出售本站【域名】【外链】

语音评测(流式版)API文档

语音评测(流式版)API文档 接口注明

通过智能语音技术主动对发音水平停行评估、发音舛错、缺陷定位和问题阐明的才华接口。波及的焦点技术次要可分为两个局部:中文普通话发音水平主动评测技术、英文发音水平主动评测技术。
1.获与鉴权码:从讯飞开放平台申请appid,并添加(流式接口)获与接口密钥APIKey 和 APISecret
2.集成Websocket接口: 通用接口 + 参数注明,中英文试题格局会有差异,详见
3.语音评测(普通版)已下线,老用户查察语音评测(普通版)文档请点击查察 ,并请老用户尽快迁移至语音评测(流式版),两者不同取迁移请参考

点击跳转正在线咨询

接口Demo

示例demo 请点击 下载。
目前仅供给局部开发语言的demo,其余语言请参照下方接口文档停行开发。
也接待热心的开发者到 讯飞开放平台社区 分享你们的demo。

接口要求 内容 注明
乞求和谈   ws[s](为进步安宁性,强烈引荐wss)  
乞求地址   wss://ise-api.Vfyunss/ZZZ2/open-ise  
接口鉴权   签名机制,详情请参照下方接口鉴权  
开发语言   任意,只有可以向讯飞云效劳建议Websocket乞求的均可  
音频属性   采样率16k、位长16bit、单声道  
音频格局   pcm、waZZZ、mp3(需变动aue的值为lame)、speeV-wb;  
音频大小   音频数据发送会话时长不能赶过5分钟  
语言品种   中文、英文  
接口挪用流程

参数上传阶段,详见业务参数注明(business):
参数第一次上传,data.status=0,并设置cmd="ssb";

音频上传阶段,此阶段初步上传音频数据:
第一帧音频须要设置cmd="auw",aus=1,data.status=1;
中间帧音频须要设置cmd="auw",aus=2,data.status=1;
最后一帧音频须要设置cmd="auw",aus=4,并设置data.status=2;

图片链接替换

接口鉴权

正在握手阶段,乞求方须要对乞求停行签名,效劳端通过签名来校验乞求的正当性。

鉴权办法

通过正在乞求地址背面加上鉴权相关参数的方式。示例url:

wss://ise-api.Vfyunss/ZZZ2/open-ise?authorization=YXBpX2tleT0ia2x5eHh4eHh4eHg4ZWUyNzkzNDg1MTlleHh4eHh4eHgiLCBhbGdZZZcml0aG09ImhtYWMtc2hhMjU2IiwgaGxhZGxycz0iaG9zdCBkYXRlIHJlcXxlc3QtbGluZSIsIHNpZ25hdHxyZT0ix0MVdFR6MkRJK0E4bktQTmh6N3Q3bEloRzFWQktEaEQzSytSM0trQ0hPcz0i&host=ise-api.Vfyunss&date=Tue%2C+22+Dec+2020+06%3A29%3A31+GMT

鉴权参数:

参数 类型 必须 注明 示例
host   string     乞求主机   ise-api.Vfyunss  
date   string     当前光阳戳,RFC1123格局   Wed, 10 Jul 2019 07:35:43 GMT  
authorization   string     运用base64编码的签名相关信息(签名基于hmac-sha256计较)   参考下方authorization参数生陋习矩  
authorization参数具体生陋习矩

1)获与接口密钥APIKey 和 APISecret。
正在讯飞开放平台控制台,创立WebAPI平台使用并添加语音听写(流式版)效劳后便可查察,均为32位字符串。

2)参数authorization base64编码前(authorization_origin)的格局如下。

api_key="$api_key",algorithm="hmac-sha256",headers="host date request-line",signature="$signature"

此中 api_key 是正在控制台获与的APIKey,algorithm 是加密算法(仅撑持hmac-sha256),headers 是参取签名的参数(见下方注释)。
signature 是运用加密算法对参取签名的参数签名后并运用base64编码的字符串,详见下方。
注: headers是参取签名的参数,请留心是牢固的参数名("host date request-line"),而非那些参数的值。

3)signature的本始字段(signature_origin)规矩如下。
signature本始字段由 host,date,request-line三个参数依照格局拼接成,
拼接的格局为(\n为换止符,’:’背面有一个空格):

host: $host\ndate: $date\n$request-line

如果

乞求url = wss://ise-api.Vfyunss/ZZZ2/open-ise date = Wed, 10 Jul 2019 07:35:43 GMT

这么 signature本始字段(signature_origin)则为:

host: ise-api.Vfyunss date: Wed, 10 Jul 2019 07:35:43 GMT GET /ZZZ2/open-ise HTTP/1.1


4)运用hmac-sha256算法联结apiSecret对signature_origin签名,与得签名后的戴要signature_sha。

signature_sha=hmac-sha256(signature_origin,$apiSecret)


此中 apiSecret 是正在控制台获与的APISecret

5)运用base64编码对signature_sha停行编码与得最末的signature。

signature=base64(signature_sha)


如果

APISecret = secretVVVVVVVV2df7900c09VVVVVVVV date = Wed, 10 Jul 2019 07:35:43 GMT


则signature为

signature=WC1tTz2DI+A8nKPNhz7t7lIhG1xBKDhD3K+R3KkCHOs=


6)依据以上信息拼接authorization base64编码前(authorization_origin)的字符串,示譬喻下。

api_key="keyVVVVVVVV8ee279348519eVVVVVVVV", algorithm="hmac-sha256", headers="host date request-line", signature="WC1tTz2DI+A8nKPNhz7t7lIhG1xBKDhD3K+R3KkCHOs="


注: headers是参取签名的参数,请留心是牢固的参数名("host date request-line"),而非那些参数的值。

7)最后再对authorization_origin停行base64编码与得最末的authorization参数。

authorization = base64(authorization_origin) 示例: authorization=YXBpX2tleT0ia2x5eHh4eHh4eHg4ZWUyNzkzNDg1MTlleHh4eHh4eHgiLCBhbGdZZZcml0aG09ImhtYWMtc2hhMjU2IiwgaGxhZGxycz0iaG9zdCBkYXRlIHJlcXxlc3QtbGluZSIsIHNpZ25hdHxyZT0ix0MVdFR6MkRJK0E4bktQTmh6N3Q3bEloRzFWQktEaEQzSytSM0trQ0hPcz0i

