AI-102 EXAM QUESTIONS 2024 MICROSOFT AZURE AI 102 CERTIFICATION COURSE PART-6

This video is part six of a case study series focusing on AI questions. The presenter moves on to the next set of questions, starting with a scenario involving a smart e-commerce project. The project requires a cognitive search solution with 99.9% availability for search and indexing. It also needs to extract information from images, manuals, and videos associated with products, and support auto-completion and suggestions based on product names in English, Portuguese, and Spanish. Additionally, it must store generated insight data for later processing and immediately update stock levels. The first question asks which three actions should be performed to implement auto-completion. The presenter explains that auto-completion involves sending queries to an endpoint that returns suggestions based on typed input. This is achieved by making API queries to the auto-complete endpoint and including the suggester name in the request body. Crucially, a "suggester" must be added to the index, specifying which fields will provide the auto-completion content. For this project, the suggester should include the three product name fields from each of the three languages (English, Portuguese, and Spanish) as source fields. The presenter clarifies that using the search endpoint is for general search, not auto-completion, and creating separate suggestors for each language field is inefficient compared to a single suggester with multiple source fields. Setting the search analyzer property is also deemed less direct for auto-completion compared to the suggester mechanism. The next question shifts to a product creation project, focusing on building a REST endpoint for multilingual product descriptions. The requirement is that data storage and processing must occur in data centers located in the United States. The presenter refers to the case study's governance and security requirements, which mandate US-based data processing. For translation, the presenter identifies the need to use a translator API. The base URL for the global translator is api.cognitiveservices.microsoft.com, but for the Americas, as specified by the requirement, the URL should be api-nam.cognitiveservices.microsoft.com. Therefore, the blank in the URL should be filled with "api-nam.cognitiveservices.microsoft.com". The second part of this question asks what should fill the second blank in the REST endpoint URL. Given the goal is to create multilingual product descriptions, the appropriate API function to call is "translate". Other options like "detect" (for language detection) or "languages" (for supported languages) are not suitable for the core translation task, and "text to speech" is for audio conversion. The following question concerns developing code to upload images for the product creation project, specifically meeting accessibility requirements. These requirements include having relevant alt text for all images, transcripts for all videos, and product description transcripts and alt text available in all three languages. The question asks how to fill the first blank in a code snippet for uploading images. The code involves using the Computer Vision client to analyze images and generate alt text. The presenter emphasizes that for multilingual support, Image Analysis 3.2 API is preferred over 4.0, as 3.2 supports English, Chinese, Portuguese, Spanish, and Japanese, covering the project's required languages. The first blank in the code, representing the image input, should be a "string" because the images are already stored in blob storage and accessible via a URL. Using a "dictionary" is incorrect as it's for key-value pairs, and "stream" would be used if the image itself (in binary form) was being provided directly without a pre-existing path. The second blank in this code snippet relates to specifying the type of visual features to extract. The presenter refers back to a previous question where "visualFeatures.description" was identified as the correct option for analyzing image descriptions and generating captions for alt text. Therefore, the blank should be filled with "visualFeatures.description". The third blank in this code snippet involves selecting the correct property from the analysis results. The goal is to obtain the image description's caption, which includes the text and a confidence score. The presenter explains that this is accessed via "results.description.captions". The code then checks if the confidence score is above a certain threshold (e.g., 0.5) to ensure the AI's confidence in the generated alt text. Finally, the video addresses a question about the process for analyzing videos within the product creation project, requiring a sequence of four actions. The first logical step is to "upload the video to blob storage," as the video needs to be stored before processing. The second step is to "index the video by using the Azure Video Analyzer for Media (formerly Video Indexer) API." This service provides advanced video analytics, including indexing and transcript extraction. The third action should be to "extract the transcript from the Video Analyzer for Media." Once the video is indexed, its transcript (captions or subtitles) needs to be extracted. The fourth and final action is to "translate the transcript by using the Translator API." This fulfills the requirement of having transcripts available in multiple languages (English, Portuguese, and Spanish). The presenter explains why other options are incorrect, such as using the Computer Vision API for video analysis instead of the specialized Video Analyzer, or attempting to send transcripts to the Language Understanding (LUIS) API before they are extracted. The presenter concludes by stating that all case studies and related questions have been covered and signs off.