[TextMatch框架] 简介

2020-10-29 10:19:27 浏览数 (2)

TextMatch

TextMatch is a semantic matching model library for QA & text search … It’s easy to train models and to export representation vectors.

[TextMatch框架] : 文本匹配/文本分类/文本embedding/文本聚类/文本检索（bow/ifidf/ngramtf-df/bert/albert/bm25/…/nn/gbdt/xgb/kmeans/dscan/faiss/….）：https://github.com/MachineLP/TextMatch

run examples

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python examples/text_search.py

examples/text_search.py

代码语言：javascript复制
import sys
from textmatch.models.text_embedding.model_factory_sklearn import ModelFactory


if __name__ == '__main__':
    # doc
    doc_dict = {"0":"我去玉龙雪山并且喜欢玉龙雪山玉龙雪山", "1":"我在玉龙雪山并且喜欢玉龙雪山", "2":"我在九寨沟", "3":"你好"}   
    # query
    query = "我在九寨沟,很喜欢"
    
    # 模型工厂，选择需要的模型加到列表中: 'bow', 'tfidf', 'ngram_tfidf', 'bert', 'albert', 'w2v'
    mf = ModelFactory( match_models=['bow', 'tfidf', 'ngram_tfidf'] )
    # 模型处理初始化
    mf.init(words_dict=doc_dict, update=True)

    # query 与 doc的相似度
    search_res = mf.predict(query)
    print ('search_res>>>>>', search_res) 
    # search_res>>>>> {'bow': [('0', 0.2773500981126146), ('1', 0.5303300858899106), ('2', 0.8660254037844388), ('3', 0.0)], 'tfidf': [('0', 0.2201159065358879), ('1', 0.46476266418455736), ('2', 0.8749225357988296), ('3', 0.0)], 'ngram_tfidf': [('0', 0.035719486884261346), ('1', 0.09654705406841395), ('2', 0.9561288696241232), ('3', 0.0)]}
    
    # query的embedding
    query_emb = mf.predict_emb(query)
    print ('query_emb>>>>>', query_emb) 
    '''
    pre_emb>>>>> {'bow': array([1., 0., 0., 1., 1., 0., 1., 0.]), 'tfidf': array([0.61422608, 0.        , 0.        , 0.4842629 , 0.4842629 ,
       0.        , 0.39205255, 0.        ]), 'ngram_tfidf': array([0.        , 0.        , 0.37156534, 0.37156534, 0.        ,
       0.        , 0.        , 0.29294639, 0.        , 0.37156534,
       0.37156534, 0.        , 0.        , 0.37156534, 0.        ,
       0.        , 0.        , 0.        , 0.        , 0.        ,
       0.        , 0.        , 0.29294639, 0.37156534, 0.        ,
       0.        , 0.        , 0.        , 0.        , 0.        ,
       0.        , 0.        , 0.        , 0.        ])}
    '''

run examples

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python tests/tools_test/faiss_test.py

tests/tools_test/faiss_test.py

代码语言：javascript复制
import sys
import json 
import time
import faiss
import numpy as np
from faiss import normalize_L2
from textmatch.config.constant import Constant as const
from textmatch.core.text_embedding import TextEmbedding
from textmatch.tools.decomposition.pca import PCADecomposition
from textmatch.tools.faiss.faiss import FaissSearch

test_dict = {"id0": "其实事物发展有自己的潮流和规律",
   "id1": "当你身处潮流之中的时候，要紧紧抓住潮流的机会",
   "id2": "想办法脱颖而出，即使没有成功，也会更加洞悉时代的脉搏",
   "id3": "收获珍贵的知识和经验。而如果潮流已经退去",
   "id4": "这个时候再去往这个方向上努力，只会收获迷茫与压抑",
   "id5": "对时代、对自己都没有什么帮助",
   "id6": "但是时代的浪潮犹如海滩上的浪花，总是一浪接着一浪，只要你站在海边，身处这个行业之中，下一个浪潮很快又会到来。你需要敏感而又深刻地去观察，略去那些浮躁的泡沫，抓住真正潮流的机会，奋力一搏，不管成败，都不会遗憾。",
   "id7": "其实事物发展有自己的潮流和规律",
   "id8": "当你身处潮流之中的时候，要紧紧抓住潮流的机会" }


if __name__ == '__main__':
    # ['bow', 'tfidf', 'ngram_tfidf', 'bert']
    # ['bow', 'tfidf', 'ngram_tfidf', 'bert', 'w2v']
    # text_embedding = TextEmbedding( match_models=['bow', 'tfidf', 'ngram_tfidf', 'w2v'], words_dict=test_dict ) 
    text_embedding = TextEmbedding( match_models=['bow', 'tfidf', 'ngram_tfidf', 'w2v'], words_dict=None, update=False ) 
    feature_list = []
    for sentence in test_dict.values():
        pre = text_embedding.predict(sentence)
        feature = np.concatenate([pre[model] for model in ['bow', 'tfidf', 'ngram_tfidf', 'w2v']], axis=0)
        feature_list.append(feature)
    pca = PCADecomposition(n_components=8)
    data = np.array( feature_list )
    pca.fit( data )
    res = pca.transform( data )
    print('res>>', res)

