Overview of Orbita Answers

One of the most common uses for chatbots and smart assistants is answering questions posed by end-users. “Where is the hospital located?” “How much is my co-pay?” “What are the early signs of breast cancer?”

To answer questions like these, a bot must do a few things:

• Understand the question being asked

• Access content with information related to the question

• Find the right information in the content to answer the question

While Orbita supports several different approaches for implementing a question-answering bot, in almost all cases the recommended approach is to use Orbita Answers.

Orbita Answers uses a fact-based approach to question-answering, which means that answers to questions are drawn from a structured knowledge base of information managed within Orbita. The knowledge base organizes information by facts. A fact is simply a structured content item containing information about a specific topic. A fact also has a set of questions associated with it.

To take an example, imagine a knowledge base of first aid information. The facts within this knowledge base might cover topics such as “influenza” and “sunburn.” Within the topic of influenza, there might be multiple facts regarding treatment, symptoms, risk factors, etc.

A single fact about influenza treatment would have questions associated with it like:

• How do you treat influenza?

• How can you treat the flu?

• treatment options for the flu?

Understanding the Knowledge Base

The structure of a Knowledge Base can be compared to an Excel spreadsheet with four columns:

• Article

• Questions

• Topic

• Relationship

The article represents a discrete piece of content containing information that authoritatively answers the questions. A topic represents the main subject of an article. A relationship represents a subcategory or attribute of the article.

For example, continuing with the first aid example, a knowledge base containing facts about the flu and sunburn might look like:

Table 1: Example table representing a knowledge base of first aid information

Each row in the spreadsheet represents a fact. In the example table above, there are three facts. One fact for sunburn, and two facts for influenza.

More about Topics & Relationships

Topics and relationships serve two purposes in Orbita Answers.

• The first is for purely organizational purposes. It makes it easier to find and manage facts when they’re organized in this way.

• The second is for accuracy purposes. You can create lists of synonyms for topics and relationships to capture the various ways a person might refer to them conversationally. For example, someone might refer to the topic “influenza” as “the flu.” Or the relationship “symptoms” as “signs.”

As you can see in Table 1, the relationship “Treatment” is used for both Influence and Sunburn. This is ideal and a sign of a well-structured knowledge base. In general, it is good practice to identify a set of general categorical attributes that are common to all topics. For example, in a first aid knowledge base, the relationships “treatments,” “symptoms,” “risks,” and “prevention” generally apply to all topics.

More about Questions

Questions should cover the various ways someone might ask for information covered in the article.

How many questions do you need per fact? There’s no hard rule. A general rule of thumb is, no less than six but the more the better. Most important is capturing the variation in how people conversationally ask the same question.

Another general best practice is to think of questions that include the topic and relationship. In situations where you have a question that does not contain the topic and relationship in the question, there may be an opportunity to include words from the question into the synonyms for the topic or relationship.

Organizing your Knowledge Base

Using a spreadsheet will make it easier to build and organize your knowledge base.

1. Create a spreadsheet with four columns: article, questions, topic, relationship

2. Organize your content into meaningful chunks. Each chunk should represent an article

3. Identify the primary topic for each article.

4. Identify the relationship of the article

5. Define a set of questions that are answered by the article.

As you go through the following steps, keep in mind the best practices and guidelines described in the previous sections (More about Topics & Relationships and More about Questions). Also, keep in mind that developing the spreadsheet will most likely be an iterative process.

With this spreadsheet complete, then:

1. For each topic, define a set of synonyms capturing the various ways someone might refer to the topic (e.g. Anaphylaxis → “severe allergic reaction,” “anaphylactic shock,” etc.)

2. For each relationship, define a set of synonyms for the same purpose (e.g. Treatment → “remedy,” “therapy,” etc.)

Best Practices for Naming Topics

• A topic is a word or phrase (most likely be a noun, noun phrase, proper noun, or compound noun).
  Example: Sunburn

• A topic name doesn’t necessarily need to be a word in a dictionary. It could be a business name, brand name, product name, etc.

• A topic name should be the singular form of the noun (e.g. Chemical Burn vs Chemical Burns). Orbita handles lexemes (burns, burning, burned, etc.)

• There are cases where a topic might be a verb (e.g. Eat) and should be the dictionary form of the word (e.g. Eat, rather than eats, eating, ate. This form is also called the lemma form).

• A topic can be an acronym (e.g. COPD). When specifying an acronym, use single letters with no spaces.

• A topic may contain multiple concepts (e.g. Cancer and Hunger) but generally should focus on a single concept (e.g. Cancer, Hunger). A best practice is to think about the questions people will ask – if someone is likely to ask about each concept separately, make it two topics. If they ask about two concepts in the question (e.g., will I lose my appetite during cancer treatments?), it’s okay to have a topic with two concepts.

Best Practices for Naming Synonyms

• A synonym is a word or phrase that means exactly or nearly the same as the topic (most likely be a noun, noun phrase, proper noun, or compound noun).

• A synonym should be the singular form of the noun (e.g. Chemical Burn vs Chemical Burns). Orbita handles lexemes (burns, burning, burned, etc.)

• A synonym should reflect the way people express topics conversationally and will not necessarily be a synonym that exists in a dictionary for your topic (e.g. severe allergic reaction or covered in hives may be synonyms for the topic anaphylaxis). Research ways people refer to concepts conversationally by looking at web search logs, call center logs, talking with customer support, etc.

• A synonym should consider the question someone might ask about the topic. For example, Assume a knowledge base of First Aid content, where a topic/relationship is Anaphylaxis/Causes. Consider the question What causes severe hives? ”severe hives” would be a synonym for Anaphylaxis

• A synonym name doesn’t necessarily need to be a word in a dictionary. It could be a business name, brand name, product name, etc.