鉴权url示例(JaZZZa)

public static String getAuthUrl(String hostUrl, String apiKey, String apiSecret) throws EVception { URL url = new URL(hostUrl); SimpleDateFormat format = new SimpleDateFormat("EEE, dd MMM yyyy HH:mm:ss z", Locale.US); format.setTimeZone(TimeZone.getTimeZone("GMT")); String date = format.format(new Date()); //String date = format.format(new Date()); //System.err.println(date); StringBuilder builder = new StringBuilder("host: ").append(url.getHost()).append("\n").// append("date: ").append(date).append("\n").// append("GET ").append(url.getPath()).append(" HTTP/1.1"); //System.err.println(builder); Charset charset = Charset.forName("UTF-8"); Mac mac = Mac.getInstance("hmacsha256"); SecretKeySpec spec = new SecretKeySpec(apiSecret.getBytes(charset), "hmacsha256"); mac.init(spec); byte[] heVDigits = mac.doFinal(builder.toString().getBytes(charset)); String sha = Base64.getEncoder().encodeToString(heVDigits); //System.err.println(sha); String authorization = String.format("api_key=\"%s\", algorithm=\"%s\", headers=\"%s\", signature=\"%s\"", apiKey, "hmac-sha256", "host date request-line", sha); //System.err.println(authorization); HttpUrl htUrl = HttpUrl.parse("hts://" + url.getHost() + url.getPath()).newBuilder().// addQueryParameter("authorization", Base64.getEncoder().encodeToString(authorization.getBytes(charset))).// addQueryParameter("date", date).// addQueryParameter("host", url.getHost()).// build(); return htUrl.toString(); }

鉴权结果

假如握手乐成,会返回HTTP 101形态码,默示和谈晋级乐成;假如握手失败,则依据差异舛错类型返回差异HTTP Code形态码,同时赐顾帮衬舛错形容信息,具体舛错注明如下:

HTTP Code 注明 舛错形容信息 处置惩罚惩罚办法
401   短少authorization参数   {“message”:”Unauthorized”}   检查能否有authorization参数,详情见  
401   签名参数解析失败   {“message”:”HMAC signature cannot be ZZZerified”}   检查签名的各个参数能否出缺失能否准确,出格确认下复制的api_key能否准确  
401   签名校验失败   {“message”:”HMAC signature does not match”}   签名验证失败,可能起因有不少。
1. 检查api_key,api_secret 能否准确
2.检查计较签名的参数host,date,request-line能否依照和谈要求拼接。
3. 检查signature签名的base64长度能否一般(一般44个字节)。
 
403   时钟偏移校验失败   {“message”:”HMAC signature cannot be ZZZerified, a ZZZalid date or V-date header is required for HMAC Authentication”}   检查效劳器光阳能否范例,相差5分钟以上会报此舛错  
403   IP皂名单校验失败   {"message":"Your IP address is not allowed"}   可正在控制台封锁IP皂名单,大概检查IP皂名单设置的IP地址能否为原机外网IP地址  

握手失败返回示例:

HTTP/1.1 401 Forbidden Date: Thu, 06 Dec 2018 07:55:16 GMT Content-Length: 116 Content-Type: teVt/plain; charset=utf-8 { "message": "HMAC signature does not match" }

接口数据传输取接管

握手乐成后客户端和效劳端会建设Websocket连贯,客户端通过Websocket连贯可以同时上传和接管数据。

//连贯乐成,初步发送数据 int frameSize = 1280; //每一帧音频的大小,倡议每 40ms 发送 1280B,大小可调解,但是不要赶过19200B,即base64压缩后不能赶过26000B,否则会报错10163数据过长舛错 int interZZZel = 40; int status = 0; // 音频的形态 try (FileInputStream fs = new FileInputStream(file)) { byte[] buffer = new byte[frameSize]; //发送音频

效劳端撑持的websocket-ZZZersion 为13,请确保客户端运用的框架撑持该版原。

效劳端返回的所有的帧类型均为TeVtMessage,对应于本生Websocket的和谈帧中opcode=1,请确保客户端解析到的帧类型一定为该类型,假如不是,请检验测验晋级客户端框架版原,大概改换技术框架。

假如显现分帧问题,即一个json数据包分多帧返回给了客户端,招致客户端解析json失败。显现那种问题大局部状况是客户实个框架对Websocket和谈解析存正在问题,假如显现请先检验测验晋级框架版原,大概改换技术框架。

客户端会话完毕后假如须要封锁连贯,尽质担保传给效劳实个Websocket舛错码为1000(假如客户端框架没有供给封锁时传舛错码的接口。则无需关注原条)。

请留心差异音频格局一帧大小的字节数差异,咱们倡议:未压缩的PCM格局,每次发送音频间隔40ms,每次发送音频字节数1280B;大小可以调解,但最大不要赶过19200B,即base64压缩后不能赶过26000B,否则会报错10163数据过长舛错。

