com.mayabot.nlp.segment.wordnet.Wordnet Maven / Gradle / Ivy
/*
* Copyright 2018 mayabot.com authors. All rights reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.mayabot.nlp.segment.wordnet;
import com.mayabot.nlp.common.utils.CustomCharSequence;
import com.mayabot.nlp.segment.common.VertexHelper;
import org.jetbrains.annotations.NotNull;
import java.util.Arrays;
import java.util.BitSet;
import java.util.HashMap;
import java.util.Map;
import java.util.function.Consumer;
/**
*
* #S# [0] #Start 品 [1] 品质 质 [2] 质 和 [3] {和、和服} 服 [4] 服务 务 [5] 务 #E# [6] #End
*
*
* sindex表示智能下标索引，用 -1 表示Start length表示End
*
* WordNodeLinkedList 是一个优化过的linkedlist，去除了中间node节点,并增量排序和特殊性质
*
* 实现CharSequence接口，可以进行当做字符串类型处理
* O
*
* @author jimichan
*/
public final class Wordnet implements CharSequence {
/**
* 节点，每一行都是前缀词，和 char数字一一对应
*/
private final VertexRow[] slotList;
/**
* 启始行 下标 -1
*/
private final VertexRow begin;
/**
* 结尾行 下标 charSize
*/
private final VertexRow end;
/**
* 原始句子对应的数组
*/
final char[] charArray;
/**
* 字符的数量
*/
private final int charSize;
/**
* 3.1.0 新增，存储上下文
*/
private Map context = null;
private Wordnet(Wordnet parent, int from, int length) {
this.charArray = Arrays.copyOfRange(parent.charArray, from, from + length);
this.charSize = length;
this.begin = new VertexRow(-1, this);
this.end = new VertexRow(charSize, this);
slotList = new VertexRow[charSize];
for (int i = 0; i < length; i++) {
VertexRow row = new VertexRow(i, this);
for (Vertex v : parent.slotList[from + i]) {
row.getOrCrete(v.length);
}
slotList[i] = row;
}
getBeginRow().put(VertexHelper.newBegin());
getEndRow().put(VertexHelper.newEnd());
}
/**
* 构建一个空的网，槽的数量是charArray.length
*
* @param charArray 字符数组
*/
public Wordnet(char[] charArray) {
this.charArray = charArray;
this.charSize = charArray.length;
this.begin = new VertexRow(-1, this);
this.end = new VertexRow(charSize, this);
// 创建一个空的数组
slotList = new VertexRow[charSize];
//初始化数组里面的对象,提前初始化好
for (int i = 0; i < charSize; i++) {
slotList[i] = new VertexRow(i, this);
}
getBeginRow().put(VertexHelper.newBegin());
getEndRow().put(VertexHelper.newEnd());
}
public Wordnet subWordnet(int from, int length) {
return new Wordnet(this, from, length);
}
/**
* 补齐空洞，这样viterbi总是能走到底
*/
public void fillNill() {
for (VertexRow row : slotList) {
if (row.first == null) {
Vertex vertex = row.getOrCrete(1);
vertex.freq = 1;
}
}
}
/**
* 寻找没有被所有路径覆盖的位置
*
* @return 返回 false的位置是没有覆盖掉的
*/
public BitSet findNoOverWords() {
BitSet noOverWords = new BitSet();
for (int i = 0; i < charSize; i++) {
VertexRow row = slotList[i];
if (row != null) {
Vertex p = row.first();
while (p != null) {
noOverWords.set(row.rowNum, row.rowNum + p.length);
p = p.next;
}
}
}
return noOverWords;
}
/**
* 寻找 存在跳转到当前row但是当前row没有跳出节点。
* 或者 根本不存在跳转到当前行的路径(前置条件是没有被路径覆盖)
* 寻找 悬空行，会导致路径中断
*
* @return bitset 孤悬节点
*/
public BitSet findDangling() {
BitSet bitSet = new BitSet(charSize);
BitSet noOverWords = findNoOverWords();
//第一行肯定是跳入的
bitSet.set(0);
for (int i = 0; i < charSize; i++) {
VertexRow row = slotList[i];
if (row != null) {
Vertex p = row.first();
while (p != null) {
noOverWords.set(row.rowNum + p.length);
p = p.next;
}
}
}
for (int i = 0; i < charSize; i++) {
//如果没人跳入，而且也没被覆盖
if (!(bitSet.get(i) || noOverWords.get(i))) {
bitSet.set(i);
}
}
//bit set 里面为 false的，那么肯定是孤力
return null;
}
/**
* 如果对应行不存在，那么会自动创建
*
* @param sindex index
* @return VertexRow
*/
public final VertexRow getRow(int sindex) {
return indexAt(sindex);
}
public final VertexRow row(int sindex) {
return indexAt(sindex);
}
/**
* 节点数字，包含了Start，end两个标记节点 动态统计,调用的时候请注意
