Document (aika 0.16 API)

java.lang.Object
- org.aika.Document

All Implemented Interfaces:

Comparable<Document>
```
public class Document
extends Object
implements Comparable<Document>
```
The Document class represents a single document which may be either used for processing a text or as training input. A document consists of the raw text, the interpretations and the activations.
When the document is not needed any more, the method clearActivations must be called, since Aika only supports a single document per thread and model.

Author:

Lukas Molzberger

Nested Class Summary

Nested Classes
Modifier and Type Class and Description

class Document.Queue

class Document.UpperBoundQueue

class Document.ValueQueue

Nested Classes
Modifier and Type	Class and Description
`class`	`Document.Queue`
`class`	`Document.UpperBoundQueue`
`class`	`Document.ValueQueue`

Field Summary

Fields
Modifier and Type	Field and Description
`TreeSet<INeuron>`	`activatedNeurons`
`TreeSet<Node>`	`activatedNodes`
`int`	`activationIdCounter`
`static Comparator<NodeActivation>`	`ACTIVATIONS_OUTPUT_COMPARATOR`
`TreeMap<NodeActivation.Key,Activation>`	`activationsByRangeBegin`
`TreeMap<NodeActivation.Key,Activation>`	`activationsByRangeEnd`
`TreeMap<NodeActivation.Key,Activation>`	`activationsByRid`
`ArrayList<Activation>`	`addedActivations`
`ArrayList<NodeActivation>`	`addedNodeActivations`
`TreeSet<Node>`	`addedNodes`
`ArrayList<Candidate>`	`candidates`
`static int`	`CLEANUP_INTERVAL`
`long`	`createV`
`TreeSet<INeuron>`	`finallyActivatedNeurons`
`int`	`id`
`static boolean`	`INCREMENTAL_MODE` Experimental code: not working yet!
`TreeSet<Activation>`	`inputNeuronActivations`
`static int`	`MAX_ROUND`
`Model`	`model`
`TreeMap<INeuron,Set<Synapse>>`	`modifiedWeights`
`Document.Queue`	`queue`
`int`	`searchNodeIdCounter`
`int`	`searchStepCounter`
`SearchNode`	`selectedSearchNode`
`SupervisedTraining`	`supervisedTraining`
`int`	`threadId`
`Document.UpperBoundQueue`	`ubQueue`
`long`	`visitedCounter`
`Document.ValueQueue`	`vQueue`

Constructor Summary

Constructors
Constructor and Description

Document(int id, String content, Model model, int threadId)

Constructors
Constructor and Description
`Document(int id, String content, Model model, int threadId)`

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`String`	`activationsToString()`
`String`	`activationsToString(boolean finalOnly, boolean withTextSnippet, boolean withLogic)`
`void`	`clearActivations()` Removes the activations of this document from the model again.
`void`	`commit()` Updates the model after the training step.
`int`	`compareTo(Document doc)`
`void`	`dumpDebugCandidateStatistics()`
`void`	`dumpOscillatingActivations()`
`void`	`generateCandidates()`
`String`	`generateOutputText()`
`Stream<Activation>`	`getActivations()`
`Stream<NodeActivation>`	`getAllActivationsStream()`
`String`	`getContent()`
`Stream<Activation>`	`getFinalActivations()`
`String`	`getText(Range r)`
`int`	`length()`
`void`	`notifyWeightModified(Synapse synapse)`
`void`	`process()` The method `process` needs to be called after all the input activations have been added to the network.
`void`	`process(Long timeoutInMilliSeconds)`
`void`	`propagate()`
`String`	`toString()`

Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

CLEANUP_INTERVAL
```
public static int CLEANUP_INTERVAL
```

MAX_ROUND
```
public static int MAX_ROUND
```

INCREMENTAL_MODE
```
public static boolean INCREMENTAL_MODE
```
Experimental code: not working yet!

