Table of Contents
Enroll Here: Text Analysis 101 Cognitive Class Exam Quiz Answers
Introduction to Text Analysis 101
Text analysis is the process of extracting meaningful information from written language. It involves various techniques to understand and interpret text, ranging from simple statistical methods to complex algorithms. The goal of text analysis is often to uncover patterns, sentiments, relationships, or other insights that can inform decision-making or further research.
There are several key components of text analysis:
- Text Preprocessing: This step involves cleaning and preparing text data for analysis. It may include tasks such as removing punctuation, converting text to lowercase, removing stop words (common words like “and,” “the,” “is”), and stemming or lemmatizing words to reduce them to their base forms.
- Tokenization: Breaking down text into smaller units, such as words or phrases (tokens), is essential for analysis. Tokenization can be straightforward for languages like English but more complex for languages without clear word boundaries.
- Statistical Analysis: Basic statistical techniques like frequency analysis (counting how often specific words occur) and distribution analysis (examining the spread of words across a text or corpus) provide foundational insights into text content.
- Sentiment Analysis: This technique aims to determine the sentiment or emotion expressed in a text. It can be done using lexicon-based approaches (matching words to sentiment dictionaries) or machine learning methods (training models to recognize sentiment).
- Named Entity Recognition (NER): Identifying and classifying named entities (such as names of people, organizations, dates, etc.) in text is crucial for tasks like information retrieval and entity linking.
- Topic Modeling: Discovering abstract topics within a collection of documents without prior labeling is achieved through topic modeling techniques like Latent Dirichlet Allocation (LDA). These methods help uncover themes or subjects that dominate a set of texts.
- Text Classification: Assigning predefined categories or labels to text documents based on their content is useful for tasks like spam detection, sentiment classification, or topic categorization.
- Natural Language Processing (NLP): NLP encompasses the broader field of computational techniques for understanding and manipulating human language. Text analysis is a subset of NLP focused specifically on extracting insights from text data.
Text analysis finds applications across various domains, including marketing (analyzing customer feedback), social media (tracking trends and sentiments), healthcare (extracting information from medical records), and many others where understanding textual data is critical.
In summary, text analysis leverages computational techniques to derive meaning and insights from textual data, enabling automated processing and understanding of large volumes of text efficiently.
Text Analysis 101 Cognitive Class Certification Answers
Module 1 – Getting to know IE Quiz Answers
Question 1: Your client would like to know how its advertising campaign impressed customers. Which IE task would provide this data?
- Event Extraction
- Co-reference resolution
- Sentiment Extraction
- Relation Extraction
Question 2: Which extraction phase can turn a dictionary match of a common first name plus an adjacent regular expression into a “potential person name” entity?
- None of these
- Entity Resolution
- Named Entity Recognition
- Feature Selection
Question 3: Consider a set of news articles that contains 20 mentions of person names. From this source, an extractor extracts 15 entities, 3 of which are incorrect. What are the Precision (P) and Recall (R) values?
- P = 0.30, R = 0.70
- P = 0.70, R = 0.30
- P = 0.80, R = 0.60
- P = 1.00, R = 1.30
Module 2 – Limitations in IE Quiz Answers
Question 1: Which of the following poses huge demands on the IE engine?
- Complex IE tasks
- Heterogeneous text inputs
- Different types of data
- All of the above
Question 2: A typical IE grammar-based workflow:
- Targets the step closest to the nature of the rule applied.
- Follows a unidirectional sequence of steps.
- Targets each step based on the nature of the input data.
- Creates a branched path according to the input and the desired output.
Question 3: We can overcome IE performance limitations by:
- Separating extractor semantics from execution strategy.
- Coupling extractor semantics with execution strategy.
- Parallel processing.
- Making faster finite state transducers.
Module 3 – Getting to know System T Quiz Answers
Question 1: What outputs do extractors generate in System T?
- Extractors
- Regular Expressions
- Annotations
- None of above
Question 2: An output refiner helps you to:
- Define multiple filters.
- Union multiple extractors.
- Define multiple extractors.
- None of above
Question 3: By selecting the Mapping Table checkbox in the Dictionary extractor, you can:
- Map dictionary terms against categories.
- Create a two-column dictionary.
- Add a column of metadata.
- All of the above.
Module 4 – IE with AQL Quiz Answers
Question 1: Which of the following statements describes AQL?
- AQL has a syntax that is similar to SQL.
- AQL has expressive power of algebra.
- AQL separates semantics from implementation.
- All of the above.
Question 2: What are the main advantages of SystemT’s approach towards Information Extraction?
- Richer and cleaner rule semantics
- Better performance through optimization
- Improved quality of results
- A and B
Question 3: Which of the following files can be part of an AQL module?
- Dictionary file
- AQL file
- UDF jar
- All of the above
Module 5 – AQL Basics Quiz Answers
Question 1: Which factor is essential for the Union All statement to work?
- The tuples should be from the same input text.
