Implement natural language processing solutions (30–35%)

Question 1

What size limitations exist in the language detection API?

Answer 1

the document size must be under 5,120 characters. The size limit is per document and each collection is restricted to 1,000 items (IDs).

Question 2

If multiple sentiments exist in a text with multiple sentences, how will the sentiment analyzer classify the text?

Answer 2

1. If all sentences are neutral, the overall sentiment is neutral.
2. If sentence classifications include only positive and neutral, the overall sentiment is positive.
3. If the sentence classifications include only negative and neutral, the overall sentiment is negative.
4. If the sentence classifications include positive and negative, the overall sentiment is mixed.

Question 3

What is the difference between question answering and conversational language?

Answer 3

Question Answering focuses on providing direct answers to user questions by creating a knowledge base of question-answer pairs.

CLU analyzes the intent and meaning behind a user’s natural language utterance, identifying key information within the text to determine the overall action needed, rather than just providing a direct answer

The two services are in fact complementary. You can build comprehensive natural language solutions that combine language understanding models and question answering knowledge bases.

Question 4

What properties are required in the body of a call to the API for question answering ?

Answer 4

Property Description
1. question - Question to send to the knowledge base.
2. top - Maximum number of answers to be returned.
3. scoreThreshold - Score threshold for answers returned.
4. strictFilters - Limit to only answers that contain the specified metadata

Question 5

What two approaches can be taken to improve performance of a question answering model?

Answer 5

After creating and testing a knowledge base, you can improve its performance with active learning and by defining synonyms.

Question 6

You want to create a knowledge base from an existing FAQ document. What should you do?

Answer 6

Create a new knowledge base, importing the existing FAQ document.

You can create a knowledge base from an existing document or web page.

Question 7

How can you enable users to use your knowledge base through email?

Answer 7

You can create a bot for your published knowledge base and configure a channel for email communication.

Question 8

What preconfigured features exist in natural language processing?

Answer 8

Summarization
Named entity recognition
PII detection
Key phrase extraction
Sentiment analysis
Language detection

Question 9

Describe conversational language understanding (CLU)

Answer 9

CLU helps users to build custom natural language understanding models to predict overall intent and extract important information from incoming utterances. CLU does require data to be tagged by the user to teach it how to predict intents and entities accurately.

Question 10

What steps do you take to build a Conversational Language Understanding project using the REST API?

Answer 10

These tasks are done asynchronously; you’ll need to submit a request to the appropriate URI for each step, and then send another request to get the status of that job.

Each call requires authentication using the header Ocp-Apim-Subscription-Key

1. Request deployment
2. Get deployment status
3. Import data
4. Train model
5. Deploy

Question 11

You have an app named App1 that analyzes social media mentions and determines whether comments are positive or negative.
During testing, you notice that App1 generates negative sentiment analysis in response to customer feedback that contains positive feedback.
You need to ensure that App1 includes more granular information during the analysis.
What should you add to the API requests?

Answer 11

opinionMining=true will add aspect-based sentiment analysis, which in turn will make the sentiment more granular so that positive and negative in a single sentence can be returned.

Question 12

What range of scores of the Bilingual Evaluation Understudy (BLEU) indicates a high quality translation?

Answer 12

between 40 and 60 indicates a high-quality translation.

Question 13

You are creating an orchestration workflow for Language Understanding.
You need to configure workflows for multiple languages. The solution must minimize administrative effort.
What should you create for each language?

Answer 13

separate workflow projects

Orchestration workflow projects do not support the multilingual option, so you need to create a separate workflow project for each language.

Question 14

How can you enable users to use your knowledge base through email?

Answer 14

You can create a bot for your published knowledge base and configure a channel for email communication.

Question 15

You are building an app that will enable users to create notes by using speech.
You need to recommend the Azure AI Speech service model to use. The solution must support noisy environments.
Which model should you recommend?

