All Projects → shibing624 → crf-seg

shibing624 / crf-seg

Licence: Apache-2.0 license
crf-seg:用于生产环境的中文分词处理工具,可自定义语料、可自定义模型、架构清晰,分词效果好。java编写。

Programming Languages

java
68154 projects - #9 most used programming language

Labels

Projects that are alternatives of or similar to crf-seg

Torchnlp
Easy to use NLP library built on PyTorch and TorchText
Stars: ✭ 233 (+1692.31%)
Mutual labels:  crf
crfsuite-rs
Rust binding to crfsuite
Stars: ✭ 19 (+46.15%)
Mutual labels:  crf
NLP-paper
🎨 🎨NLP 自然语言处理教程 🎨🎨 https://dataxujing.github.io/NLP-paper/
Stars: ✭ 23 (+76.92%)
Mutual labels:  crf
Pytorch ner bilstm cnn crf
End-to-end Sequence Labeling via Bi-directional LSTM-CNNs-CRF implement in pyotrch
Stars: ✭ 249 (+1815.38%)
Mutual labels:  crf
keras-crf-ner
keras+bi-lstm+crf,中文命名实体识别
Stars: ✭ 16 (+23.08%)
Mutual labels:  crf
fastai sequence tagging
sequence tagging for NER for ULMFiT
Stars: ✭ 21 (+61.54%)
Mutual labels:  crf
Open Sesame
A frame-semantic parsing system based on a softmax-margin SegRNN.
Stars: ✭ 170 (+1207.69%)
Mutual labels:  crf
CRFasRNNLayer
Conditional Random Fields as Recurrent Neural Networks (Tensorflow)
Stars: ✭ 76 (+484.62%)
Mutual labels:  crf
xinlp
把李航老师《统计学习方法》的后几章的算法都用java实现了一遍,实现盒子与球的EM算法,扩展到去GMM训练,后来实现了HMM分词(实现了HMM分词的参数训练)和CRF分词(借用CRF++训练的参数模型),最后利用tensorFlow把BiLSTM+CRF实现了,然后为lucene包装了一个XinAnalyzer
Stars: ✭ 21 (+61.54%)
Mutual labels:  crf
deepseg
Chinese word segmentation in tensorflow 2.x
Stars: ✭ 23 (+76.92%)
Mutual labels:  crf
video-quality-metrics
Test specified presets/CRF values for the x264 or x265 encoder. Compares VMAF/SSIM/PSNR numerically & via graphs.
Stars: ✭ 87 (+569.23%)
Mutual labels:  crf
ChineseNER
中文NER的那些事儿
Stars: ✭ 241 (+1753.85%)
Mutual labels:  crf
Gumbel-CRF
Implementation of NeurIPS 20 paper: Latent Template Induction with Gumbel-CRFs
Stars: ✭ 51 (+292.31%)
Mutual labels:  crf
Pytorch Bert Crf Ner
KoBERT와 CRF로 만든 한국어 개체명인식기 (BERT+CRF based Named Entity Recognition model for Korean)
Stars: ✭ 236 (+1715.38%)
Mutual labels:  crf
korean ner tagging challenge
KU_NERDY 이동엽, 임희석 (2017 국어 정보 처리 시스템경진대회 금상) - 한글 및 한국어 정보처리 학술대회
Stars: ✭ 30 (+130.77%)
Mutual labels:  crf
Fancy Nlp
NLP for human. A fast and easy-to-use natural language processing (NLP) toolkit, satisfying your imagination about NLP.
Stars: ✭ 233 (+1692.31%)
Mutual labels:  crf
mahjong
开源中文分词工具包,中文分词Web API,Lucene中文分词,中英文混合分词
Stars: ✭ 40 (+207.69%)
Mutual labels:  crf
keras-crf-layer
Implementation of CRF layer in Keras.
Stars: ✭ 76 (+484.62%)
Mutual labels:  crf
Hierarchical-Word-Sense-Disambiguation-using-WordNet-Senses
Word Sense Disambiguation using Word Specific models, All word models and Hierarchical models in Tensorflow
Stars: ✭ 33 (+153.85%)
Mutual labels:  crf
BiLSTM-CRF-NER-PyTorch
This repo contains a PyTorch implementation of a BiLSTM-CRF model for named entity recognition task.
Stars: ✭ 109 (+738.46%)
Mutual labels:  crf

crf-seg

crf-seg是CRF模型用于自然语言处理(NLP)的Java工具包,目标是普及自然语言处理在生产环境中的应用。 crf-seg具备性能高效、架构清晰、语料时新、可自定义语料、可自定义模型的特点。

author:xuming(shibing624)

environment:jdk 1.8

演示页面 http://www.borntowin.cn:8080/xmnlp


CRF模型对新词有很好的识别能力,对繁体字的处理及专有名词识别良好,但开销较大。是目前中文分词效果最好的模型,可用于生产环境。

使用

模型文件需要另外下载,并不包含在源码中,网盘下载:http://pan.baidu.com/s/1skQW35j,放置在 data/model/segment 下。

crf-seg调用方便:

System.out.println(Xmnlp.crfSegment("你好,欢迎使用CRF分词工具!"));

训练自定义模型

  1. 使用GenerateBMESDemo(位于test中的org.xm.xmnlp.demo下)生成自己数据的序列标注集,之后用crf++生成crf模型。

  2. 提供熟语料参考文件以及我用GenerateBMESDemo生成的测试文件,可以作为格式参考,网盘下载:http://pan.baidu.com/s/1eStd0jg

  3. 提供人民日报2014版标注的分词数据,网盘下载:链接:http://pan.baidu.com/s/1gfae4Zh 密码:l506 。尊重版权,传播请注明出处。

  4. 提供linux版和windows版的crf++模型生成工具,网盘下载:http://pan.baidu.com/s/1skKkTgL

  5. 请通过命令行参数指定CRF++生成txt格式的模型,比如:

    crf_learn -f 3 -c 4.0 template train.bmes.txt crf-simple.model -t

    然后将生成的 crf-simple.model.txt 的路径替换到配置项CRFSegmentModelPath,首次运行后会得到相应的 crf-simple.model.txt.bin 文件; 下次加载时会直接从bin缓存加载,速度会快很多。

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].