200行java代码实现ChatDOC

需求：

PDF,DOCX和TXT格式的文档，通过我们的模型短时间训练后，您就可以和自己的文档对话

package gpt3;

import com.colorbin.common.base.JsonTable;
import com.colorbin.common.constant.ProjectConstant;
import com.colorbin.entity.gpt.GptTernary;
import com.colorbin.entity.gpt.model.GptConvertUtils;
import com.colorbin.entity.gpt.model.GptTernaryVo;
import com.colorbin.rpa.c_magic_ai.c02_nlp.nlpUtil;
import com.colorbin.rpa.d_software_auto.d10_folder.fileUtil;
import com.colorbin.service.gpt.GptTernaryService;
import com.colorbin.service.gpt.impl.GptTernaryServiceImpl;
import com.colorbin.util._listUtils;
import com.fasterxml.jackson.core.JsonFactory;
import com.fasterxml.jackson.core.JsonParser;
import com.fasterxml.jackson.core.JsonToken;
import com.hankcs.hanlp.HanLP;
import com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLSentence;
import com.hankcs.hanlp.corpus.dependency.CoNll.CoNLLWord;
import org.apache.poi.hwpf.HWPFDocument;
import org.apache.poi.hwpf.extractor.WordExtractor;
import org.apache.poi.hwpf.usermodel.Range;

import java.io.File;
import java.io.FileInputStream;
import java.io.IOException;
import java.lang.reflect.Field;
import java.util.*;
import java.util.concurrent.CompletableFuture;
import java.util.concurrent.LinkedBlockingQueue;
import java.util.concurrent.ThreadPoolExecutor;
import java.util.concurrent.TimeUnit;

public class TestGpt2 {
    public static void main(String[] args) throws Exception{
        //1、清空t_gpt_ternaryh
        Class<?> cls = GptTernary.class;
        Field cid = cls.getDeclaredField("id");
        JsonTable annotation = cid.getAnnotation(JsonTable.class);
        String name= annotation.value();
        String fileStr1="C:\\data\\json\\"+name+ ProjectConstant.DATA_JSON;
        fileUtil.writeTxtFile(fileStr1,"[]");
        //2、获取文档内容
        String a = getDoc();
        String as[] = a.split("。");
        //3、获取三元关系
        for (String s: as) {
            nlpDoc(s);
        }
        //4、智能问答
        String q = "消防几点演习？";
        String answer = send2(q);
        System.out.println(answer);

