Sentiment Analysis Archives | Comidor Platform

AI/ML Application Cases

Comidor team — Wed, 23 Dec 2020 16:21:59 +0000

Artificial Intelligence (AI) in BPM is ideal in complicated situations where huge data volumes are involved and humans need to make decisions. Machine learning is the part of artificial intelligence (AI) that focuses on the development of computer programs that can access data and use it to learn for themselves.

Comidor platform offers the ability to build your own Low-Code App through App Builder, and include both AI and ML components, in order to determine the writer’s attitude, get predictions, classify text and enhance digital process automation.

In this article, we will give two AI/ML application cases of real business problems where we have included AI and ML in the solution.

Case 1. Loan approval process

Business Problem

A loan approval process starts when a potential borrower reaches out to the organisation. The first-phase employee should input all customer details and check the customer’s creditworthiness. In the next phase, a second-level employee should review all data and decide whether to approve or reject the loan request, which might be demanding, of high-risk, and time-consuming especially for a new employee.

There was not a central system that could handle and manage all loan requests and process steps. The main need in this case was to enhance the accuracy of the decision-making process.

The solution

As a solution to the above, Comidor offers a Low-Code application to monitor all Loan approval processes in one place along with a workflow that orchestrates all process steps.

In the workflow we have included:

Public form for process initiation by the potential borrower outside of the Comidor environment
Task allocation to the responsible users and groups
ML Predictive Model that predicts the loan approval decision based on historic data and variables such as the annual salary and credit score of the borrower

User forms & fields for data input and display
Gateways and conditions for path determining
Automated emails

The Loan approval process steps in detail are:

The loan request process is triggered by the customer on their personal web banking portal, with Comidor embedded public forms. The customer adds personal details and the loan information, and selects the type of loan and loan interest.
The first-phase employee is notified about the new Loan request, reviews it and adds further information (Credit score)
Based on the predefined range of variables in the loan process and historical data on the approval process, the Comidor ML Predictive model provides an instant, high-confidence
suggested decision.
Then, the next-level employee is informed about the loan request and the available ML prediction. The employee can then take the final approval/rejection decision.
Finally, the customer receives an automated email with the final decision about the loan request.

What we achieved:

Big data analysis
Robust credit decisions within minutes
Automation of the loan request process
Pattern identification
Human error elimination
Improved and faster risk assessment
Customer-Self service

Case 2. Customer request management

Business Problem

The Customer request management process starts when a new customer need rises. In this case, there are 4 types of customer requests: individual, corporate, support and complaint.
There was a lack of one central channel of communication between the company and its customers. Resolution time could take too long due to the huge volume of requests and therefore, complaints were increased.

The solution

For this business problem, the solution is a Low-Code application to monitor all Customer requests in one place, along with a reporting dashboard. A workflow that orchestrates all process steps is also included.

In the workflow we have included:

A public form allowing non Comidor users to trigger internal processes
Automated emails with process details
ML text classification model that assists in request categorisation
AI Sentiment analysis that analyses customer’s sentiment
Scripts to change the priority of the request upon certain conditions
Task allocation to the responsible users and groups
User forms & fields for data input and display
Gateways, conditions for path determining, and loops
Timer for auto-closing the process after a certain period of time

1. Customer request initiation

We have added a Comidor public form to our client’s website so as to allow non Comidor users to trigger Customer request processes. The public form is an embedded form similar to the initiation quick add form in Comidor, including all user fields and business rules such as customer request details. Once the customer completes the public form, a new process starts in Comidor.
Alternatively, a Comidor user from the customer service department can manually trigger the same process within the Comidor environment, in case the customer places the request by phone, email or another source.