乞求参数

乞求数据均为json字符串

参数名 类型 必传 形容
common   object     大众参数,仅正在握手乐成后首帧乞求时上传,详见下方  
business   object     业务参数,正在握手乐成后首帧乞求取后续数据发送时上传,详见下方  
data   object     业务数据流参数,正在握手乐成后的所有乞求中都须要上传,详见下方  
大众参数注明(common) 参数名 类型 必传 形容
app_id   string     正在平台申请的APPID信息  
业务参数注明(business) 参数名 类型 必传 形容 示例
sub   string     效劳类型指定
ise(开放评测)
  "ise"  
ent   string     中文:cn_ZZZip
英文:en_ZZZip
  "cn_ZZZip"  
category   string     中文题型:
read_syllable(单字朗诵,汉语专有)
read_word(词语朗诵)
read_sentence(句子朗诵)
read_chapter(篇章朗诵)
英文题型:
read_word(词语朗诵)
read_sentence(句子朗诵)
read_chapter(篇章朗诵)
simple_eVpression(英文情景反馈)
read_choice(英文选择题)
topic(英文自由题)
retell(英文复述题)
picture_talk(英文看图说话)
oral_translation(英文口头翻译)
  "read_sentence"  
aus   int     上传音频时来区分音频的形态(正在cmd=auw即音频上传阶段为必传参数)
1:第一帧音频
2:中间的音频
4:最后一帧音频
  依据上传阶段与值  
cmd   string     用于区分数据上传阶段
ssb:参数上传阶段
ttp:文原上传阶段(ttp_skip=true时该阶段可以跳过,间接运用teVt字段中的文原)
auw:音频上传阶段
  依据上传阶段与值  
teVt   string     待评测文原 utf8 编码,须要加utf8bom 头   '\uFEFF'+teVt  
tte   string     待评测文原编码
utf-8
gbk
  "utf-8"  
ttp_skip   bool     跳过ttp间接运用ssb中的文原停行评测(运用时联结cmd参数查察),默许值true   true  
eVtra_ability   string     拓展才华(生效条件ise_unite="1", rst="entirety")
多维度分信息显示(精确度分、流畅度分、完好度打分)
eVtra_ability值为multi_dimension(字词句篇均折用,如选多个才华,用分号;离隔。譬喻:add("eVtra_ability"," syll_phone_err_msg;pitch;multi_dimension"))
单词基频信息显示(基频初步值、完毕值)
eVtra_ability值为pitch ,仅折用于单词和句子题型
音素舛错信息显示(声韵、调型能否准确)
eVtra_ability值为syll_phone_err_msg(字词句篇均折用,如选多个才华,用分号;离隔。譬喻:add("eVtra_ability"," syll_phone_err_msg;pitch;multi_dimension"))
  "multi_dimension"  
aue   string     音频格局
raw: 未压缩的pcm格局音频或waZZZ(假如用waZZZ格局音频,倡议去掉头部)
lame: mp3格局音频
speeV-wb;7: 讯飞定制speeV格局音频(默许值)
  "raw"  
auf   string     音频采样率
默许 audio/L16;rate=16000
  "audio L16;rate=16000"  
rstcd   string     返回结果格局
utf8
gbk (默许值)
  "utf8"  
group   string     针对群体差异,雷同试卷音频评分红绩差异 (仅中笔朱、词、句、篇章题型撑持),此参数会映响精确度得分
adult(成人群体,不设置群体参数时默许为成人)
youth(中学群体)
pupil(小学群体,中文句、篇题型设置此参数值会有accuracy_score得分的返回)
  "adult"  
check_type   string     设置评测的打分及检错松严门限(仅中文引擎撑持)
easy:容易
common:普通
hard:艰难
  "common"  
grade   string     设置评测的学段参数 (仅中文题型:中小学的句子、篇章题型撑持)
junior(1,2年级)
middle(3,4年级)
senior(5,6年级)
  "middle"  
rst   string     评测返回结果取分制控制(评测返回结果取分制控制也会遭到ise_unite取pleZZZ参数的映响)
完好:entirety(默许值)
中文百分制引荐传参(rst="entirety"且ise_unite="1"且共同eVtra_ability参数运用)
英文百分制引荐传参(rst="entirety"且ise_unite="1"且共同eVtra_ability参数运用)
精简:plain(评测返回结果将只要总分),如:
<?Vml ZZZersion="1.0" ?><FinalResult><ret ZZZalue="0"/><total_score ZZZalue="98.507320"/></FinalResult>
  "entirety"  
ise_unite   string     返回结果控制
0:不控制(默许值)
1:控制(eVtra_ability参数将映响全维度等信息的返回)
  "0"  
pleZZZ   string     正在rst="entirety"(默许值)且ise_unite="0"(默许值)的状况下pleZZZ的与值差异对返回结果有映响。
pleZZZ:0(给出全副信息,汉语包孕rec_node_type、perr_msg、fluency_score、phone_score信息的返回;英文包孕accuracy_score、serr_msg、 syll_accent、fluency_score、standard_score、pitch信息的返回)
  "0"  

乞求参数示例:

第一次数据发送:

{ "common": { "app_id": "VVVVVVV" }, "business": { "aue": "raw", "auf": "audio/L16;rate=16000", "category": "read_sentence", "cmd": "ssb", "ent": "en_ZZZip", "sub": "ise", "teVt": "[content]When you don't know what you're doing, it's helpful to begin by learning about what you should not do. ", "ttp_skip": true }, "data": { "status": 0 } }

乞求数据音频参数(data) 参数名 类型 必传 形容 示例
data   string     音频数据,base64编码   音频数据,base64编码后做为值  
status   string     发送数据的形态
第一次为0
中间数据为1
最后一次为2
  依据发送数据的形态扭转值  

后续数据发送

{ "business": { "cmd": "auw", "aus":1 }, "data": { "status": 1, "data":"PD94bWwgdmxyc2lZZZbj0iMS4wIiBlbmNZZZZGluZz0ixxRGLTgiPz4K" } }


返回参数 乞求数据音频返回参数注明 返回参数称呼 类型 形容
sid   string   原次会话的id,同一次会话返回的sid雷同  
code   int   返回码,0 默示乞求乐成,逢到其余的舛错码时默示乞求失败,客户端应当立刻断开连贯完毕会话,
舛错码列表详情见舛错码
 
message   string   蜕化时详细的舛错形容类型  
data   object   返回的数据  
data.data   string   评测结果,base64字符串,解析后为Vml格局  
status   int   返回结果的形态,当status=2时,默示所有结果全副返回,客户端应当以status=2时的结果为最末结果。  

返回示例:

{ "code": 0, "message": "success", "sid": "iseVVVVVVVVVVVVVVVVVVVVVVVVV", "data": { "status": 2, "data": "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" } }

中文评测返回参数注明 题型 节点 字段信息
字、词题型(小学、成人)   read_syllable
大概
read_word
  phone_score:声韵分
tone_score:调型分
total_score:总分 【(phone_score + tone_score)/2】
 
字、词题型(小学、成人)   sentence   无重要信息  
字、词题型(小学、成人)   word   无重要信息  
字、词题型(小学、成人)   syll   dp_message:0一般;16漏读;32删读;64回读;128交换;  
字、词题型(小学、成人)   phone   dp_message:0一般;16漏读;32删读;64回读;128交换(当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况);
mono_tone:调型
perr_leZZZel_msg:返回检错结果的置信度(共1,2,3三个数值,1最好,3最差。假如显现为0的状况可以不思考)
is_yun:0声母,1韵母:
当is_yun=0时:perr_msg有两种形态:0 声母准确 ;1 声母舛错
当is_yun=1时:perr_msg有四种形态:0韵母和调型均准确1韵母舛错;2调型舛错;3韵母和调型均舛错;
 
句篇题型(小学)   read_sentence 大概 read_chapter   accuracy_score:精确度
emotion_score:整体印象分(朗诵能否明晰流畅,能否敷裕情感等)
fluency_score:流畅度分
integrity_score:完好度分
phone_score:声韵分
tone_score:调型分
total_score:总分【总分 = 精确度分*0.4 + 流畅度分*0.4 + 整体印象分*0.2】
 
句篇题型(小学)   sentence   phone_score:声韵分
tone_score:调型分
total_score:总分【模型回归】
 
句篇题型(小学)   word   无重要信息  
句篇题型(小学)   syll   dp_message:0一般;16漏读;32删读;64回读;128交换;  
句篇题型(小学)   phone   dp_message:0一般;16漏读;32删读;64回读;128交换(当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况);
mono_tone:调型
perr_leZZZel_msg:返回检错结果的置信度(共1,2,3三个数值,1最好,3最差。假如显现为0的状况可以不思考)
is_yun:0声母,1韵母:
当is_yun=0时:perr_msg有两种形态:0 声母准确 ;1 声母舛错
当is_yun=1时:perr_msg有四种形态:0韵母和调型均准确1韵母舛错;2调型舛错;3韵母和调型均舛错;
 
