| Date | November 2019 | Marks available | 2 | Reference code | 19N.2.HL.TZ0.8 |
| Level | HL | Paper | 2 | Time zone | no time zone |
| Command term | Outline | Question number | 8 | Adapted from | N/A |
Question
Optical character recognition (OCR) is a method where printed text or handwritten text is converted to machine-encoded text.
Recognizing handwritten characters presents more of a problem because people have different handwriting styles.
For example, the digitized handwritten letter X in Figure 3 does not exactly match the digitized letter X in Figure 4.
Artificial neural networks (ANNs) can be used to assist with recognizing handwritten characters that do not exactly match the expected pattern.
Predictive text, where the computer predicts the next word in the sentence, can be programmed to utilize a neural network.
The sentence “The child is feeling” is entered into an application that uses predictive text and three options are suggested: better, like, a. Upon entering the two characters “hu” the word hungry is suggested.
Outline one problem that may lead to printed text characters not being detected correctly.
Outline why an ANN can be used to overcome the challenges outlined in this scenario.
Explain how ANN pattern recognition techniques are applied to ensure that the handwritten letter X in Figure 3 is recognized as a letter X.
Identify two features that would be required by the ANN to predict the next word in the sentence.
Explain how the application uses a neural network to suggest suitable words.
Outline two potential problems with training the ANN to suggest appropriate words.
Markscheme
Award [2 max].
Printed text quality may be poor or faint/may use an unusual typeface e.g. BrushScript/In certain typefaces some characters can look similar e.g. sans serif;
For example, 5 and S may be confused/Numbers may be confused with letters 1 or I;
Award [2 max].
ANN can be trained to recognise handwriting styles / supervised learning;
ANN can learn to apply existing knowledge to new handwriting styles / unsupervised learning;
ANN can recognize parts of the image to determine if it is a match / doesn’t require the entire image to be a match;
Award [4 max].
The ANN accesses a database of correctly stored characters;
ANN breaks the image into smaller parts / applies a filter (2 or 3 pixels) to a section of the image;
Calculate whether the pixels match / multiple each image pixel to the feature pixel, add them up, divide by number of pixels / perfect match will be 1;
Apply convolution / repeat the application of the filter over and over;
Apply a ReLU (Rectified Linear Units) layer (remove negative values) to reduce the mathematical calculations;
Apply pooling to shrink the image stack;
Use the stack of filter images / convolution layer to see if the image is a match;
Award [2 max].
Dictionary of words;
Predictive text algorithms;
The ability to recall the previous words in the sentence / memory;
An understanding of language structure e.g. nouns, verbs, adverbs, etc.;
Previous words must be added back into the neural network;
Award [6 max].
If they have used a diagram should include:
Award [1] Input for previous words;
Award [1] Input for new characters the user enters;
Award [1] Hidden layers;
Award [1] Weights inputted into the network;
Award [1] Combining the two inputs;
Award [1] Non-linear regression / sigmoid function;
Award [1] Merge layers to produce output;
Award [1] Back propagation / Output that re-enters the ANN;
Award [4 max].
Unsupervised learning may have a poor set of words / text-speak is used;
So the ANN database doesn’t contain text-speak words so doesn’t learn;
Vanishing gradient problem / Exploding gradient problem;
Gradient signal is multiplied many times by weight matrix / The gradient signal can become smaller at every training step (vanishing) / can become excessively large at every training step (exploding) / This can make learning very slow or stops it completely;
Examiners report
Syllabus sections
- 17N.2.HL.TZ0.8c.i: State the number of neurons that would be in the input layer.
- 17N.2.HL.TZ0.8a: Explain one reason why ANNs are suitable for solving the types of problems found in these areas.
- 17N.2.HL.TZ0.8c.ii: State the number of neurons that would be in the output layer.
-
19N.2.HL.TZ0.8a:
Outline one problem that may lead to printed text characters not being detected correctly.
-
19N.2.HL.TZ0.8e:
Explain how the application uses a neural network to suggest suitable words.
