Week 11 UAS

Question 1

The difference between Temporal-Difference Learning and Adaptive Dynamic Programming?

Answer 1

• TDL is to match estimation utility value of state with follow state to reach and makes one matching the value of utility estimation of each transition state.
• ADP is to match estimation utility value of state with all sate to reach and makes some matching value needed for consistent between utility estimation value with environment modelling.

Question 2

Approaches in Active Reinforcement Learning?

Answer 2

1. Reflex agent: learn policy directly, function
   mapping from states to actions
2. Q-learning: learns an action-value function, or Q function, giving the expected utility of taking a given action in a given state
3. Utility-based agent: learn utility values for states, use it to select actions that maximize the expected outcome utility

Question 3

The agent in Active Reinforcement Learning can get profit value from 2 method

Answer 3

1. Greedy approach: maximize utility by using estimation model
2. Wacky approach: random for all environment

Question 4

What is Direct Utility Estimation?

Answer 4

Using least mean squares concept (reward to go) which use Bellman Equations

Question 5

What is Active reinforcement Learning?

Answer 5

learn about optimizing mapping from states and actions

Question 6

What is Temporal Difference Learning?

Answer 6

update utility value appropriately which is affected by final state

Question 7

What is Natural Language Processing?

Answer 7

The Agent which want to add the information needs to understand (at least partially) of the human language which is sometimes ambiguous and unclear.

Question 8

3 ways to find information based on information retrieval perspective in NLP?

Answer 8

1. Text classification
2. information retrieval
3. information extraction

Question 9

N-gram character models

Answer 9

N-gram character models is defined as a Markov chain of order that in a Markov chain the probability of character depends only on the immediately preceding characters, not on any other characters.

Question 10

Smoothing approach N-gram models?

Answer 10

Linear Interpolation smoothing (backoff model) combines unigram, bigram, and trigram using linear interpolation.

Question 11

Text classification also known as?

Answer 11

Categorization: given a text of some kind, decide which of a predefined set of classes it belongs to. Language identification and genre classification are examples of text classification, as is sentiment analysis (classifying a movie or product review as positive or negative) and spam detection (classifying an email message as spam or not-spam).
can be done with Naive bayes.

Question 12

What is Information retrieval?

Answer 12

Information retrieval (searching information like Google) is the task of finding documents that are relevant to a user’s need for information. The best-known examples of information retrieval systems are search engines on the World Wide Web.
characteristics:
1. A collection of writings (document).
The system must determine which one want to be considered as a document (paper). Example: a paragraph, a page, etc.
2. User Query
The query is a formula used to find the information needed by the user.
In its simplest form, a query is a keyword and documents that contain the keywords are the searched documents.
3. Set of Results
The results from the queries. A part of the documents
in which is relevant to the query.
4. Display of result sets
Can be a list of results in a ranking of the title documents.

previously works with boolean models

Question 13

What is Information extraction?

Answer 13

Information extraction is the process of acquiring knowledge by skimming a text and looking for occurrences of a particular class of object and for relationships among objects.
The simplest type of information extraction system is an attribute-based extraction system TEMPLATE REGULAR EXPRESSION that assumes that the entire text refers to a single object and the task is to extract attributes of that object.

Question 14

4 approximation for Information Extraction?

Answer 14

– Deterministic to stochastic
– Domain-spesific to general
– Hand-crafted to learned
– Small-scale to large-scale

Question 15

Relational-based extraction system?

Answer 15

FASTUS.
divided to be 5 stages:
1. Tokenization: divide the characters into a token
2. Complex-word handling: handle words that contains grammatical rules
3. Basic-group handling: sort by morphological of the words
4. Complex-phrase handling: merge basic group to be an arrangement of words
5. Structure merging: combining structure that have been resulted

Question 16

What is Machine Translation?

Answer 16

System that could read on its own and build up its own database. Such a system would he relation-independent; would work for any relation.

Question 17

A representative machine-reading system is

Answer 17

TEXTRUNNER

Question 18

Information Retrieval applications?

Answer 18

1. PageRank Algorithm
2. HITS (Hyperlink-Induced Topic Search) Algorithm
3. Question Answering, when the query type is questions, then the result is not a ranking list of the documents, but the form of short response, it could be a sentence or phrase.