故事复述

更新时间:

该文档是开发者在产品功能集成阶段 使用的英文故事复述题型的语音评测技术文档

本题型适用场景:

  • 先给答题者播放一段音频,然后让答题者将听到的内容复述出来。

评测请求参数

字段

类型

可选

说明

coreType

String

必选

en.retell.score表示英文故事复述题型

rank

int

可选

总分评分分制

attachAudioUrl

int

可选

评测结果中是否返回音频url
0: 不返回
1: 返回

precision

double

可选

设置打分精度,只支持0.1、0.5或者1,默认值为1(如果设置的值不是0.10.5则按1处理)

lm

array

必选

数组字典,复述内容的范例 (注:lm字段是LM的小写形式)

grade

int

可选

指定学段;只支持12,默认值为1 1:表示初中学段(覆盖要点到90%左右才能得满分) 2:表示高中学段(因为高中话题有难度,只要覆盖80%左右就能得高分)

phones

object

可选

指定单词的发音。如果单词有多发音,要用”|”符号分隔开;每个音素之间及音素与音节分隔符”-“之间,必须使用空格隔开。 例:"phones":{"directory":"d ih - 'r eh k - tr ih|d ax - 'r eh k - t ax - r ih"};

points

array

可选

设置答案要点; 每个要点一个text

rateScale

float

可选

打分宽松度,设置的范围0.8~1.5,默认1.0;如果设置值小于0.8,按0.8计算;如果大于1.5则按1.5计算。这个参数可以看作是个乘数,值越高打分越高。

参数示例

{
    "coreType": "en.retell.score",
    "rank": 100,
    "precision": 0.5,
    "attachAudioUrl":1,
    "lm": [
        {
            "text": "My aunt Judy was a single mother of fifty, working as a secretary. When doctors discovered she had a very serious heart disease, she didn't give in. She began studying medical reports and found an article in a magazine about a well known heart doctor, Dr. Michael Baker, of Houston, Texas. The article said Dr. Baker's fees were very high; Aunt Judy couldn't afford them. But could he tell her of someone whose fee she could pay? So Aunt Judy wrote to him. She simply listed her reasons for wanting to live: her three children, who would be independent in three or four more years, her little girl dream of travelling and seeing the world. There wasn't a word of self pity only warmth and humor and the joy of living. She mailed the letter, not really expecting an answer. A few days later, Aunt Judy rushed into my house, and read aloud a letter: your beautiful letter moved me very deeply. If you can come to Houston, there will be no charge for either the hospital or the operation. Michael Baker. That was seven years ago. Since then, Aunt Judy has been around the world. Her three children are happily married."
        }
    ],
    "points": [
        {
            "text": "My aunt Judy was a single mother of fifty, working as a secretary."
        },
        {
            "text": "When doctors discovered she had a very serious heart disease, she didn't give in."
        },
        {
            "text": "She began studying medical reports and found an article in a magazine about a well known heart doctor, Dr. Michael Baker, of Houston, Texas."
        },
        {
            "text": "The article said Dr. Baker's fees were very high; Aunt Judy couldn't afford them."
        },
        {
            "text": "But could he tell her of someone whose fee she could pay? So Aunt Judy wrote to him."
        },
        {
            "text": "She simply listed her reasons for wanting to live: her three children, who would be independent in three or four more years, her little girl dream of travelling and seeing the world."
        },
        {
            "text": "There wasn't a word of self pity only warmth and humor and the joy of living."
        },
        {
            "text": "She mailed the letter, not really expecting an answer."
        },
        {
            "text": "A few days later, Aunt Judy rushed into my house, and read aloud a letter: your beautiful letter moved me very deeply. If you can come to Houston, there will be no charge for either the hospital or the operation. Michael Baker."
        },
        {
            "text": "That was seven years ago. Since then, Aunt Judy has been around the world. Her three children are happily married."
        }
    ]
}

返回结果说明

类型

说明

使用建议

recordId

string

音频文件的唯一标识

可供参考

tokenId

string

用户请求的唯一标识

可供参考

applicationId

string

提供的授权账号,即Appkey

内部参数

dtLastResponse

string

云端响应的时间点

内部参数

eof

int

0表示返回未结束,后续还有其它的返回结果 1:表示本次评测所有的返回结束

内部参数

params

object

回显客户端传入的评测请求参数

- app

object

- - applicationId

string

提供的授权账号,即Appkey

- - userId

string

用户标识 ,自定义字符串,方便定位排查问题

- - timestamp

string

start时间

- - sig

string

校验码

- - clientId

string

- request

object

评分请求参数节

- audio

object

评分请求的音频参数节点数

result

object

评测结果

- res

string

评测本题型时,使用的资源名称

内部参数

- version

string

引擎的版本,包含发布时间

内部参数

- rank

int

评分分制

内部参数

- precision

float

评分精度

内部参数

- pretime

int

云端调用Start接口本身耗时,单位ms

内部参数

- systime

int

评测整个过程的总耗时,单位ms

内部参数

- delaytime

int

云端从feed音频结束到获取结果的耗时,单位ms

内部参数

- wavetime

int

音频时长,单位ms

可供参考

- overall

float

总分

建议字段

- fluency

float

流利度分(表述流畅情况,与正确答案无关)

建议字段

- pron

float

发音分(体现发音情况,与正确答案无关,只与单词发音准确程度有关)

建议字段

- rhythm

float

韵律分(暂未使用)

内部参数

- integrity

float

完整度分

内部参数

- keypoints_hit

array

points顺序分别对应每个要点的命中情况,无points字段时,该字段不输出;

建议字段

- fake_words

object

单词不在词典(集外词),才会输出此字段。如果需要对集外词进行评测,可以通过refText传音标的方式进行评测。

- -char

string

- info

object

- - snr

float

信噪比,值越高越清晰,范围(0~40dB) 此参数影响评分时,会设置相应的tipId

可供参考

- - clip

float

音频声音太高,出现截幅,范围(0\~1)。 此参数影响评分时,会设置相应的tipId

内部参数

- - volume

int

录音音量,范围(0~180dB)

可供参考

- - tipId

int

音频质量,详情请参考 tipId提示信息说明

建议字段

  • 注:返回结果中其它未在上面展示的字段是内部参数,产品层面不需要关注

结果示例

维度

说明

详细

overall

总分

题型的评测得分

pron

发音分

fluency

流利度

keypoints_hit

各个要点的命令率

参考对应的题型说明

评测失败时的返回结果

如果返回的评测结果中result结构为空,说明评测失败。通过返回结果可以获取 errId 和 error,需要根据相应错误码进行后续判断和处理。

错误码

错误码参考:评测错误码