    }
    private static ThreadPoolExecutor threadPool =
            new ThreadPoolExecutor(100, Integer.MAX_VALUE, 60L,
                    TimeUnit.MILLISECONDS,
                    new LinkedBlockingQueue<>(),
                    new ThreadPoolExecutor.CallerRunsPolicy());
    private static String send2(String name) throws Exception {
        final List<String>[] list7 = new List[]{new ArrayList<>()};
        CompletableFuture one = CompletableFuture.runAsync(() -> {
            try {
                String question = name;
                List<String> list3 = com.colorbin.rpa.c_magic_ai.c02_nlp.nlpUtil.getJG(new String[]{question});
                Set<String> stringSet = new HashSet<>();
                stringSet.addAll(list3);
                list7[0] = new ArrayList<>(stringSet);
            } catch (Exception e) {
                throw new RuntimeException(e);
            }
        }, threadPool);
        CompletableFuture.allOf(one).get();
        //去除停用词
        List<String>[] list6 = new List[]{new ArrayList<>()};
        CompletableFuture two = CompletableFuture.runAsync(() -> {
            try {
                list6[0] = com.colorbin.rpa.c_magic_ai.c02_nlp.nlpUtil.removeStopWords(list7[0]);
            } catch (Exception e) {
                throw new RuntimeException(e);
            }
        }, threadPool);
        CompletableFuture.allOf(two).get();
        String fileName = ProjectConstant.DATA_JSON_PATH + "t_gpt_ternary" +ProjectConstant.DATA_JSON;
        List<GptTernary>[] list9 = new List[]{new ArrayList<>()};
        //读取大文件 时间较长
        List<GptTernary> list8 = new ArrayList<>();
        JsonFactory factory = new JsonFactory();
        JsonParser parser = factory.createParser(new File(fileName));
        GptTernary m2 = new GptTernary();
        while (!parser.isClosed()) {
            JsonToken token = parser.nextToken();
            if (token == null) {
                break;
            }
            if (JsonToken.FIELD_NAME.equals(token)) {
                String fieldName = parser.getCurrentName();
                parser.nextToken();
                String fieldValue = parser.getValueAsString();
                if (fieldName.equals("subject")){
                    m2.setSubject(fieldValue);
                }
                if (fieldName.equals("object")){
                    m2.setObject(fieldValue);
                }
                if (fieldName.equals("predicate")){
                    m2.setPredicate(fieldValue);
                }
                if(m2.getSubject()!=null && m2.getObject()!=null && m2.getPredicate()!=null){
                    if(containsKeyFlag(m2.getSubject(),(list6[0])) || containsKeyFlag(m2.getObject(),(list6[0]))
                            || containsKeyFlag(m2.getPredicate(),(list6[0]))){
                        list8.add(m2);
                    }
                    m2 = new GptTernary();
                }
            }
        }
        parser.close();
        Set<GptTernary> metaFaqSet1 = new HashSet<>();
        metaFaqSet1.addAll(list8);
        list9[0] = new ArrayList<>(metaFaqSet1);
        List<GptTernaryVo> list10 = new ArrayList<>();
        CompletableFuture four = CompletableFuture.runAsync(() -> {
            List<GptTernaryVo> lists = GptConvertUtils.INSTANCE.doToDTO(list9[0]);
            for (GptTernary dataInterview : lists) {
                double similarity = nlpUtil.getSimilarity(name, dataInterview.getSubject()+dataInterview.getPredicate()+dataInterview.getObject());
                GptTernaryVo metaFaqVo = new GptTernaryVo();
                metaFaqVo.setId(dataInterview.getId());
                metaFaqVo.setSubject(dataInterview.getSubject());
                metaFaqVo.setObject(dataInterview.getObject());
                metaFaqVo.setPredicate(dataInterview.getPredicate());
                metaFaqVo.setSim(similarity);
                list10.add(metaFaqVo);
            }
            _listUtils.sort(list10, false, "sim");
        }, threadPool);
        CompletableFuture.allOf(four).get();
        if (list10.size() > 0) {
            return list10.get(0).getSubject()+list10.get(0).getPredicate()+list10.get(0).getObject();
        }else {
            return "没有找到相似答案";
        }

