Training with eva Automated Learning
A software to ease up
For questions less frequently asked but still relevant to the virtual agent's context, there is an application in the Dialog Manager that dispenses intent training: Automated Learning.
This feature is available starting from the version 3.3.1.0
For example, imagine that you're working in a banking virtual agent project and there are much of institutional information, such as the story of the bank, CEO's profile, the board of directors... Although this is important information, rarely a customer asks for that, right?
So, for cases like that, you can save many hours for your team using Automated Learning, a cognitive engine developed by NTT Data with the latest advances in NLP
In summary, it's a question-answering software that interprets users’ questions and answers them based on documents uploaded beforehand.
Here, is what Automated Learning can do for your project:
Finds the answer to the user's question in one or multiple documents, without previously training any data, like it's needed for the intents engine
Has a FAQs functionality for documents by registering pairs of questions and answers. Those answers are brief and accurate
No need to execute specific training to work with a text knowledge base. Automated Learning is multi-language, and it was tested in Spanish, English and Portuguese
Generates an output (likelihood) indicating the confidence of the model response
Ability to identify if there is no answer to a question, given the context of the document
Important: Frequently accessed subjects and transactional subjects need to be accessed by the Intent engine
Although Automated Learning can read and extract answers from any text, some best practices make its assertiveness better.
a) Longer paragraphs work better than short ones
Automated Learning works with context.
So, the longer the material is, the inference is better. Paragraphs with 4 sentences works better than paragraphs with 2 sentences. If you have 2 small paragraphs, it is better to join them and make a larger, 4-sentence paragraph.
For example, instead of:
John is a nice man. John lives in New York. John likes zucchini.
Prefer:
John is a nice man. He lives in New York. John likes zucchini.
b) Avoid bulleted lists
Bullet lists are the extreme example of the lack of context that small paragraphs have. Try to make a paragraph of those lists.
Instead of a list, such as:
- Gazpacho
- Tomato
- Onion
- Garlic
- Stale bread
- Salt, olive oil and vinegar
Try condensing in a paragraph, such as:
To make gazpacho, you will need tomato, onion, garlic, stale bread, salt, olive oil and vinegar.
The answers delivered by Automated Learning are short, up to 10 words. If you have a large list, try to split it.
For example, if you have a list with 30 items, you can split it into three lists in three different paragraphs, like this:
List A – Items 1 to 10
List B – Items 11 to 20
List C – Items 21 to 30
c) Avoid tables and forms
Automated Learning does not read well tables and forms. It is better to rewrite the contents in tables and forms as paragraphs or dismiss them.
d) Remove special characters, emojis, and images. Non-ASCII characters tend to turn to strings of characters that are not readable by Automated Learning. Although some special characters are common (like the dollar sign, $), some are more obscure and can hamper eva NLP inference. For example, the character ♠. Emojis count as special characters and are better removed. Images also might interfere with eva NLP AL functioning.
e) Remove headers, footers, summaries, references and any other elements that are not relevant to the document.
Headers, footers, summaries, and elements alike add unnecessary information that might make it harder for the cognitive engine to deliver precise answers.
f) Pay special attention to punctuation
Errors in punctuation hamper understanding much more for a cognitive engine than for a human reader. Punctuation matters a lot. “I’m sorry; I love you” is an apology. “I’m sorry I love you” is cruel.
g) Avoid passive voice
“Gazpacho is an example of cold tomato soup” works better than “An example of cold tomato soup is gazpacho”.
h) Avoid subordinate clauses
If you have a sentence with subordinate clauses, try to dismember it in smaller sentences. “Gazpacho is a cold tomato soup from Spain that, although characterized by the tomato, didn’t took tomato until the 19th century” is a long sentence with lots of subordinate clauses. “Gazpacho is a cold tomato soup. It is from Spain. Although characterized by the tomato content, it didn’t take tomato until the 19th century” works better.
i) Pay attention to the voice and tone of your documents
As the answer inference is automatic, if you extend your virtual agent voice tone to its documents, you will guarantee a consistent virtual agent personality.
Some tips when uploading FAQs (Frequently Asked Questions):
Before uploading a FAQ on Automated Learning, make sure that you remove the questions and upload only the answers. Automated Learning works on FAQs better if the questions are absent.
When writing a FAQ, think of your text as a question - answers pairs. First, write them and then remove the questions.
Think of answers to short questions. Not only most doubts can be expressed with short questions, but Automated Learning works better when the questions are short.
Important:
Automated Learning only accepts documents in .txt format
To achieve greater assertiveness and productivity, upload documents up to 400kb
Using Automated Learning
After you adapted your documents following those guidelines and converted them to .txt, you can upload them to Automated Learning.
Clicking on the Automated Learning tab in the Cockpit will lead you to the Automated Learning first page. Here are the steps:
First, you have to enable Automated Learning in eva, as it comes disabled by default.
After that, you can upload documents.
To import a document, click on “Import document”. After the first document is imported, you will be taken to the documents’ repository.
There, you can check the documents that are part of a virtual agent knowledge base, how many and which questions were coupled to a particular document, delete, and disable or enable a document.
Important: Remember to always train the virtual agent after you imported a document
By clicking “view all” on the Questions column, you will see the questions tied to this particular document.
By clicking on “view document” (eye icon), the selected document will be opened, and you can see its details.
Once you are in a document, you can enable or disable it and create specific question and answer pairs to this document.
Important: When you disable or delete a document, all questions attached to it are also disabled or deleted
By clicking on “Create question”, you will be taken to the question creation modal, where you can couple a question and its variants to a specific answer.
Although Automated Learning can answer users’ questions without prior registration, by creating questions, you can tie a user query to a specific answer and edit this answer.
After you name (and optionally describe) a question, you will have to add the question itself and a number of variants. Simply insert the question and its variants in the “add question” field. You can use tags to classify this question.
Example: in a text about soups, you want to tie the question “What is gazpacho?” to the answer “Gazpacho is a cold soup from Andalusia, Spain”.
Some people might ask this question as “What is gazpacho?”, but might ask as “What is gaspacho?” or even “What gazpacho is?”. There are numerous variants to any question. Try to add the most variants, so most forms are covered.
After you created the question, Automated Learning will infer the answer and deliver to you while opening a modal, so you can edit this answer.
Important:
Automated Learning infer the answer after the first question is registered
After the answer is inferred, a modal named “Curate answer” opens with the answer (like a regular answer modal, but with the inferred answer in the text box).
You can copy any part of the text and paste it on the text box or rewrite the answer as you wish. As in a regular answer, you can add buttons and/or technical text.
Example:
For the question “What is gazpacho?”, eva NLP AL found the answer “A cold soup from Andalusia, Spain”. You can edit it so it becomes “Gazpacho is a cold soup from the region of Andalusia, Spain”.
If you want to link the answer to other text, you can swap the document that the answer belongs by clicking the “change document” button. A modal with documents open so you can select the new document.ortant: every question must be attached to a document.
After you click “save”, the question/answer pair will be stored in the Questions repository.
In this repository, you can see the question name, how many variants this question has, who last edited it and which document it belongs.
By clicking on the filter icon, you can see filter questions by a document.
Filter icon By clicking the pencil icon, you can edit the question.
Last updated