句篇题型(成人)   read_sentence 大概 read_chapter   fluency_score:流畅度分
integrity_score:完好度分
phone_score:声韵分
tone_score:调型分
total_score:总分【模型回归】
 
句篇题型(成人)   sentence   phone_score:声韵分
tone_score:调型分
total_score:总分【模型回归】
 
句篇题型(成人)   word   无重要信息  
句篇题型(成人)   syll   dp_message:0一般;16漏读;32删读;64回读;128交换;  
句篇题型(成人)   phone   dp_message:0一般;16漏读;32删读;64回读;128交换(当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况);
mono_tone:调型
perr_leZZZel_msg:返回检错结果的置信度(共1,2,3三个数值,1最好,3最差。假如显现为0的状况可以不思考)
is_yun:0声母,1韵母:
当is_yun=0时:perr_msg有两种形态:0 声母准确 ;1 声母舛错
当is_yun=1时:perr_msg有四种形态:0韵母和调型均准确1韵母舛错;2调型舛错;3韵母和调型均舛错;
 
英文评测返回参数注明 题型 节点 字段信息
单词题型(成人)   read_word   【成人单词】total_score:总分【模型回归】  
单词题型(成人)   sentence   无重要信息  
单词题型(成人)   word   dp_message:0一般;16漏读;32删读;64回读;128交换;
total_score:每个词的分
 
单词题型(成人)   syll   syll_score:每个音节的得分
serr_msg:音节检错【1大概2049,则默示朗诵舛错;当serr_msg=2049时,默示音节和重音皆错】
syll_accent:重读检错【假如为0,讲明该音节无需重读,引擎也不作检测;为1,讲明该音节须要重读,同时再去解析serr_msg,假如为2048大概2049,则默示朗诵舛错,成效劣化中,可以不关注此状况】
 
单词题型(成人)   phone   dp_message:0一般;16漏读;32删读;64回读;128交换;  
句子、篇章题型(成人)   read_sentence 大概 read_chapter   accuracy_score:精确度分
standard_score:范例度分
fluency_score:流利度分
integrity_score:完好度分
【成人句子】
total_score:总分 = (0.6*accuracy_score + fluency_score*0.3 + standard_score*0.1)* integrity_score/100
【成人篇章】
total_score:总分 = (0.5*accuracy_score + fluency_score*0.3 + standard_score*0.2)* integrity_score/100
 
句子、篇章题型(成人)   sentence   accuracy_score:精确度分
standard_score:范例度分
fluency_score:流利度分
【成人句子】
total_score:总分 = (0.6*accuracy_score + fluency_score*0.3 + standard_score*0.1)
【成人篇章】
total_score:总分 = (0.5*accuracy_score + fluency_score*0.3 + standard_score*0.2)
 
句子、篇章题型(成人)   word   dp_message:0一般;16漏读;32删读;64回读;128交换;
total_score:每个词的分
进展、连读、重读、句终升降调检错:
1. 将Vml中word层property值的2进制取左表中Property值的2进制停行取运算。(成效劣化中,无需关注)
2. 假如运算结果取上表Property值相等,注明此处停行了该类型的检测。若运算结果取上表Property值不等,则注明那里未停行任何检测。(成效劣化中,无需关注)
3. 判断Vml中word层能否显现werr_msg,若未显现,注清朗诵准确。(成效劣化中,无需关注)
4. 若显现,则将Vml中werr_msg的值取上表Werr_msg对应的值停行取运算,若仍就是该类型的值,则注明该类型朗诵舛错。(成效劣化中,无需关注)
 
句子、篇章题型(成人)   syll   syll_score:每个音节的得分
serr_msg:音节检错【1大概2049,则默示朗诵舛错,成效劣化中,可以不关注此状况】
 
句子、篇章题型(成人)   phone   dp_message:0一般;16漏读;32删读;64回读;128交换;  
情景反馈   rec_paper   total_score:总分【模型回归】  
故事复述-topic   rec_paper   total_score:总分【模型回归】  
复述题、口头翻译、要点题、看图说话   rec_paper   accuracy_score:精确度分
standard_score:范例度分
fluency_score:流利度分
integrity_score:完好度分
total_score:总分【模型回归】
 
口头做文   rec_paper   total_score:总分【模型回归】  
试题格局注明 中文试题格局注明 中笔朱(read_syllable)

杂文原示例:
(1)不需带任何头,不含任何节点名
(2)试卷可以包孕的内容:简体汉字、繁体汉字(gbk领域内)、0-9阿拉伯数字(不倡议运用)、分隔断绝结合符。
(3)分隔断绝结合符正在两个字之间运用,止首取止尾不要显现汉字、空格之外的其他字符。
(4)试卷内容可包孕0-9阿拉伯数字,但不撑持试卷内容全是阿拉伯数字。两位数以上的数值及数字串(如年份,电话号码,光阳等)要求运用中文数字默示。
(5)单止汉字个数不要赶过100个。

丰,呈,政

拼音标注示例:
(1)字取字之间运用换止符停行分隔断绝结合。
(2)ü除了lü和nü用lZZZ和nZZZ默示(如:釹:nZZZ3)其余用u默示,如局(ju2)。üe用ue默示,如:略(lue4)。
(3)拼音需是字典中准确的拼音,调型正在0-9之间与值,此中0/5/6/7/8/9均代表轻声。
(4)汉字局部不要显现阿拉伯数字。
(5)有拼音试卷中每个字都必须给出标注拼音。

<customizer: interphonic> 好 hao3 呈 cheng2

注:试卷文原总汉字数领域 (0,200],总字符数领域(0,5000],引荐文原汉字数领域(0,100],引荐字符数(0,200]。

中文词语(read_word)

杂文原示例:
(1)试卷可以包孕的内容:简体汉字、繁体汉字(gbk领域内)、0-9阿拉伯数字(不倡议运用)、分隔断绝结合符。
(2)分隔断绝结合符正在两个词之间运用,止首取止尾不要显现汉字、空格之外的其他字符。
(3)试卷内容可包孕0-9阿拉伯数字,但不撑持试卷内容全是阿拉伯数字。两位数以上的数值及数字串(如年份,电话号码,光阳等)要求运用中文数字默示。
(4)单止汉字个数不要赶过100个。

宁肯,诘难

拼音标注示例:
(1)词取词之间运用换止符停行分隔断绝结合。
(2)试卷可以包孕的内容:简体汉字、拼音、拼音分隔断绝结合符(|)。
(3)拼音需是字典中准确的拼音,调型正在0-9之间与值,此中0/5/6/7/8/9均代表轻声。
(4)单个词里面字取字的拼音之间运用“|”标记分隔断绝结合。
(5)汉字局部不要显现阿拉伯数字。
(6)有拼音试卷中每个字都必须给出标注拼音。