   

    pre = text_embedding.predict("潮流和规律")
    feature = np.concatenate([pre[model] for model in ['bow', 'tfidf', 'ngram_tfidf', 'w2v']], axis=0)
    test = pca.transform( [feature] )

    faiss_search = FaissSearch( res, sport_mode=False )
    faiss_res = faiss_search.predict( test )
    print( "faiss_res:", faiss_res )
    '''
    faiss kmeans result times 8.0108642578125e-05
    faiss_res: [{0: 0.7833399, 7: 0.7833399, 3: 0.63782495}]
    '''

    
    faiss_search = FaissSearch( res, sport_mode=True )
    faiss_res = faiss_search.predict( test )
    print( "faiss_res:", faiss_res )
    '''
    faiss kmeans result times 3.266334533691406e-05
    faiss_res: [{0: 0.7833399, 7: 0.7833399, 3: 0.63782495}]
    '''

run examples：基于召回和排序的文本搜索

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python tests/core_test/text_search_test.py

tests/core_test/text_search_test.py

代码语言：javascript复制
import sys
from textmatch.core.text_match import TextMatch
from textmatch.core.qa_match import QMatch, AMatch, SemanticMatch
from textmatch.models.text_search.model_factory_search import ModelFactorySearch



def text_match_recall(testword, doc_dict):
    # QMatch
    q_match = QMatch( q_dict=doc_dict, match_models=['bow', 'tfidf', 'ngram_tfidf', 'albert']) 
    q_match_pre = q_match.predict(testword, match_strategy='score', vote_threshold=0.5, key_weight = {'bow': 1, 'tfidf': 1, 'ngram_tfidf': 1, 'albert': 1})
    # print ('q_match_pre>>>>>', q_match_pre )
    return q_match_pre 

def text_match_sort(testword, candidate_doc_dict):
    text_match = TextMatch( q_dict=candidate_doc_dict, match_models=['bm25', 'edit_sim', 'jaccard_sim'] )
    text_match_res = text_match.predict( query, match_strategy='score', vote_threshold=-100.0, key_weight = {'bm25': 0, 'edit_sim': 1, 'jaccard_sim': 1} )
    return text_match_res 


if __name__ == '__main__':
    doc_dict = {"0":"我去玉龙雪山并且喜欢玉龙雪山玉龙雪山", "1":"我在玉龙雪山并且喜欢玉龙雪山", "2":"我在九寨沟", "3":"我在九寨沟,很喜欢", "4":"很喜欢"}   
    query = "我在九寨沟,很喜欢"
    
    # 直接搜索
    mf = ModelFactorySearch( match_models=['bm25', 'edit_sim', 'jaccard_sim'] )
    mf.init(words_dict=doc_dict)
    pre = mf.predict(query)
    print ('pre>>>>>', pre) 

    
    # 先召回
    match_pre = text_match_recall( query, doc_dict )
    print( '召回的结果:', match_pre )

    candidate_doc_dict = dict( zip( match_pre.keys(), [doc_dict[key] for key in match_pre.keys()] ) )
    print ("candidate_doc_dict:", candidate_doc_dict)

    # 再排序
    # ['bm25', 'edit_sim', 'jaccard_sim']
    text_match_res = text_match_sort( query, candidate_doc_dict )
    print ('排序的score>>>>>', text_match_res) 



    '''
    # 排序
    mf = ModelFactorySearch( match_models=['bm25', 'edit_sim', 'jaccard_sim'] )
    mf.init(words_dict=candidate_doc_dict) 
    pre = mf.predict(query)
    print ('排序的结果>>>>>', pre) 
    '''

run tests/core_test （qa/文本embedding）

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
pip install -r requirements.txt
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python tests/core_test/qa_match_test.py
python tests/core_test/text_embedding_test.py

run tests/models_test （模型测试）

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
pip install -r requirements.txt
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python tests/models_test/bm25_test.py                    (bm25)
python tests/models_test/edit_sim_test.py                (编辑距离)
python tests/models_test/jaccard_sim_test.py             (jaccard)
python tests/models_test/bow_sklearn_test.py             (bow)
python tests/models_test/tf_idf_sklearn_test.py          (tf_idf)
python tests/models_test/ngram_tf_idf_sklearn_test.py    (ngram_tf_idf)
python tests/models_test/w2v_test.py                     (w2v)
python tests/models_test/albert_test.py                  (bert)

run tests/ml_test （机器学习模型测试）

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
pip install -r requirements.txt
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python tests/ml_test/lr_test.py                          (lr)
python tests/ml_test/gbdt_test.py                        (gbdt)
python tests/ml_test/gbdt_lr_test.py                     (gbdt_lr)
python tests/ml_test/lgb_test.py                         (lgb)
python tests/ml_test/xgb_test.py                         (xgb)

run tests/tools_test （聚类/降维工具测试）

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
pip install -r requirements.txt
export PYTHONPATH=${PYTHONPATH}:../TextMatch
python tests/tools_test/kmeans_test.py                   (kmeans)
python tests/tools_test/dbscan_test.py                   (dbscan)
python tests/tools_test/pca_test.py                      (pca)
python tests/tools_test/faiss_test.py                    (faiss)

run tests/tools_test （词云）

代码语言：javascript复制
git clone https://github.com/MachineLP/TextMatch
cd TextMatch
pip install -r requirements.txt
cd tests/tools_test
python generate_word_cloud.py
python 模型测试 腾讯云测试服务

0 人点赞