Document (aika 0.24 API)

java.lang.Object
- network.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

static class Document.ActKey

class Document.Queue

class Document.UpperBoundQueue

class Document.ValueQueue

Nested Classes
Modifier and Type	Class and Description
`static class`	`Document.ActKey`
`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<Activation>`	`ACTIVATIONS_OUTPUT_COMPARATOR`
`TreeMap<Integer,Activation>`	`activationsById`
`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`
`Linker`	`linker`
`int`	`logicNodeActivationIdCounter`
`static int`	`MAX_ROUND`
`Model`	`model`
`TreeMap<INeuron,Set<Synapse>>`	`modifiedWeights`
`int`	`positionIdCounter`
`TreeMap<Integer,Position>`	`positions`
`Document.Queue`	`queue`
`static int`	`ROUND_LIMIT`
`int`	`searchNodeIdCounter`
`TreeMap<Integer,Double>`	`searchNodeWeights`
`int`	`searchStepCounter`
`SearchNode`	`selectedSearchNode`
`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)`
`String`	`activationsToString(boolean finalOnly, boolean withTextSnippet, boolean withLogic)`
`void`	`addActivation(Activation act)`
`void`	`append(String txt)`
`char`	`charAt(int i)`
`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()`
`Collection<Activation>`	`getActivations(boolean onlyFinal)`
`Collection<Activation>`	`getActivationsByPosition(int fromSlot, Position fromPos, boolean fromInclusive, int toSlot, Position toPos, boolean toInclusive)`
`Collection<Activation>`	`getActivationsByPosition(Position fromPos, boolean fromInclusive, Position toPos, boolean toInclusive)`
`Stream<NodeActivation>`	`getAllActivationsStream()`
`String`	`getContent()`
`Activation`	`getNextActivation(Activation currentAct)`
`int`	`getNumberOfActivations()`
`String`	`getText(Integer begin, Integer end)`
`String`	`getText(Position begin, Position end)`
`int`	`length()`
`Position`	`lookupFinalPosition(int pos)`
`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
```

ROUND_LIMIT
```
public static int ROUND_LIMIT
```

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
```

logicNodeActivationIdCounter

public int logicNodeActivationIdCounter

searchNodeIdCounter
```
public int searchNodeIdCounter
```

searchStepCounter
```
public int searchStepCounter
```

positionIdCounter
```
public int positionIdCounter
```

model
```
public Model model
```

threadId
```
public int threadId
```

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

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

ubQueue

public Document.UpperBoundQueue ubQueue

linker
```
public Linker linker
```

positions

public TreeMap<Integer,Position> positions

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

searchNodeWeights

public TreeMap<Integer,Double> searchNodeWeights

activationsById

public TreeMap<Integer,Activation> activationsById

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

addedNodeActivations

public ArrayList<NodeActivation> addedNodeActivations

selectedSearchNode
```
public SearchNode selectedSearchNode
```

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

createV
```
public long createV
```

ACTIVATIONS_OUTPUT_COMPARATOR

public static Comparator<Activation> ACTIVATIONS_OUTPUT_COMPARATOR

Constructor Detail

Document

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

Method Detail

append
```
public void append(String txt)
```

charAt
```
public char charAt(int i)
```

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

length
```
public int length()
```

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

toString in class Object

lookupFinalPosition

public Position lookupFinalPosition(int pos)

getText

public String getText(Position begin,
                      Position end)

getText

public String getText(Integer begin,
                      Integer end)

addActivation

public void addActivation(Activation act)

getActivations

public Collection<Activation> getActivations(boolean onlyFinal)

getActivationsByPosition

public Collection<Activation> getActivationsByPosition(int fromSlot,
                                                       Position fromPos,
                                                       boolean fromInclusive,
                                                       int toSlot,
                                                       Position toPos,
                                                       boolean toInclusive)

getActivationsByPosition

public Collection<Activation> getActivationsByPosition(Position fromPos,
                                                       boolean fromInclusive,
                                                       Position toPos,
                                                       boolean toInclusive)

getNextActivation

public Activation getNextActivation(Activation currentAct)

getNumberOfActivations
```
public int getNumberOfActivations()
```

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)

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

ROUND_LIMIT

INCREMENTAL_MODE

id

visitedCounter

activationIdCounter

logicNodeActivationIdCounter

searchNodeIdCounter

searchStepCounter

positionIdCounter

model

threadId

queue

vQueue

ubQueue

linker

positions

activatedNodes

activatedNeurons

finallyActivatedNeurons

inputNeuronActivations

modifiedWeights

searchNodeWeights

activationsById

addedNodes

addedNodeActivations

selectedSearchNode

candidates

createV

ACTIVATIONS_OUTPUT_COMPARATOR

Constructor Detail

Document

Method Detail

append

charAt

getContent

length

toString

lookupFinalPosition

getText

getText

addActivation

getActivations

getActivationsByPosition

getActivationsByPosition

getNextActivation

getNumberOfActivations

compareTo

propagate

generateCandidates

process

process

dumpDebugCandidateStatistics

notifyWeightModified

commit

clearActivations

generateOutputText

activationsToString

activationsToString

activationsToString

getAllActivationsStream

dumpOscillatingActivations