-
19N.2.HL.TZ0.8c:
Explain how ANN pattern recognition techniques are applied to ensure that the handwritten letter X in Figure 3 is recognized as a letter X.
-
19M.2.HL.TZ0.8a:
Describe the difference between a genetic algorithm and a neural network.
-
19M.2.HL.TZ0.8d:
Explain why the machine learning capabilities of A Doll Called Alicia may lead to instances when the child and the doll cannot communicate effectively.
-
18M.2.HL.TZ0.8a:
Outline the basic steps involved in supervized learning algorithms.
-
18M.2.HL.TZ0.9b:
Discuss the appropriateness of using genetic algorithms to compare the processed images with those stored in the knowledge base.
-
19M.2.HL.TZ0.8b:
Identify two ways in which the neural network could be modified that may improve its performance.
-
19M.2.HL.TZ0.8c:
Describe the difference between supervised learning and unsupervised learning.
-
18N.2.HL.TZ0.8c:
Many companies now use an automated response system when answering basic telephone queries from customers. This system uses speech analysis together with recorded messages.
Explain why the content of the messages given by these systems must be carefully chosen.
- 18N.2.HL.TZ0.8a: Genetic algorithms follow an iterative process in developing a solution to a...
- 18N.2.HL.TZ0.8b.i: Outline why neural networks need to be trained before they can be used in this analysis.
-
18N.2.HL.TZ0.8b.ii:
Explain why supervised learning may be preferred to unsupervised learning for training these networks to make predictions on the progress of this parasitic disease.
-
18M.2.HL.TZ0.10:
Machine translators are regularly used to translate text from one language to another. In spite of the advances that have been made in this field, the output may still need to be proofread by a human.
Describe two problems that these translators may encounter when translating from one language to another.
-
20N.2.HL.TZ0.8a:
Outline two advantages of developing NPCs using genetic algorithms.
-
17N.2.HL.TZ0.8b:
Describe an appropriate set of data that could be used to train the network.
-
17N.2.HL.TZ0.8d:
Explain the way in which the output from neuron D will be determined.
-
20N.2.HL.TZ0.8b:
Explain how supervised learning of a genetic algorithm could be used to support the way NPCs learn.
-
18M.2.HL.TZ0.8b:
The following block diagram illustrates a neural network used by a supervized learning algorithm to calculate a resulting value RES, for a given set of inputs.
The neural network in the block diagram contains 60 nodes in the input layer (block I), 36 nodes in the hidden layer (block H) and 6 nodes in the outermost layer (block O). The solid black arrows indicate that all nodes in one layer are feeding all nodes in the successive layer. In particular, the value RES is calculated by using all nodes in O.
Calculate the number of weights that the above neural network uses in producing a value for RES. Show the details of your calculation.
-
17N.2.HL.TZ0.9:
Google Translate is an algorithm whose function is to translate text from one language to another. One of the resources that it uses is the body of documents produced by the United Nations which are routinely translated by humans into various languages.
Discuss the reasons why Google Translate takes a probabilistic approach in preference to a cognitive rule-based approach.
-
19N.2.HL.TZ0.8d:
Identify two features that would be required by the ANN to predict the next word in the sentence.
-
19N.2.HL.TZ0.8f:
Outline two potential problems with training the ANN to suggest appropriate words.
-
17N.2.HL.TZ0.8e:
The network is set up with initial values. The outputs are compared with the desired outputs.
Identify the steps that now take place to train the network.
-
20N.2.HL.TZ0.8d:
The last stage of NPC development is verbal communication with player characters and other NPCs. To assist with this process, it was decided to explore research related to chatbots.
Explain the benefits of giving NPCs the functionality of chatbots.
-
20N.2.HL.TZ0.8c:
Describe the way an NPC in a game could adapt its behaviour when moving and interacting with its environment.
-
19M.2.HL.TZ0.8e:
Companies such as MAGS are considering products that use unsupervised learning rather than supervised learning.
Explain the benefits of unsupervised learning in developing products such as A Doll Called Alicia.