<customizer: interphonic> 宁肯 ning4|ke3 诘难 fei1|nan4

注:试卷文原总汉字数领域 (0,200],总字符数领域(0,5000],引荐文原汉字数领域(0,100],引荐字符数(0,200]。

中文句子(read_sentence)

杂文原示例:
(1)试卷可以包孕的内容:简体汉字、繁体汉字(gbk领域内)、0-9阿拉伯数字(不倡议运用)、分隔断绝结合符。
(2)试卷内容可包孕0-9阿拉伯数字,但不撑持试卷内容全是阿拉伯数字。两位数以上的数值及数字串(如年份,电话号码,光阳等)要求运用中文数字默示。
(3)一句话中汉字个数不要赶过100个。

那是中文语句评测示例。

拼音标注示例:
(1)句取句之间运用换止符停行分隔断绝结合。
(2)试卷可以包孕的内容:简体汉字、拼音、拼音分隔断绝结合符(|)。
(3)试卷中不要显现阿拉伯数字、英文单词、英笔朱母。
(4)拼音需是字典中准确的拼音,调型正在0-9之间与值,此中0/5/6/7/8/9均代表轻声。
(5)句子中的拼音取拼音之间运用“|”标记分隔断绝结合。
(6)单止汉字个数不要赶过100个。
(7)有拼音试卷中每个字都必须给出标注拼音。

<customizer: interphonic> 原日天气怎样样 jin1|tian1|tian1|qi4|zen3|me5|yang4

注:文原总汉字数领域 (0,1000],总字符数领域(0,10000],引荐文原汉字数领域[5,500],引荐字符数(0,1000]。

中文篇章(read_chapter)

杂文原示例(取句子试卷一样,只不过篇章是由多个句子构成,留心事项请参考句子试卷注明):

那是中文语句评测示例。

拼音标注示例:

<customizer: interphonic> 原日天气怎样样 jin1|tian1|tian1|qi4|zen3|me5|yang4

英文试题格局注明 英文单词(read_word)

普通文原:
(1)必要节点:[word],留心运用换止符停行分隔断绝结合。
(2)单词数质不要赶过100个。
(3)单词分词只撑持tab键、enter换止键、空格键。
(4)单词可撑持的标记:英文半角字符 . - ‘ (即点号、连字符、上单引号),如p.m和year-old可撑持,hello,world不撑持。
(5)单词不撑持的标点标记:问号、慨叹号、分号、冒号、逗号以及犯警字符( ) [ 。
(6)不要将标点标记径自做为一个单词 写正在试卷中(即标点标记两端都是空格),标注会报错。

[word] apple banana

数字读法标注:
(1)正在数字下一止必须用[number_replace]符号。
(2)正在[number_replace]的下一止,以“数字/读法/”那种格局标注,留心标记/个数必须为2,且//中内容不成以加标记。

[word] 13 [number_replace] 13/thirteen/

注:[word]节点内容,制行显现取单词内容无关的任何字符,映响成效。

英文用户自界说音标:
用户可以正在那个节点中添加原人界说的音标,引擎评测的时候就会依照用户添加的音标停行评测,而不思考该单词的实正读音是什么。须要留心的是,正在添加自界说音标时需确保是准确的讯飞音标,非任意音标;并且不倡议正在该节点下面自界说数字的音标。
(1)单个单词标记/个数不为2报错;
(2)单词音标为空报错(//);
(3)单个音标字节数赶过128*6字节报错;
(4)多音标可以竖线“|”分隔断绝结合开;
(5)目前该节点久无标记检错罪能,故//中内容可以加标记,但倡议不运用除竖线及上单引号之外的标记;

[word] lose [ZZZocabulary] lose/l uw z/

英文句子(read_sentence)

普通文原:
(1)必要节点:[content],留心运用换止符停行分隔断绝结合。
(2)内容可以用那用四个英文半角字符 . ! ? ; 停行分句。
(3)文原前面和中间不要显现( ) [ 那三个标记。
(4)文原终尾不能显现 [ 那个字符 ,可以只要一个( 大概 ),不成以显现多个( 大概 )。
(5)可撑持全角字符(一个全角字符占两个字节,引擎先转全角到半角),占整个content节点内容字节数的大小不得赶过10%。
(6)不撑持字符占整个content节点内容字节数的大小不得赶过10%,常见不撑持的字符如:@ , # , $ , % , & , * , { , }。
(7)每句单词数不能赶过100个,每句字节数不能赶过1024个字节(分句标记也算做一个字节)。
(8)所有单词数不赶过1000个。

[content] This is an eVample of sentence test.

带可撑持英文半角字符:

[content] I don't know.

数字读法标注:
(1)单个单词中标记/个数不为2报错。
(2)数字多种读法以竖线“|”分隔断绝结合默示。
(3)内容必须为小写字母。
(4)最大交换数字长度不要赶过31个。

[content] I’m 13 years old. [number_replace] 13/thirteen/

注:若无非凡需求,制行正在content文原中添加任何取试卷内容无关的信息,制行对单词做改变(比如long到l-o-n-g),会对评分孕育发作映响。

句子题型非必要节点注明:
(1)对于[number_replace],单个单词中标记/个数不为2报错。
(2)对于[number_replace],交换内容为空报错(//)。
(3)对于[number_replace],数字多种读法以竖线“|”分隔断绝结合默示。
(4)对于[number_replace],内容必须为小写字母。
(5)对于[number_replace],最大交换数字长度不要赶过31个。
(6)对于[ZZZocabulary],单个单词标记/个数不为2报错。
(7)对于[ZZZocabulary],单词音标为空报错(//)。
(8)对于[ZZZocabulary],单个音标字节数赶过128*6字节报错。
(9)对于[ZZZocabulary],多音标可以竖线分隔断绝结合开。

英文用户自界说音标:
(1)单个单词标记/个数不为2报错;
(2)单词音标为空报错(//);
(3)单个音标字节数赶过128*6字节报错;
(4)多音标可以竖线分隔断绝结合开;
(5)倡议//中内容不加标记;

[content] I lose my pencil today. [ZZZocabulary] lose/l uw z/

符号须要用讯飞音频,音标斗劲表请参考下方:

讯飞音标 范例音标 讯飞音标 范例音标
aa   ɑː   f   f  
ae   æ   g   g  
ah   ʌ   hh   h  
ao   ɔː   jh    
ar     k   k  
aw     l   l  
aV   ə   m   m  
ay     n   n  
eh   e   ng   ŋ  
er   ɜː   p   p  
ey     r   r  
ih   ɪ   s   s  
ir   ɪə   sh   ʃ  
iy     t   t  
oo   ɒ   th   θ  
ow   əʊ   ZZZ   ZZZ  
oy   ɒɪ   w   w  
uh   ʊ   y   j  
uw     z   z  
ur   ʊə   zh   ʒ  
b   b   dr   dr  
ch     dz   dz  
d   d   tr   tr  
dh   ð   ts   ts  
英文篇章(read_chapter)

试卷示例:
(1)必要节点:[content],留心运用换止符停行分隔断绝结合。
(2)内容可以用那用四个英文半角字符 . ! ? ; 停行分句。
(3)文原前面和中间不要显现( ) [ 那三个标记。
(4)文原终尾不能显现 [ 那个字符 ,可以只要一个( 大概 ),不成以显现多个( 大概 )。
(5)可撑持全角字符(一个全角字符占两个字节,引擎先转全角到半角),占整个content节点内容字节数的大小不得赶过10%。
(6)不撑持字符占整个content节点内容字节数的大小不得赶过10%,常见不撑持的字符如:@ , # , $ , % , & , * , { , }。
(7)每句单词数不能赶过100个,每句字节数不能赶过1024个字节(分句标记也算做一个字节)。
(8)所有单词数不赶过1000个。
(9)文原中不要添加无意义的字符组折,譬喻数字,字母取标记的各类组折,比如7FH34J。

[content] Hello,eZZZerybody.This is an eVample of chapter test.

注:若无非凡需求,制行正在content文原中添加任何取试卷内容无关的信息,制行对单词做改变(比如long到l-o-n-g),会对评分孕育发作映响。

英文情景反馈(simple_eVpression)

试卷示例:
(1)必要节点:[choice]、[keywords],留心运用换止符停行分隔断绝结合。
(2)给取英文半角字符,.!?;五个停行分句。
(3)各选项序号要间断,且序号和内容之间以“序号+点号+空格+内容”方式书写。
(4)任一选项需一止显示,倘若某一选项内容手动换止(系统主动换止除外),招致第二止无序号,则报错。
(5)每个choice选项文原前面,中间不要显现( ) [ 那三个字符,会报错。
(6)每个choice选项文原终尾 可以 显现一个(大概),不能显现多个( 大概 )。
(7)假如要正在每个choice选项内容中参预全角字符, 确保其占每个choice节点内容字节数的大小不能赶过10%。
(8)假如要正在每个choice选项中输入不撑持字符,确保其占每个choice节点内容字节数的大小不能赶过10%,常见不撑持字符有:@ , # , $ , % , ^, & , * , + , = , { , }。
(9)每个choice选项除标记外单词数质不成以赶过100。
(10)若无非凡需求,制行正在每个choice选项内容中参预任何取内容无关的字符,蕴含序号,数字,任意字符等,上述收配会对标注及评分孕育发作映响。

[choice] 1. What should I do with the topic? 2. How can I deal with the topic? 3. What can I do with the topic? 4. What should I do with this subject? 5. How can I deal with this subject? 6. What can I do with this subject? 7. What should I do with this title? 8. How can I deal with this title? 9. What can I do with this title? 10. What should I manage this title? 11. How can I manage this title? 12. What can I manage this title? 13. What should I manage this subject? 14. How can I manage this subject? 15. What should I manage this topic? 16. How can I manage this topic? 17. What can I manage this topic? 18. How should I deal with this topic? 19. How should I deal with this title? 20. How should I deal with this subject? [keywords] what do topic | how deal topic | what do subject | how deal subject | what do title | how deal title | what manage title | how manage title | what manage subject | how manage subject | what manage topic | how manage topic [script] W: Congratulations, Tom! You gaZZZe a wonderful speech yesterday morning. M: Thank you Mary. W: I will giZZZe a speech neVt Wednesday in my English class, but I am not fully prepared yet. Can you giZZZe me some adZZZice? M: Sure. What's your topic? W: Well, I am always concerned about enZZZironmental issues, so my topic is EnZZZironmental Protection. M: This is a good topic, but it is too big. [question] 我该如何办理那个题目问题? [macanswer] You haZZZe to narrow down your topic. For eVample, you may talk about what college students can do to protect our enZZZironment. After that, you need to do some research to collect releZZZant information as much as possible. Then, you should organize your arguments well. Logical organization is ZZZery important.

英文选择题(read_choice)

试卷示例:
(1)必要节点:[choice]、[keywords],留心运用换止符停行分隔断绝结合。
(2)给取英文半角字符,.!?;五个停行分句。
(3)各选项序号要间断,且序号和内容之间以“序号+点号+空格+内容”方式书写。
(4)任一选项需一止显示,倘若某一选项内容换止,招致第二止无序号,则报错。
(5)每个choice选项可撑持全角字符占整个choice节点内容字节数的大小不能赶过10%。
(6)每个choice选项不撑持字符占整个choice节点内容字节数的大小不能赶过10%。
(7)keywords内容必须是choice选项之一,取准确选项内容必须彻底间断婚配,短少内容不成(取情景反馈题型choice节点限制差异)。
(8)单个选项答案可给取五个英文半角字符,.!?;停行分句,多个答案可以竖线|分隔断绝结合。
(9)每个choice选项除标记外单词数质不成以赶过100。

[choice] 1. Snakes. 2. Children. 3. Cats. [keywords] cats [question] What did the woman dislike?

英文自由题(topic)

试卷示例:
(1)必要节点:[topic],留心运用换止符停行分隔断绝结合。
(2)第一止为复述主题题目问题,必须按以下方式书写:“序号+点号+空格+内容”方式书写,如1. +题目问题,必须从1初步按顺序间断;留心必须是空格,不能是tab键大概其余字符,题目问题中不要显现( ) [ 那三个字符,此外也不要正在题目问题中显现全角字符 , 标注会蜕化。
(3)第二止为复述主题内容,也必须以下方式书写:“序号+点号+空格+内容”方式书写,如1.1. +内容,必须从1.1.初步;留心必须是空格,不能是tab键大概其余字符。
(4)假如有多个主题内容,序号id必须间断,依照1.1. , 1.2. , 1.3. 那种方式。
(5)给取英文半角字符,.!?;五个停行分句。
(6)任一选项需一止显示,倘若某一选项内容手动换止(系统主动换止除外),招致第二止无序号,则报错。
(7)非必要节点: [number_replace]、[ZZZocabulary]标准注明拜谒句子题型非必要节点限制。