*
* @return 包含了Start，end两个标记节点的数量
*/
public int size() {
int count = 0;
for (int i = charSize - 1; i >= 0; i--) {
VertexRow r = slotList[i];
if (r != null) {
count += r.size();
}
}
count += this.begin.size();
count += this.end.size();
return count;
}
/**
* 返回第多少行的链表
*
*
* length = 5
* sloat length = 7
*
* -1 0 1 2 3 4 5 优化过的智能下标
* 0 1 2 3 4 char array 下标
* 0 1 2 3 4 5 6 sloat数组下标
*
*
* @param sindex SmartIndex -1表示Start
* @return VertexRow
*/
private VertexRow indexAt(int sindex) {
if (sindex == -1) {
return begin;
}
if (sindex == charSize) {
return end;
}
return slotList[sindex];
}
/**
* 非空的行数
*
* @return 非空的行数
*/
public int notNullRowNums() {
int count = 0;
for (int i = charSize - 1; i >= 0; i--) {
VertexRow r = slotList[i];
if (r != null && !r.isEmpty()) {
count++;
}
}
if (!this.begin.isEmpty()) {
count++;
}
if (!this.end.isEmpty()) {
count++;
}
return count;
}
////////////// 顶点操作////////////////
/**
* 添加顶点,重复添加就忽略 返回被替换的节点
*
* @param charOffset sindex 下标和char对应
* @param vertex 顶点
* @return 返回被替换的节点, 新增节点返回null
*/
public Vertex put(int charOffset, Vertex vertex) {
return getRow(charOffset).put(vertex);
}
/**
* put一个 ，但是返回的是一个最新的Vertext对象，然后可以继续设置属性
*
* @param offset 偏移量
* @param length 长度
* @return 返回的是一个最新的Vertex对象
*/
public Vertex put(int offset, int length) {
Vertex vertex = new Vertex(length);
getRow(offset).put(vertex);
return vertex;
}
/**
* 行首节点
*
* @param sindex
* @return Vertex
*/
public Vertex getRowFirst(int sindex) {
VertexRow row = row(sindex);
return row.getFirst();
}
/**
* 获取某一行长度为length的节点 没有就返回null
*
* @param sindex
* @param length
* @return Vertex
*/
public Vertex getVertex(int sindex, int length) {
VertexRow row = row(sindex);
return row.get(length);
}
/**
* 检查是否包含 offset-length的vertext
*
* @param sindex
* @param length
* @return boolean true not contain
*/
public boolean isNotContains(int sindex, int length) {
VertexRow row = row(sindex);
return !row.contains(length);
}
/**
* 返回一行里面有几个节点
*
* @param sindex
* @return int size
*/
public int sizeInRow(int sindex) {
VertexRow row = row(sindex);
if (row == null) {
return 0;
}
return row.size();
}
/**
* 访问网络里面所有的Vertex节点. 从后向前了
*
* @param consumer
*/
public final void accessAllVertex(Consumer consumer) {
for (int i = slotList.length - 1; i >= 0; i--) {
VertexRow row = slotList[i];
if (row != null) {
for (Vertex v = row.first(); v != null; v = v.next()) {
consumer.accept(v);
}
}
}
}
/**
* 根据当前的最优路径，设定已经选择的最优路径中的Vertex的最优网络标记为true
*/
@Override
public String toString() {
return new WordNetToStringBuilder(this, false).toString();
}
public String toMoreString() {
return new WordNetToStringBuilder(this, true).toString();
}
public VertexRow getBeginRow() {
return begin;
}
public VertexRow getEndRow() {
return end;
}
/**
* 比如字符串长度为5，那么这个length返回5
*
* @return char size
*/
public int getCharSizeLength() {
return charSize;
}
/**
* 原始的
*
* @return chars
*/
public @NotNull
char[] getCharArray() {
return charArray;
}
@Override
public int length() {
return charArray.length;
}
@Override
public char charAt(int index) {
if (index < 0) {
return ' ';
}
if (index >= charSize) {
return ' ';
}
return charArray[index];
}
@Override
public CharSequence subSequence(int start, int end) {
return new CustomCharSequence(charArray, start, end - start);
}
public VertexRow[] getSlotList() {
return slotList;
}
public T get(String key){
if (context == null) {
return null;
}else{
return (T) context.get(key);
}
}
public void set(String key, Object value) {
if(context == null){
context = new HashMap<>();
}
context.put(key, value);
}
}