id
```
public final int id
```

visitedCounter
```
public long visitedCounter
```

activationIdCounter
```
public int activationIdCounter
```

searchNodeIdCounter
```
public int searchNodeIdCounter
```

searchStepCounter
```
public int searchStepCounter
```

model
```
public Model model
```

threadId
```
public int threadId
```

queue
```
public Document.Queue queue
```

vQueue
```
public Document.ValueQueue vQueue
```

ubQueue

public Document.UpperBoundQueue ubQueue

activatedNodes
```
public TreeSet<Node> activatedNodes
```

activatedNeurons

public TreeSet<INeuron> activatedNeurons

finallyActivatedNeurons

public TreeSet<INeuron> finallyActivatedNeurons

inputNeuronActivations

public TreeSet<Activation> inputNeuronActivations

modifiedWeights

public TreeMap<INeuron,Set<Synapse>> modifiedWeights

supervisedTraining

public SupervisedTraining supervisedTraining

activationsByRangeBegin

public TreeMap<NodeActivation.Key,Activation> activationsByRangeBegin

activationsByRangeEnd

public TreeMap<NodeActivation.Key,Activation> activationsByRangeEnd

activationsByRid

public TreeMap<NodeActivation.Key,Activation> activationsByRid

addedNodes
```
public TreeSet<Node> addedNodes
```

addedNodeActivations

public ArrayList<NodeActivation> addedNodeActivations

addedActivations

public ArrayList<Activation> addedActivations

selectedSearchNode
```
public SearchNode selectedSearchNode
```

candidates
```
public ArrayList<Candidate> candidates
```

createV
```
public long createV
```

ACTIVATIONS_OUTPUT_COMPARATOR

public static Comparator<NodeActivation> ACTIVATIONS_OUTPUT_COMPARATOR

Constructor Detail

Document

public Document(int id,
                String content,
                Model model,
                int threadId)

Method Detail

getContent
```
public String getContent()
```

length
```
public int length()
```

toString
```
public String toString()
```
Overrides:

toString in class Object

getText
```
public String getText(Range r)
```

getFinalActivations

public Stream<Activation> getFinalActivations()

getActivations

public Stream<Activation> getActivations()

compareTo
```
public int compareTo(Document doc)
```
Specified by:

compareTo in interface Comparable<Document>

propagate
```
public void propagate()
```

generateCandidates
```
public void generateCandidates()
```

process
```
public void process()
```
The method process needs to be called after all the input activations have been added to the network. It performs the search for the best interpretation.

process

public void process(Long timeoutInMilliSeconds)
             throws SearchNode.TimeoutException

Throws:: SearchNode.TimeoutException

dumpDebugCandidateStatistics

public void dumpDebugCandidateStatistics()

notifyWeightModified

public void notifyWeightModified(Synapse synapse)

commit
```
public void commit()
```
Updates the model after the training step. It applies the weight and bias delta values and reflects the changes in the logic node structure.

clearActivations
```
public void clearActivations()
```
Removes the activations of this document from the model again.

generateOutputText
```
public String generateOutputText()
```

activationsToString
```
public String activationsToString()
```

activationsToString

public String activationsToString(boolean finalOnly,
                                  boolean withTextSnippet,
                                  boolean withLogic)

getAllActivationsStream

public Stream<NodeActivation> getAllActivationsStream()

dumpOscillatingActivations

public void dumpOscillatingActivations()

Class Document

Nested Class Summary

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

CLEANUP_INTERVAL

MAX_ROUND

INCREMENTAL_MODE

id

visitedCounter

activationIdCounter

searchNodeIdCounter

searchStepCounter

model

threadId

queue

vQueue

ubQueue

activatedNodes

activatedNeurons

finallyActivatedNeurons

inputNeuronActivations

modifiedWeights

supervisedTraining

activationsByRangeBegin

activationsByRangeEnd

activationsByRid

addedNodes

addedNodeActivations

addedActivations

selectedSearchNode

candidates

createV

ACTIVATIONS_OUTPUT_COMPARATOR

Constructor Detail

Document

Method Detail

getContent

length

toString

getText

getFinalActivations

getActivations

compareTo

propagate

generateCandidates

process

process

dumpDebugCandidateStatistics

notifyWeightModified

commit

clearActivations

generateOutputText

activationsToString

activationsToString

getAllActivationsStream

dumpOscillatingActivations