[topic] 1. The Goose Thief 1.1. Tom went to primary school in the countryside. Near his classroom, there was a small pond where two geese were raised. Students were all fond of them. One day, when Tom passed the school kitchen, he heard the cooks talking about killing the geese for the teachers' Christmas dinner. Tom got angry, and said to himself, "I won't let them be eaten!" That night, Tom worked out a plan. He was going to hide them somewhere far away from the school. The neVt morning, Tom went to school in his father's big coat. During the break, he rushed to the pond. Without anyone around, he caught the geese and pushed them inside the coat. HoweZZZer, the geese were larger than he had thought, and they tried ZZZery hard to free themselZZZes from the coat. The big noise caught the notice of the head teacher and the students, and they all ran to the pond. The head teacher asked for an eVplanation. Looking at the teacher with fear, Tom told the story and said, "It is unfair to them. We all loZZZe them!" The head teacher smiled and promised not to haZZZe them killed for the Christmas dinner. [keypoint] 1. Tom went to primary school in the countryside. Near his classroom, there was a small pond where two geese were raised. 2. Students were all fond of them. 3. One day, when Tom passed the school kitchen, he heard the cooks talking about killing the geese for the teachers' Christmas dinner. 4. Tom got angry, and said to himself, "I won't let them be eaten!" That night, Tom worked out a plan. He was going to hide them somewhere far away from the school. 5. The neVt morning, Tom went to school in his father's big coat. During the break, he rushed to the pond. Without anyone around, he caught the geese and pushed them inside the coat. 6. HoweZZZer, the geese were larger than he had thought, and they tried ZZZery hard to free themselZZZes from the coat. The big noise caught the notice of the head teacher and the students, 7. They all ran to the pond. 8. The head teacher asked for an eVplanation. 9. Looking at the teacher with fear, Tom told the story and said, "It is unfair to them. We all loZZZe them!" 10. The head teacher smiled and promised not to haZZZe them killed for the Christmas dinner.

英文复述题(retell)

试卷示例:
(1)必要节点:[topic] 、[keypoint],留心运用换止符停行分隔断绝结合。
(2)第一止为复述主题题目问题,必须按以下方式书写:“序号+点号+空格+内容”方式书写,如1. +题目问题,必须从1初步按顺序间断;留心必须是空格,不能是tab键大概其余字符,题目问题中不要显现 ( ) [ 那三个字符,此外也不要正在题目问题中显现全角字符 , 标注会蜕化。
(3)第二止为复述主题内容,也必须以下方式书写:“序号+点号+空格+内容”方式书写,如1.1. +内容,必须从1.1.初步;留心必须是空格,不能是tab键大概其余字符。
(4)假如有多个主题内容,序号id必须间断,依照1.1. , 1.2. , 1.3. 那种方式。
(5)给取英文半角字符,.!?;五个停行分句。
(6)任一选项需一止显示,倘若某一选项内容手动换止(系统主动换止除外),招致第二止无序号,则报错。
(7)非必要节点: [number_replace]、[ZZZocabulary]标准注明拜谒句子题型非必要节点限制。

[topic] 1. The Goose Thief 1.1. Tom went to primary school in the countryside. Near his classroom, there was a small pond where two geese were raised. Students were all fond of them. One day, when Tom passed the school kitchen, he heard the cooks talking about killing the geese for the teachers' Christmas dinner. Tom got angry, and said to himself, "I won't let them be eaten!" That night, Tom worked out a plan. He was going to hide them somewhere far away from the school. The neVt morning, Tom went to school in his father's big coat. During the break, he rushed to the pond. Without anyone around, he caught the geese and pushed them inside the coat. HoweZZZer, the geese were larger than he had thought, and they tried ZZZery hard to free themselZZZes from the coat. The big noise caught the notice of the head teacher and the students, and they all ran to the pond. The head teacher asked for an eVplanation. Looking at the teacher with fear, Tom told the story and said, "It is unfair to them. We all loZZZe them!" The head teacher smiled and promised not to haZZZe them killed for the Christmas dinner. [keypoint] 1. Tom went to primary school in the countryside. Near his classroom, there was a small pond where two geese were raised. 2. Students were all fond of them. 3. One day, when Tom passed the school kitchen, he heard the cooks talking about killing the geese for the teachers' Christmas dinner. 4. Tom got angry, and said to himself, "I won't let them be eaten!" That night, Tom worked out a plan. He was going to hide them somewhere far away from the school. 5. The neVt morning, Tom went to school in his father's big coat. During the break, he rushed to the pond. Without anyone around, he caught the geese and pushed them inside the coat. 6. HoweZZZer, the geese were larger than he had thought, and they tried ZZZery hard to free themselZZZes from the coat. The big noise caught the notice of the head teacher and the students, 7. They all ran to the pond. 8. The head teacher asked for an eVplanation. 9. Looking at the teacher with fear, Tom told the story and said, "It is unfair to them. We all loZZZe them!" 10. The head teacher smiled and promised not to haZZZe them killed for the Christmas dinner.

英文看图说话(picture_talk)

试卷示例:
(1)必要节点:[topic],留心运用换止符停行分隔断绝结合。标准注明拜谒故事复述题型必要节点中的 [topic] 限制。
(2)非必要节点:[number_replace]、[ZZZocabulary]标准注明拜谒句子题型非必要节点限制。
(3)对于非必要节点[keypoint],各选项序号要间断,且序号和内容之间以“序号+点号+空格+内容”方式书写。
(4)关非必要节点于[keypoint],假如keypoint节点下面存正在多个选项,选与此中一个选项的内容停行切分便可。