2. Process Flow

An automated email is sent to the customer confirming the receipt of the request.
Then, the ML text classification model makes a suggestion based on the customer’s request subject. The ML model has been trained with historical data to ensure the accuracy of classification.
An AI Sentiment Analysis model is used to identify and categorise opinions expressed in the request description and determine whether the customer’s attitude is positive, negative or neutral.
Based on the sentiment, the ticket priority changes accordingly, e.g. for negative sentiment, the ticket priority is set to top.
The Account Manager is notified about the ML text classification and the sentiment and then makes the final decision.
Then, the responsible department handles and resolves the ticket.
The Account Manager reviews the resolution. If the resolution is confirmed, an automated email is sent to the customer. If not, the ticket loops back to the department for resolution.
Finally, the Account Manager awaits for customer’s confirmation. If the customer agrees the ticket is closed. If not, the ticket loops back once again to the department for resolution.

What we achieved:

Real-time monitoring and reporting of all customer requests
Involvement of non Comidor users in internal processes
Lower resolution time with automatic request categorization
Increased productivity since manual steps have been removed
Better customer experience due to automatic prioritization

Find more information about AI/ML and Workflow elements that you can include in your workflows.

The post AI/ML Application Cases appeared first on Comidor Low-code Automation Platform.

Leia Chatbot & Knowledge Base

Comidor team — Thu, 07 Apr 2022 05:22:51 +0000

Leia Chatbot

Introducing Leia. Comidor’s new intelligent virtual assistant.

Every organisation deals with multistage internal processes, workflows, forms, rules, and regulations.

It’s vital for every employee to have access to essential information in order to perform their work efficiently and effectively.

Leia is an AI-enabled assistant that helps employees and teams work smarter, remotely, and more efficiently.

This chatbot can have quite an influence on how your employees experience their day-to-day duties. It can assist them in a more natural, more engaging, and ultimately, more human way. It is available at any time, and accessible at the right bottom of your screen.

To access the Leia chatbot, click on the chat icon at the right bottom of your screen, and a chat window will pop up.

Simply ask a question and Leia answers the question with specific data or recommends a useful reading source.
If the answer is large, you can see it in full view, as per Leia’s suggestion.

Give your feedback on Leia’s answer to assist with the chatbot’s training. Simply click on Yes/No at “Was this answer helpful?”.

Furthermore, the user can choose to send the question via email to a pre-defined email account, in case the question is not properly answered by the bot.
- The administrator’s email should be defined by an application parameter.
In case your question can be answered in multiple ways, the chatbot will suggest more than one answer. So, you can click on the one you wish to view in full view.

Other Actions

Leia’s capabilities have been enhanced, and apart from answering questions based on the Knowledgebase system, users can interact with Leia via slash commands:

/add

Leia can understand the “/add” command and insert records (tasks, processes, and contacts) in Comidor.

For example, if you use the “/add” command to create a task, Leia asks proper questions to fill in mandatory fields.

After the creation, a link with the new record is displayed. Also, the new process can be created by choosing your desired process template.

/quick

You can initiate an app with the “quick add” form via the chatbot.

When you write the “/quick” command, Leia populates a list of all “quick add” forms from the custom apps in your account. Once the desired app is chosen, Leia asks proper questions to fill in all fields from the selected form, in order to initiate a process.

clear

This command deletes the chat history.

Don’t miss any of your important upcoming meetings or activities. Leia will alert you about your upcoming tasks.

Leia frequently checks and alerts the user of the tasks that are scheduled for the next 30 minutes. The user receives a pop-up alert with the scheduled tasks in Leia’s chat.
Leia’s reminder includes the title of the upcoming task and a link to open the task in full view.

Knowledge base

Through the Comidor Knowledge base unit, you can create the brain of your Leia bot.

Leia, the AI chatbot, retrieves data from a knowledge base and delivers information instantly to the end-users.

Comidor allows you to create your own knowledge base, the central repository for all the information your chatbot needs to support your employees and answer questions.

To access the Knowledge base go to Workplace > Knowledge base

Create an Answer

Click on the “+” icon to add a new record.

For every question you are able to add:

A category, so you can group all your answers
keywords separated in commas, that will be used as tags in the bot’s answers
The relevant answer; type a paragraph with your answer here
Any supportive links can be also included in your answer:
- Add an explanatory URL link
- Define the Quick Form JS Link Name and the Quick Form JS Link, so the user when clicking this link, a respective quick add form will appear.
- Choose the Entity Link, from the list of all Comidor Entities (comidor units & custom apps). Then, choose the record that the user will open when clicking on this link.
Click on save, save and new for multiple entries or cancel.

