premise and Fully Customisable Chat Bot - Part 4 - Integrating the Natural Language Processor NLP
Disclaimer: I am the author of this package.
Abodit NLP (https://nlp.abodit.com) can do what you want but it's .NET only at present.
In particular you can easily connect it to databases and can provide custom Tokens that are queries against a database. It's all strongly-typed and adding new rules is as easy as adding a method in C#.
It's also particularly adept at turning date time expressions into queries. For example "next month on a Thursday after 4pm" becomes ((((DatePart(year,[DATEFIELD])=2019) AND (DatePart(month,[DATEFIELD])=7)) AND (DatePart(dw,[DATEFIELD])=4)) AND DatePart(hour,[DATEFIELD])>=16)