Free SAS Viya Natural Language Processing and Computer Vision Practice Questions

Sample Practice Questions

Try these 10 sample questions from the SAS Viya Natural Language Processing and Computer Vision practice bank. Each includes the answer and explanation. Continue with the interactive practice above to review the full question set.

1. 1Which encoding should text data use when it is sent to SAS Visual Text Analytics in SAS Viya?

   1. A. UTF-16
   2. B. UTF-8
   3. C. US-ASCII
   4. D. WLATIN1

   Answer: B

   Explanation: The guide states, "SAS Viya processes all character strings as UTF-8." It also says, "Text data should be sent as UTF-8". UTF-8 is therefore the recommended preparation encoding for text sent to this workflow.

   Reference: SAS Visual Text Analytics: User’s Guide — Chapter 1, Encoding Text Data, p. 9

2. 2A text analysis should consider only a maintained vocabulary, rather than all terms except a few exclusions. Which Text Parsing resource directly implements this vocabulary restriction?

   1. A. A synonym list mapping terms to parents
   2. B. A stop list containing the maintained vocabulary
   3. C. A start list containing the maintained vocabulary
   4. D. A category variable containing vocabulary names

   Answer: C

   Explanation: The guide says, "If you use a start list, then only terms that are included in that list appear in parsing results." A stop list would exclude the listed vocabulary, while a synonym list combines terms and a category variable supplies labels.

   Reference: SAS Visual Text Analytics: User’s Guide — Chapter 1, Start Lists and Stop Lists, p. 6

3. 3At one convolution position, the input patch is [[2, 1], [3, 4]] and the filter is [[1, 0], [0, -1]]. With zero bias and identity activation, what output value results from the elementwise dot product?

   1. A. -2 (negative two)
   2. B. 2 (positive two)
   3. C. 6 (positive six)
   4. D. -6 (negative six)

   Answer: A

   Explanation: SAS describes convolution as computing "the dot product between the filter and the input image during each step." For this patch the sum is 2×1 + 1×0 + 3×0 + 4×(-1) = -2. Zero bias and identity activation leave that value unchanged.

   Reference: SAS Viya: Deep Learning Programming Guide — Convolution — Convolution, Overview of Convolution, Filter, and Stride

4. 4Document text has been converted correctly, but joins on character document IDs that include Japanese text are unreliable in a SAS Visual Text Analytics workflow. Which additional encoding check is specifically relevant?

   1. A. Check that document IDs use UTF-8.
   2. B. Check that document IDs use UTF-16.
   3. C. Check that document IDs use WLATIN1.
   4. D. Check that document IDs use US-ASCII.

   Answer: A

   Explanation: The guide says, "if the document ID is a character variable, it should be in UTF-8." It explains that otherwise "the merging of tables based on document IDs might not be applied correctly." Correctly encoded document text alone does not establish that the IDs are suitable for merging.

   Reference: SAS Visual Text Analytics: User’s Guide — Chapter 1, Encoding Text Data, p. 9

5. 5An analyst wants automobile and car to contribute as one analysis term while retaining car as the representative label. Which resource is designed for this grouping?

   1. A. A synonym list with automobile mapped to car
   2. B. A stop list with automobile and car included
   3. C. A start list with automobile and car included
   4. D. A category variable with automobile and car labels

   Answer: A

   Explanation: A synonym list "identifies pairs of words that should be combined as single terms for the purposes of analysis." The PARENT supplies the representative term. Start and stop lists control inclusion rather than grouping; a category variable labels documents.

   Reference: SAS Visual Text Analytics: User’s Guide — Chapter 1, Terms and Synonyms, p. 5

6. 6A CNN convolution uses a 31 × 31 input, a 3 × 3 filter, stride 2 in both dimensions, and explicitly zero padding. Only complete filter placements are evaluated. What is each output feature map’s spatial size?

   1. A. 14 × 14
   2. B. 16 × 16
   3. C. 15 × 15
   4. D. 29 × 29

   Answer: C

   Explanation: SAS defines stride as "the distance (for example, number of pixels) that the filter traverses across the input matrix between each convolution stride." Valid starting positions along either dimension are 0, 2, …, 28: (31−3)/2 + 1 = 15 positions. Thus each output map is 15 × 15.

   Reference: SAS Viya: Deep Learning Programming Guide — Convolution — Convolution, Overview of Convolution, Filter, and Stride

7. 7An analyst is preparing a document collection for a model that will later process customer correspondence. Which selection principle does SAS recommend?

   1. A. Select documents with nearly identical lengths.
   2. B. Select the shortest documents from each source.
   3. C. Select documents representative of the later workload.
   4. D. Select documents from the most frequent category.

   Answer: C

   Explanation: SAS recommends selecting documents "that is representative of the documents that you want to process later." This matters because "The terms and patterns that exist in the input document collection influence the creation of any models." Length or category frequency alone does not ensure representative training material.

   Reference: SAS Visual Text Analytics: User’s Guide — Chapter 3, Preparing the Document Collection, p. 24

8. 8A predefined concept almost meets a project’s extraction needs, but extra matching behavior is required. Which change is supported in SAS Visual Text Analytics?

   1. A. Rename the predefined concept to replace its built-in behavior.
   2. B. Edit the predefined concept’s hidden base definitions directly.
   3. C. Replace its priority value with the text of the extra rule.
   4. D. Add rules in the rule code editor to extend its behavior.

   Answer: D

   Explanation: The guide states, "You cannot rename predefined concepts, nor can you view or edit their base definitions." It permits "additional rules in the rule code editor of the Edit Concept tab to modify or extend their behavior." Priority resolves overlapping matches; it does not hold rule text.

   Reference: SAS Visual Text Analytics: User’s Guide — Chapter 1, Concepts, p. 3

9. 9A standard convolutional layer produces 20 × 20 spatial maps and uses nFilters=12. What is the resulting output volume, expressed as width × height × depth?

   1. A. 20 × 20 × 3
   2. B. 20 × 20 × 12
   3. C. 12 × 12 × 20
   4. D. 20 × 12 × 1

   Answer: B

   Explanation: The CNN documentation illustrates that using 12 filters yields an output volume with depth 12: "If you decide to use 12 filters, a volume might result such as 32 x 32 x 12." Applying the same filter-count rule to the specified 20 × 20 maps gives 20 × 20 × 12.

   Reference: SAS Viya: Deep Learning Programming Guide — CNN Layers — CNN Layers, Convolution Layer

10. 10After SAS Data Explorer imports a directory of documents for text analysis, what does one row of the resulting in-memory table represent?

    1. A. One extracted term and its corpus frequency
    2. B. One source directory and its document count
    3. C. One document page and its image coordinates
    4. D. One source document and its metadata

    Answer: D

    Explanation: The import overview states, "Each row in the table represents one document and its metadata." The output is organized by document, rather than by term, directory, or page.

    Reference: SAS Data Explorer: User’s Guide — Importing Documents from a Caslib Directory, Overview, p. 59