Edit an Answer

Go to Workplace > Knowledge base
Select one or more Sentiment Analysis records.
Click on the pencil icon, apply any change you wish and then save.

You can also apply multiple actions to one or more answers

Select one or more records.
Click on Delete to delete one or multiple records at the same time. A confirmation pop-up box appears.

Create your Model

After adding your questing and answers, you need to create your ai model.

Go to Workplace > Knowledge base
Click on the actions button>Create Model.

Now, Leia will be equipped with all those answers!

The post Leia Chatbot & Knowledge Base appeared first on Comidor Low-code Automation Platform.

Sentiment Analysis & Document Analyzer Models

Comidor team — Tue, 14 Jul 2020 06:34:41 +0000

Artificial Intelligence (AI)

Artificial Intelligence (AI) in BPM is ideal in complicated situations where huge data volumes are involved and humans need to make decisions. Cognitive Automation is capable of performing complex tasks that require extensive human thinking and activities.

Comidor offers some of the key capabilities of Cognitive Automation such as:

Natural Language Processing (NLP): the ability to understand language, with Sentiment Analysis
Optical Character Recognition (OCR): the ability to read and understand written text, with Document Analyzer Models

Use the Sentiment Analysis element in a workflow to identify and categorize opinions expressed in a text field of a workflow, and determine whether the writer’s attitude is positive, negative, or neutral.

Benefits of Sentiment Analysis within organisations:

Develop a more insightful marketing strategy
Understand your customers better
Measure your marketing campaigns
Re-think your brand perception
Give an extra boost to your customer service

With Comidor Document Analyzer Models enterprises can scan documents like invoices and create digital copies. In particular, the text that is extracted from the document is saved in a text field and can be used within simple or more advanced workflows.

Use Document Analyzer Models and achieve:

Higher Productivity by retrieving data quicker
Cost Reduction
High Accuracy
Increased Storage Space
Massively Improved Customer Service

Sentiment Analysis

To access Sentiment Analysis, go to App Factory Icon > Process Automation > Sentiment Analysis.

Click on the + Icon at the top of the screen to open the Create Form.
Type a Title to your Sentiment Analysis.
Link this Analysis with a Connected Application or select the respective Entity.
Define the field in which the Sentiment Analysis should be performed, in Sentiment Analysis Field.
- Select a text or memo field.
Set the Response Field, where the sentiment result will be stored.
Select the desired Save option (refer to the Quick Reference Guide).

Edit Sentiment Analysis

Go to App Factory Icon > Process Automation > Sentiment Analysis.
Select the Sentiment Analysis record to edit.
Click on the Edit button to open the Edit Form.
Edit the information you want and click on the desired Save option (refer to Quick Reference Guide)

Test Sentiment Analysis

After creating your Sentiment Analysis model, you can test the sentiment results based on data that you type.

Go to App Factory Icon > Process Automation > Sentiment Analysis.
Select the Sentiment Analysis record to test.
Click on the Actions button > Test.
- In the pop-up window, type your test data, click on the Test button and you will see the sentiment result (POSITIVE, NEUTRAL, NEGATIVE).

Delete Sentiment Analysis Models

Go to App Factory Icon > Process Automation > Sentiment Analysis.
Check one or more Sentiment Analysis records.
Click on Delete to delete one or multiple Sentiment Analysis records at the same time. A confirmation pop-up box appears.

Document Analyzer Models

To access Document Analyzer Models, go to App Factory Icon > Process Automation > Document Analyzer Models.

Click on the + Icon at the top of the screen to open the Create Form.

Select the preferred OCR provider (Amazon, Comidor, or HTML converter).
Define the Excel Extraction Method(Amazon or PDFTables), if the document you wish to analyze includes tables
Type a Title to your Document Analyzer Model.
Choose a file type among PDF, JPG, or PNG from your PC to upload the Document for analysis. This Document will be used as the template in the workflow Document Analyzer.
Define the Response type to be per Line or per Word.
Select the desired Save option (refer to the Quick Reference Guide).