    }

    public static void saveRelation(GptTernaryService gptTernaryService, String a, String b, String c) throws Exception{
        GptTernary gptTernary = new GptTernary();
        gptTernary.setSubject(a);
        gptTernary.setPredicate(b);
        gptTernary.setObject(c);
        gptTernaryService.saveOrUpdate(gptTernary);
    }
    private static boolean containsKeyFlag(String query, List<String> strings) {
        for (String str:strings){
            if(query.contains(str)){
                return true;
            }
        }
        return false;
    }
    public static void nlpDoc(String text) throws Exception{
        CoNLLSentence sentence = HanLP.parseDependency(text);
        Map<String, List<CoNLLWord>> maps = new HashMap<>();
        for (CoNLLWord word : sentence) {
            String b = word.DEPREL;
            maps.put(b, new ArrayList<>());
        }
        for (Map.Entry<String, List<CoNLLWord>> entry : maps.entrySet()) {
            List<CoNLLWord> list = new ArrayList<>();
            for (CoNLLWord word : sentence) {
                if (entry.getKey().equals(word.DEPREL)) {
                    list.add(word);
                    maps.put(entry.getKey(), list);
                }
            }
        }
        GptTernaryService gptTernaryService = new GptTernaryServiceImpl();
        if(maps.containsKey("主谓关系") && maps.containsKey("动宾关系")){
            CoNLLWord entity1 = maps.get("主谓关系").get(0);
            CoNLLWord entity2 = maps.get("动宾关系").get(0);
            String relation = entity1.HEAD.LEMMA;
            saveRelation(gptTernaryService, entity1.LEMMA.trim(), relation.trim().replaceAll("\r\n","").replaceAll(" ",""),
                    entity2.LEMMA);
        }
        Map<String, List<CoNLLWord>> dic = maps;
        if (dic.containsKey("主谓关系") && dic.containsKey("动补结构")){
            CoNLLWord entity1 = dic.get("主谓关系").get(0);
            CoNLLWord prep = dic.get("动补结构").get(0);
            if (maps.containsKey("介宾关系")){
                CoNLLWord entity2 = maps.get("介宾关系").get(0);
                String relation = entity1.HEAD.LEMMA + prep.LEMMA;
                saveRelation(gptTernaryService, entity1.LEMMA.trim(), relation.trim().replaceAll("\r\n","").replaceAll(" ",""),
                        entity2.LEMMA);
            }
        }
        if (dic.containsKey("前置宾语")){
            CoNLLWord entity2 = dic.get("前置宾语").get(0);
            if (dic.containsKey("状中结构")){
                CoNLLWord adverbial = dic.get("状中结构").get(0);
                if (maps.containsKey("介宾关系")){
                    CoNLLWord entity1 = maps.get("介宾关系").get(0);
                    String relation = adverbial.HEAD.LEMMA;
                    saveRelation(gptTernaryService, entity1.LEMMA.trim(), relation.trim().replaceAll("\r\n","").replaceAll(" ",""),
                            entity2.LEMMA);
                }
            }
        }
    }


    public static String getDoc() throws Exception{
//        readAndWriterTest3();
        File file = null;
        WordExtractor extractor = null;
        String a = "";
        try {
            file = new File("D:\\Personal\\Desktop\\消防演习通知.doc");
            FileInputStream fis = new FileInputStream(file.getAbsolutePath());
            HWPFDocument document = new HWPFDocument(fis);
            extractor = new WordExtractor(document);
            String[] fileData = extractor.getParagraphText();
            for(int i=0;i<fileData.length;i++){
                if(fileData[i] != null)
//                    System.out.println(fileData[i]);
                a+=fileData[i];
            }
        } catch(Exception exep) {
            exep.printStackTrace();
        }
        return a;
    }

    public static void readAndWriterTest3() throws IOException {
        File file = new File("D:\\Personal\\Desktop\\光大店2023.5.26消防演习通知.doc");
        String str = "";
        try {
            FileInputStream fis = new FileInputStream(file);
            HWPFDocument doc = new HWPFDocument(fis);
            String doc1 = doc.getDocumentText();
            System.out.println(doc1);
            StringBuilder doc2 = doc.getText();
            System.out.println(doc2);
            Range rang = doc.getRange();
            String doc3 = rang.text();
            System.out.println(doc3);
            fis.close();
        } catch (Exception e) {
            e.printStackTrace();
        }
    }

}

写的代码都是问GPT的。

1. ChatGPT API节截至2023年6月1日前全面免费试用；
2. 免费提交Prompt机器人，并进行多轮对话。不仅仅只有ChatGPT应用开发者才是开发者，只要您能编写出有趣、实用的Prompt，也将成为新时代的开发者；
3. ChatDOC智能对话：只需提交PDF, DOC和TXT文档，您就能与您的文档进行智能对话，方便您更快速地阅读文档。

可以上传你的PDF,DOCX和TXT格式的文档，通过我们的模型短时间训练后，您就可以和自己的文档对话，现在支持单文档，后续支持多文档。