- The schemas of the tuples should be different.
- The schemas of the tuples should be from a single view.
- The schemas of the tuples should be same.
Question 2: Which of the following options is a valid consolidate policy?
- ContainsButNotEqual
- RightToLeft
- ExactEqual
- ContainedInside
Question 3: When is the Minus statement useful?
- When the two sets of input tuples have different schemas
- When you want to find matches for a sequence pattern
- When you want to subtract a set of tuples from another set of tuples
- All of the above
Module 6 – Advanced AQL Quiz Answers
Question 1: Which type of text can be extracted using the Detag statement?
- Semi-structured text
- Unstructured text
- Structured text
- None of above
Question 2: When should you use a standard tokenizer?
- When token boundaries are defined by punctuation and whitespace.
- When extraction of person names from Chinese text is needed.
- When extraction of parts of speech is required.
- All of the above.
Question 3: Which best practices should you use when developing an AQL module?
- Place large dictionaries and tables in separate modules.
- Avoid using the output view statement when developing extractor libraries.
- Document the source code using AQL Doc.
- All of the above.
Module 7 – Declarative IE and the System T optimizer Quiz Answers
Question 1: Which of the following leads to mistakes when two rules match the same region of text?
- Limited expressivity
- Lossy sequencing
- Rigid matching priority
- None of the above
Question 2: Which of the following strategies can overcome lossy sequencing?
- Expand rule patterns to include features such as aggregation.
- Impose modular tokenization.
- Include matching regimes that increase flexibility on priority.
- Use grammar rules that operate on graphs rather than sequences of annotations.
Question 3: In which stage of the SystemT optimizer do you merge block plans into a single operator graph?
- Post-processor
- Planner
- Pre-processor
- None of the above
Module 8 – Best Pratices Quiz Answers
Question 1: Why is it that the first document in a collection is often at the top of the AQL Profiler’s “hot” document’s view?
- The optimizer is trying to produce plans that are sensitive to each input document.
- This is because of how Java implements regex.
- This is due to the Java compiler.
- System T sorts documents by length for processing, so the first document is the longest.
Question 2: Which of the following is NOT a best practice for writing AQL?
- Use the AQL profiler to find and address hot spots.
- Follow simple rules of thumb when writing AQL.
- Don’t hand-tune while writing AQL.
- Always ignore throughput levels when designing extractors.
Question 3: Why is it necessary to be selective about performance tuning?
- It might adversely affect code readability
- It might reduce the quality of your results
- It might make your code more difficult to maintain
- A and B.
- A and C.
Text Analytics 101 Final Exam Answers
Question 1: Identify the logical sequence of phases in an IE system.
- Entity Identification > Feature Selection > Entity Resolution
- Entity Identification > Entity Resolution > Feature Selection
- Feature Selection > Entity Resolution > Entity Identification
- Feature Selection > Entity Identification > Entity Resolution
Question 2: Consider a set of news articles that contains 100 mentions of organizations. From this source, an extractor extracts 75 entities, 50 of which are correct. What are the Precision (P) and Recall (R) values of this extractor?
- P = 0.75, R = 0.50
- P = 0.67, R = 1.50
- P = 0.67, R = 0.50
- P = 0.50, R = 0.67
Question 3: What problem is caused by an IE system having a rigid matching priority?
- Regular expressions cannot be used when specifying rules.
- There is no support for matching strings spanning more than one token.
- The system cannot express aggregation operations.
- When multiple rules match the same region of text, mistakes are likely to occur.
Question 4: The System T consolidate policy:
- Applies a filtering predicate to output tuples.
- Specified how to handle tuples with overlapping spans.
- Specifies which tuple columns to group on.
- Specifies a tuple ordering.
Question 5: Which of the following AQL statements uses expressions, dictionaries, and sequence patterns to perform extraction?
- Relational style statement.
- Extract statement.
- Create table statement.
- Select statement.
Question 6: Which of the following statements are part of an AQL file?
- Create external table statements.
- Import statements.
- Create external dictionary statements.
- Export statements.
- All of the above.
Question 7: Which of the following types is a return value for table UDFs?
- Tuples.
- Integer.
- Span.
- Boolean.
Question 8: Which predicate would you use to check if a span is exactly equal to one of a predefined set of words?
- FollowsTok.
- MatchesRegex.
- MatchesDict.
- ContainsDict.
Question 9: Why is correct text tokenization important?
- Dictionary evaluation and many extraction operators, such as regex, are done on token boundaries, and incorrect tokenization will lead to incorrect results.
- Several built-in predicates and functions are token sensitive.
- AQL extract statements will not compile if tokenization is incorrect.
- A and B.
- A and C.
Question 10: Which of the following is NOT a best practice rule of thumb to follow when writing AQL?
- Use dictionaries instead of regex whenever possible.
- Make sure each module has its own copy of every dictionary.
- Avoid using UDFs as join predicates.
- Avoid Cartesian products.