Answer 15

The custom speech-to-text model is correct, as you need to adapt the model because a factory floor might have ambient noise, which the model should be trained on.

Question 16

You are building an app that will recognize the intent and entities of user utterances in real-time.
You are evaluating the use of intent recognition with the Azure AI Speech and Azure AI Language services or simple pattern matching.
When should you use pattern matching?

Answer 16

You are only interested in matching strictly what the user said.

Question 17

You are building a custom translation model.
You need to use bilingual training documents to teach the model your terminology and style.
Which rule should you follow?

1. Be Literal
2. Be Strict
3. Be restrictive

Answer 17

Be liberal is correct. Any in-domain human translation is better than machine translation. Add and remove documents as you go and try to improve the Bilingual Evaluation Understudy (BLEU) score.
Be strict is incorrect. Compose them to be optimally representative of what you are going to translate in the future.
Be restrictive is also incorrect. A phrase dictionary is case-sensitive, and any word or phrase listed is translated in the way you specify. In many cases, it is better to not use a phrase dictionary and let the system learn.

Question 18

You are building an app that uses Azure AI Services Document Translation.
You need to improve the quality of the translation for user-uploaded documents.
What should you ask the users to include when they upload a document?

Answer 18

If the language of the content in the source document is known, it is recommended to specify the source language in the request to get a better translation.

Question 19

You are creating an orchestration workflow for Language Understanding.
You need to configure workflows for multiple languages. The solution must minimize administrative effort.
What should you create for each language?

Answer 19

Orchestration workflow projects do not support the multilingual option, so you need to create a separate workflow project for each language.

Question 20

You are building a multilingual conversational app by using Conversational Language Understanding (CLU), part of Azure AI Language service.
You create a CLU model that will serve multiple languages.
You need to optimize the performance of the model. The solution must minimize development effort.
What should you do?

Answer 20

Add utterances for languages that are performing poorly in the model.

With CLU, there is no need to use multiple projects for a model. For example, you can train a model in English and query it in German. There is no project language, therefore, adding utterances for languages in the model that are performing poorly is the appropriate solution to increase performance.

Question 21

What are the guidelines for entering utterances in a Conversational Language Understanding model?

Answer 21

1. Capture multiple different examples, or alternative ways of saying the same thing
2. Vary the length of the utterances from short, to medium, to long
3. Vary the location of the noun or subject of the utterance. Place it at the beginning, the end, or somewhere in between
4. Use correct grammar and incorrect grammar in different utterances to offer good training data examples
5. The precision, consistency and completeness of your labeled data are key factors to determining model performance.
   5.1 Label precisely: Label each entity to its right type always. Only include what you want extracted, avoid unnecessary data in your labels.
   5.2 Label consistently: The same entity should have the same label across all the utterances.
   5.3 Label completely: Label all the instances of the entity in all your utterances.

Question 22

What are the three different types of entities that can be used in a CLU model?

Answer 22

1. Learned entities are the most flexible kind of entity, and should be used in most cases. You define a learned component with a suitable name, and then associate words or phrases with it in training utterances. When you train your model, it learns to match the appropriate elements in the utterances with the entity.
2. List entities are useful when you need an entity with a specific set of possible values - for example, days of the week. You can include synonyms in a list entity definition, so you could define a DayOfWeek entity that includes the values “Sunday”, “Monday”, “Tuesday”, and so on; each with synonyms like “Sun”, “Mon”, “Tue”, and so on.
3. Prebuilt entities are useful for common types such as numbers, datetimes, and names.

Question 23

What is the limit of prebuilt components per entity in a CLU model?

Answer 23

You can have up to five prebuilt components per entity

Question 24

What are the steps to building an iterative CLU model?

Answer 24

1. Train a model to learn intents and entities from sample utterances.
2. Test the model interactively or using a testing dataset with known labels
3. Deploy a trained model to a public endpoint so client apps can use it
4. Review predictions and iterate on utterances to train your model