多轮query改写_自然语言处理(NLP)-阿里云帮助中心

1. 参数

1.1 入参

一级参数

参数	类型	是否必传	说明
algorithm	string	是	标识符
model	string	否	模型版本，默认不传。可选择 dashscope-conv-rewrite-1.8b dashscope-conv-rewrite-14b
input	map<string, object>	是	输入参数
parameters	map<string, object>	否	配置参数
debug	boolean	否	调试模式，默认false

二级参数

类型

是否必传

说明

input.query

string

是

当前轮输入内容

input.history

array(map/dict/json)

否

历史对话记录

通常为如下结构：

[

{"role": "user", "content": "user_content"}, {"role": "assistant", "content": "assistant_content"}

]

{
    "algorithm": "conversation_rewriter",
    "model": "dashscope-conv-rewrite-14b",
    "input": {
        "query": "贵吗",
        "history": [
            {
                "role": "user",
                "content": "什么是黑梓木"
            },
            {
                "role": "assistant",
                "content": "黑梓木是一种用材最广的木头，在我国东北地区也称之为臭梧桐，分布比较广泛，产量也大，很多装饰部件都会用黑梓木制作。"
            }
        ]
    }
}

1.2 出参

一级参数

参数	类型	说明
requestId	string	请求id
status	int	请求结果状态
message	string	响应信息
data	map<string, object>	多轮改写结果
debugInfo	map<string, object>	调试信息

二级参数

二级参数	类型	说明
data.conversationRewriter	array(string)	多轮改写后的query list
data.timeCost	float	服务时间消耗，单位ms
data.inputTokens	int	输入token数
data.outputTokens	int	输出token数

{
    "requestId": null,
    "data": {
        "conversationRewriter": [
            "黑梓木的价格贵吗"
        ],
        "outputTokens": 8,
        "timeCost": 424.08,
        "inputTokens": 61
    },
    "debugInfo": null,
    "message": null,
    "status": 0
}

2. SDK调用

2.1 Java SDK

2.1.1 Maven依赖

<dependency>
    <groupId>com.aliyun</groupId>
    <artifactId>alinlp20200629</artifactId>
    <version>2.7.0</version>
</dependency>

<!--出现java.lang.NoSuchMethodError: com.aliyun.credentials.Client.getCredential()Lcom/aliyun/credentials/models/CredentialModel;异常则引入-->
<dependency>
    <groupId>com.aliyun</groupId>
    <artifactId>credentials-java</artifactId>
    <version>0.3.0</version>
</dependency>

2.1.2 调用示例

/**
 * 使用AK&SK初始化账号Client
 * @param accessKeyId
 * @param accessKeySecret
 * @return Client
 * @throws Exception
 */
public static Client createClient(String accessKeyId, String accessKeySecret) throws Exception {
    Config config = new Config()
    // 必填，您的 AccessKey ID
    .setAccessKeyId(accessKeyId)
    // 必填，您的 AccessKey Secret
    .setAccessKeySecret(accessKeySecret);
    // Endpoint 请参考 https://api.aliyun.com/product/alinlp
    config.endpoint = "alinlp.cn-beijing.aliyuncs.com";
    return new Client(config);
}

public static void main(String[] args) throws Exception {
    String accessKeyId = "xxx";
    String accessKeySecret = "xxx";
    Client client = createClient(accessKeyId, accessKeySecret);
    PostISConvRewriterRequest request = new PostISConvRewriterRequest();
    request.setAlgorithm("conversation_rewriter");
    String input = "{\n" +
            "    \"query\": \"贵吗\",\n" +
            "    \"history\": [\n" +
            "        {\n" +
            "            \"role\": \"user\",\n" +
            "            \"content\": \"什么是黑梓木\"\n" +
            "        },\n" +
            "        {\n" +
            "            \"role\": \"assistant\",\n" +
            "            \"content\": \"黑梓木是一种用材最广的木头，在我国东北地区也称之为臭梧桐，分布比较广泛，产量也大，很多装饰部件都会用黑梓木制作。\"\n" +
            "        }\n" +
            "    ]\n" +
            "}";
    Map<String, Object> inputMap = JSONObject.parseObject(input, Map.class);
    request.setInput(inputMap);
    PostISConvRewriterResponse response = client.postISConvRewriter(request);
    System.out.println(JSONObject.toJSONString(response.getBody()));
}

2.2 Python SDK

2.2.1 pip源

pip install alibabacloud-alinlp20200629

2.2.2 调用示例

import json

from alibabacloud_alinlp20200629 import client
from alibabacloud_tea_openapi import models as api_models
from alibabacloud_alinlp20200629 import models


def conversation_rewriter():
    config = api_models.Config(
        access_key_id='xxx',
        access_key_secret='xxx',
        region_id="cn-beijing")
    nlp_client = client.Client(config)
    request = models.PostISConvRewriterRequest()
    input_str = '{\n' + \
                '    \"query\": \"贵吗\",\n' + \
                '    \"history\": [\n' + \
                '        {\n' + \
                '            \"role\": \"user\",\n' + \
                '            \"content\": \"什么是黑梓木\"\n' + \
                '        },\n' + \
                '        {\n' + \
                '            \"role\": \"assistant\",\n' + \
                '            \"content\": \"黑梓木是一种用材最广的木头，在我国东北地区也称之为臭梧桐，分布比较广泛，产量也大，很多装饰部件都会用黑梓木制作。\"\n' + \
                '        }\n' + \
                '    ]\n' + \
                '}'
    request.input = json.loads(input_str)
    request.algorithm = 'conversation_rewriter'
    response = nlp_client.post_isconv_rewriter(request)
    print(json.dumps(response.body.data, ensure_ascii=False))


if __name__ == '__main__':
    conversation_rewriter()