• A synonym name can be an acronym. When specifying an acronym, use single letters with no spaces (e.g. COPD)

• A good synonym reflects the ways people refer to the concept conversationally.

Defining a Knowledge Base in Orbita

Using your spreadsheet as a guide, you’re now ready to create your Knowledge Base in Orbita.

Create a New Knowledge Base

In Experience Manager, select your project, then:

1. Navigate to Answers

2. Click the “+” icon

Open

In the dialog:

1. Provide a Name for your knowledge base

2. You can choose either the default or custom schema, you created, for the facts edit screen. How do I create Schemas?

3. Check the checkbox titled “Generate intent per fact in KB”.

4. Click “Add”

Open

From the list of Knowledge Bases, click and open your newly created Knowledge Base.

Create Your First Fact

To add a fact, click on the (+) icon.

Adding an Article

1. On the Answer tab, enter the article content into the Multi-modal Content Editor (default schema). If you have a custom schema, this view will change according to the schema selected.

The fact you are working on will be locked for other users to edit. Any other user who tries to edit the locked fact will get a pop-up message and can choose to force edit by clicking on Continue.

Adding a Topic

1. Enter the name of the topic in the

Topics field. You’ll be able to select existing topics once they have been created.

1. Click “OK” when prompted. For example, if you create a topic for Sunburn, the following prompt will appear.

1. Click the Synonyms icon to specify alternative names for your topic. For example, “sunburn” might be referred to as “sun poisoning.” (e.g.

What are the signs of sun poisoning?)

Adding a Relationship

1. Enter the name of the relationship in the

Relationships field. You’ll be able to select existing relationships once they have been created.

1. Click “OK when prompted. For example, if you create a relationship for Treatment, the following prompt will appear.

1. Click the Synonyms icon to specify alternative names for your relationship. For example, “treatment” might be alternatively referred to as “remedies” or “home remedies” (e.g.

What are home remedies for the flu?)

Adding Questions

1. Click on the Questions tab

2. Enter in questions capturing the various ways someone might ask for information covered in the article. For example, if the article includes location information for a hospital, questions might include “Where is the hospital located?” “Where’s the hospital?”, etc.

3. Click Save button

Best Practices for creating Facts

It is a best practice to add less than 8 Relationships on a Fact to avoid a known performance issue.

If you have 8 or more Relationships to be added to the Fact, as a workaround, instead of adding all the Relationships to a single Fact, you could create two separate Facts, dividing the number of Relationships across them.

For example, instead of having one single Fact with 1 Topic + 8 Relationships, you could have two Facts, with 1 Topic + 4 Relationships each.

Managing a Knowledge Base in Orbita

Viewing a List of All Topics

1. From the list of knowledge bases, find the desired knowledge base and click its action menu icon (three dots on the right of the row)

1. From the action menu, select Settings

1. From the Settings screen, you can click the “Topics” to see a list of topics or relationships.

Adding a Related Topic

Topics can be related to one another. For example, you might have topics for Asthma, Bronchitis, Pneumonia, and COPD. Because these topics all deal with lung disease, you might want to relate them to one another.

1. Follow the steps above for “Viewing a List of All Topics”

2. Find the desired topic from the list

3. Click the topic’s action menu icon

4. Choose Edit from the menu

1. Open the

Related Topics tab.

1. Click

Add (+) to add a related topic. You must select an existing topic from the list.

Your related topic now appears on the list of related topics

Managing Advanced Settings

1. From the list of knowledge bases, find the desired knowledge base and click its action menu icon (three dots on the right of the row)

1. From the action menu, select Settings

1. From the Settings screen, click “Advanced” to see advanced settings options

Generating an Intent per Fact

Enabling this option creates an intent per fact, using the questions associated with the fact (on the Questions tab) to train the natural language processor. It is recommended to enable this option as it increases accuracy and provides sufficient and relevant questions to the question tab for each fact.

The new tab Intents will capture all the Auto-generated intents for each fact.

Refer, How do I create an intent? and How to use Annotations in Intents for more information on intents.

Setting a Primary Topic

In most cases, a knowledge base will be created to capture information on a specific topic. For example, imagine a knowledge base created to contain information about sleep & wellness. Topics in this knowledge base might include general information about sleep or specific information about insomnia treatment.

Identifying a primary topic helps Orbita Answers to disambiguate questions that include multiple topics. For example, the question “I have insomnia and can’t sleep” contains two topics from Orbita Answers' perspective: Insomnia and Sleep (e.g. “I have insomnia and can’t sleep”). By identifying “sleep” as a Primary Topic, Orbita Answers understands that this particular question is asking about “insomnia” and not generally about “sleep.”

Using a Fallback

The fallback is used to define a relationship that will be used when one is not present in the question. For example, imagine someone is looking for information about “insomnia” but just says the phrase “insomnia” instead of asking a question “What is insomnia?”

By defining a fallback relationship, the fallback relationship will be used in cases where an end-user asks about a topic, without expressing a relationship in the question (e.g. “insomnia”, “COVID-19”, etc)

A general best practice is to create a relationship covering general information and to use it as a fallback relationship.

See also How do I set up and use Answers?

Click on the links below to launch the training and other content.

Video Content

https://cdn.orbita.cloud/assets/orbita/answers/Orbitaanswers.mp4

Supporting Documentation

https://orbita.atlassian.net/wiki/spaces/OCS/pages/872185926

Guide to Orbita Answers

Useful Links

http://orbita.ai

Related Content