Analyze your Document

After uploading the document, and while being at the view form of the Document Analyzer Model, click on the Analyze button.

OCR provider: Amazon example

The Document has been analyzed per the selected options, and the response is saved in Document Analyzer Response.
You can see the preview of the document, marked with blocks, wherever text was spotted.
Drag-and-drop the blocks you wish to the right side of the screen and give a name to a parameter to define a section in your document.
- This parameter will be used in the Document Analyzer Parameters of the Document Analyzer component in the Workflow design.
Click on the + icon to create more correlations, or on the – icon to remove a row.
Finally, click on “Send” to save your Parameters in Model Parameters.

OCR provider: HTML converter example

The Document has been analyzed, converted to HTML and the response is saved in Document Analyzer Response.
Click on the pencil button to add Free text parameters:
- In the Key type the parameter name. This parameter will be used in the Document Analyzer Parameters of the Document Analyzer component in the Workflow design.
- Specify the HTML tag as the Document Element Start and Document Element End to define the area of the document you wish to be captured.
  - You can also add [eol] in Document Element End for fetching the text until the end of the current line.
- Use the Index to limit the area of the document you wish to be captured. E.g. if you want to fetch 19 chars prior to Document element start, add “-19,S”. If you want to fetch 7 chars after the Document element end, add “+7,E”.
- In case you want the data from your document to be displayed as a table specify is table Yes in your parameter.
- Add as many parameters as you wish by clicking on the + icon, or remove a row with the – icon.
Metadata can also be captured from the document. Type the key of the parameter and choose the desired metadata element from the list (Author, Creation date, Title, Subject, etc)

Edit Document Analyzer Models

Go to App Factory Icon > Process Automation > Document Analyzer Models.
Select a Document Analyzer Model to edit.
Click on the Edit button to open the Edit Form.
Edit the information you want and click on the desired Save option (refer to Quick Reference Guide).

Delete Document Analyzer Models

Go to App Factory Icon > Process Automation > Document Analyzer Models.
Check one or more Document Analyzer Models.
Click on Delete to delete one or multiple Document Analyzer Models at the same time. A confirmation pop-up box appears.

To access Workflows go to App Factory Icon > Business Automation > Workflows

Sentiment Analysis

Drag-and-drop the Sentiment Analysis element.
Give a Title to the element.
Give the Parent Stage which is the stage of the parent process as soon as this step is reached.
Select which Model you would like to run at this step, from the list of the Sentiment Analysis Models that you have already created.
Define the field in which the Sentiment Analysis should be performed, in Sentiment Analysis Field.
Set the Response Field, where the sentiment result will be stored.

Document Analyzer

Drag-and-drop the Document Analyzer element.
Give a Title to the element.
Give the Parent Stage which is the stage of the parent process as soon as this step is reached.
Select which Model you would like to run at this step, from the list of the Document Analyzer Models that you have already created.
Document Analyzer Field: set the binary field where the user will upload the document to be analyzed.
Define the Document Analyzer Parameters by correlating the variables (parts of the template document that was saved and analyzed in Document Analyzer Model) with the fields where the text of the document should be stored in.
You can create a text field and set it as the Response Field, to see the response of this component.

Image Classification

Drag-and-drop the Image Classification element.
Give a Title to the element.
Give the Parent Stage which is the stage of the parent process as soon as this step is reached.
Select the binary field where you would upload the Template File. The image classification component will search for the wanted image in the Template file.
Wanted image: upload the .png file from your desktop that you wish the image classification to search for.
You can create a text field and set it as the Response Field, to see the response of this component.
Add the Actual value that you would like to be returned in case the wanted image is found in the Template file. If it is not found, the actual value would get -1 as a value.

Find out more on how to create and manage workflows step by step and learn all about Comidor Workflow Elements.

The post Sentiment Analysis & Document Analyzer Models appeared first on Comidor Low-code Automation Platform.