what is tokenization give an example in english were

Post New Homework

1. Why don't we use grep for information retrieval?

2. Why don't we use a relational database for information retrieval?

3. Google does not always interpret the query as a boolean conjunction of its terms. Give examples.

4. What is a term-document incidence matrix?

5. In constructing the index, which step is most expensive/complex?

6. Complex Boolean retrieval systems like Westlaw use many operations that go beyond strictly Boolean

7. operators. Name some of them.

8. Define the number of types/tokens in a sentence.

9. An IR system can normalize terms by defining equivalence classes. E.g., "suit" and "suits" could be

10. in an equivalence class. What is the limitation of this model in IR?

11. What is tokenization?

12. Give an example in English were tokenization is nontrivial

13. What is a stop list?

14. What is lemmatization? Give an example.

15. What is stemming? Give an example that is not also a lemmatization example.

16. Name a particular stemmer.

17. Give an example of a pair of words that a typical stemmer would put in one equivalence class and we

18. would expect improved performance of the IR system.

19. Give an example of a pair of words that a typical stemmer would put in one equivalence class and we

20. would expect decreased performance of the IR system.

21. Name two data structures that support phrase queries.

22. Name a data structure that supports proximity queries.

23. Which data structures are typically used for locating the entry for a term in the dictionary?

24. Which data structure is best used for locating the entry for a term in the dictionary if the collection

25. is static?

26. Which data structure is best used for locating the entry for a term in the dictionary if prefix search

27. must be supported?

28. Which special strings are stored in the permuterm index for the word "car"?

29. What sequence of letters is looked up in the permuterm index for the following wildcard queries?

30. X, X*, *X, *X*, X*Y

31. What is the difference between the regular inverted index used in IR and the k-gram index?

32. Give an example of a query that cannot be corrected using isolated-word spelling correction.

33. Define Levenshtein edit distance.

34. Define Damerau-Levenshtein edit distance.

35. Give the formula for Zipf's law.

36. Give the formula for Heaps' law.

37. What is the feast or famine problem?

38. Define the Jaccard coefficient

39. What is the bag of words model?

40. What is the advantage of idf weighting compared to inverse-collection-frequency weighting?

41. What is the tf-idf weight of term t in document d?

42. What is the relationship between term frequency and collection frequency?

43. Why don't we use Euclidean distance of tf-idf vectors to rank documents with respect to a query?

44. Write down the formula for cosine similarity between query q and document d.

45. Explain the notation ddd.qqq

46. What is the advantage of pivot normalization compared to regular cosine normalization?

47. What is document-at-a-time processing?

48. What index organization does document-at-a-time processing require?

49. What is term-at-a-time processing?

50. What data structure does term-at-a-time processing require that document-at-a-time processing does

51. not require?

52. What is a tiered inverted index?

53. Name two criteria that can be used for deciding as to whether to put a document d in tier 1 of a tiered

54. index.

55. Name three criteria for evaluating a search engine.

56. What are the components of an information retrieval benchmark?

57. What is the difference between the concepts of query and information need?

58. Define precision

59. Define recall

60. Define F1

61. What is the harmonic mean of two numbers?

62. Why is F1 defined as the harmonic mean?

63. What is an easy way of maximizing the recall of an IR engine?

64. What is an easy way of maximizing the precision of an IR engine?

65. What is a precision-recall curve?

66. An evaluation benchmark ideally should tell us for any document-query pair whether the document is

67. relevant to the query. Why is Cranfield the only collection that actually satisfies this desideratum?

68. Define the kappa measure

69. What is the minimum and maximum of the kappa measure?

70. What is the significance of kappa being less than / greater than 0?

71. What is A/B testing?

72. What does marginal relevance measure?

73. What distinguishes a dynamic from a static summary?

74. What is a simple heuristic for computing a dynamic summary if you can display n characters?

Post New Homework
Captcha

Looking tutor’s service for getting help in UK studies or college assignments? Order Now