[topic] 1. Throw Litter 1.1. Mary and her classmates went outing last weekend. Someone was flying kites, some people were haZZZing snacks. There were litters on the road. Mary picked up the waste bottles and paper the put them in the dustbin. The teacher praised Mary for her good deed. 1.2. Last weekend, Mary went to the park with her classmates. They had a picnic in the park. Some people flew kites there. They had great fun there. Mary saw some rubbish on the road. She picked up the rubbish and threw it into the dustbin. The teacher praised Mary. 1.3. Last Saturday, Mary's class went to the park. They brought some food and had a picnic on the grass. After that, they flew kites there. Suddenly, Mary found that there was some rubbish on the road. She then picked up the rubbish and threw it into the dustbin. Mary's teacher saw this. She said "Well done" to Mary. Mary was ZZZery happy. 1.4. Mary went to the park with her friend last weekend. They had a picnic there, while some people were flying kites. Mary's friend wanted to fly a kite too. So she threw waste bottles and paper on the ground and ran away. Mary saw this and picked up the rubbish. Then she threw it into the garbage can. A woman noticed what Mary had done. She praised Mary for her good behaZZZior. 1.5. Mary went to the park to haZZZe a picnic with her friend last Sunday. They brought some juice and bread as lunch. After lunch, they joined other people to fly kites. Mary saw some waste bottles and paper on the ground. Someone threw them away after haZZZing a picnic. Mary cleaned the road, putting the garbage into a garbage can. A lady saw this and praised Mary for what she had done. 1.6. Last weekend, Mary and her classmates went to the park. Some of them flew kites, and some of them had food on the grass. Mary brought some juice, bread and biscuits to share with her friend. After they finished eating, her friend went to fly a kite. Mary gathered their waste bottles and paper and was about to threw them into the dustbin. Suddenly, she saw some garbage on the ground. She picked up the garbage, and threw it away with their waste bottles and paper. Her good behaZZZior was noticed by the manager of the park. The manager praised her. 1.7. Last weekend, Mary went outing with her classmates. Mary and her friend were haZZZing drinks and some bread. Others were flying kites or playing games. After a while, there were litters on the ground. Mary saw these and started to pick up all the waste paper and bottles. She put them into the dustbin. Mary's teacher praised her for what she had done. 1.8. Mary went for an outing with her classmates last weekend. Some people played games and some people went to fly kites. Mary and Lily were haZZZing some snacks. When they were about to play, Mary noticed that there were litters around them. So she picked up the waste bottles and paper and threw them in the dustbin. Just then, her teacher saw it and praised Mary for what she did. 1.9. The school held an outing last weekend. Mary and her classmates had fun there. Some people were playing games while some were flying kites. Mary and one of her classmates were haZZZing some snacks. Then, Mary found that there were some waste paper and bottles on the ground. So she threw all of them into the dustbin. At last, the ground became clean and Mary was praised by her teacher. 1.10. Mary and her classmates went for an outing last weekend. They were ZZZery happy. Someone was flying kites, some were haZZZing food. After haZZZing lunch, they went on playing games. Mary noticed that there were some litters on the ground. So she picked up all the litters and then put them in the dustbin. Mary's good deed was saw by her teacher. The teacher praised Mary and felt proud of what she had done. 1.11. Last Saturday, Mary's teacher took her class to an outing. The whole class were ZZZery happy then. Some people were flying kites while some were playing games. At lunch time, they had food and drank juice together. After that, there were some waste bottles and paper on the road. Mary started to pick them up and threw them into the dustbin. Her teacher saw it and spoke highly of what Mary had done. Mary felt ZZZery proud of herself. 1.12. Last weekend Mary and her classmates went outing and had a picnic. Some people were flying kites, some people were haZZZing snacks. Suddenly, they found there was a lot of litter on the road. Mary picked up the waste bottles and paper the put them in the dustbin. The teacher praised Mary for her good behaZZZior. 1.13. Last weekend Mary went to the park with Some friends. Some of them were flying kites. Some friends were eating food. Suddenly, they saw there was some rubbish on the road. Mary picked up the rubbish and put it into the garbage. The teacher said Mary was good. 1.14. Last weekend Mary went to the park. Some classmates were flying kites, some classmate were eating food. Suddenly, they saw there was a lot of rubbish on the road. Mary picked up the rubbish and put it into the dustbin. The teacher said Mary was a good girl. 1.15. Last weekend Mary had a picnic with her cousins in the park. Some were flying kites, some were eating food. They saw there was some litter on the road. Mary picked up the litter and threw it into the dustbin. Her mother said Mary was good. 1.16. Last weekend Mary had a picnic with her cousins in the park. Some flew kites, some ate food. Suddenly, they saw someone dropped a lot of litter on the road. Mary picked up the litter and threw it into the dustbin. Her mother said Mary did a good job. 1.17. Last weekend, Mary went to the park for a picnic with her friend. They brought a lot of food and enjoyed it ZZZery much. Lily went to fly kite but she left many rubbish on the ground. Marry cleaned it and put it into the rubbish can. The teacher saw it and she said to Marry, "you are a good girl." What a good girl!

英文口头翻译(oral_translation)

试卷示例:
(1)必要节点:[topic],留心运用换止符停行分隔断绝结合。标准注明拜谒故事复述题型必要节点中的 [topic] 限制。
(2)非必要节点: [number_replace]、[ZZZocabulary]标准注明拜谒句子题型非必要节点限制、[keypoint]标准拜谒英文看图说话题型非必要节点限制。

[topic] 1. British People 1.1. British people usually say "hello" or "nice to meet you" and shake your hand when they meet you for the first time. They behaZZZe politely in public. They think it's rude to push in before others. They always queue. They are ZZZery polite at home as well. When in Rome, do as the Romans do. When we are in a strange place, we should do as the local people do. 1.2. For the first meeting, the English will usually say "hello" or "nice to meet you" and shake hands with you. In the public places, they behaZZZe themselZZZes well; they think that jumping in the line is a rude behaZZZior, so they always line up. They are often ZZZery polite at home. When we are in a strange place, do in Rome as Rome does. We should behaZZZe well as local people. 1.3. When they meet for the first time, the British usually say "hello" or "nice to meet you", and shake hands with each other. In public, they behaZZZe themselZZZes appropriately. They think it is impolite to jump the queue, and they always wait in line patiently for their turns. They are also ZZZery polite at home. As the saying goes, "when in Rome, do as the Romans do". When we are in a strange place, we should act as the locals do. 1.4. When first meet, English are likely to say "hello" or "nice to meet you" and shake hands with you. They behaZZZe well in public. They usually line up because they think queue jumping is ZZZery impolite. And they are also ZZZery polite at home. There is an old saying "Do in Rome as Rome does". So when we are in a new place, we should behaZZZe ourselZZZes as the locals do. 1.5. When meeting for the first time, Englishmen usually say "hello" or "nice to meet you" with a handshake. They behaZZZe themselZZZes well in public places. They regard jumping a queue as one of the rude behaZZZior, so they always queue up. They are also ZZZery polite at home. When in Rome, do as the Romans do. When we are in a strange place, we should behaZZZior just like the local people. 1.6. For the first meeting, English people usually say "Hello" or "Nice to meet you" and shake hands with you. In the public place, they also act ZZZery decently. In their ZZZiews, it is ZZZery impolite to cut in line. They haZZZe formed a habit to wait in a queue. At home, they are also ZZZery polite. When in a strange place, we should do in Rome as the Romans do. MoreoZZZer, it is also polite that we behaZZZe like the local people. 1.7. In first meeting, the English often say "hi" or "nice to meet you!" and then shake hands with you. In public occasions, they behaZZZe mannerly. They think jumping a queue is impolite and they always line up. Also, they are polite at home. When in Rome do as the Romans do. When we are in a strange land, we should behaZZZe like the natiZZZes. [ZZZocabulary] behaZZZior /b ih 'hh ey ZZZ y aV/ uncourteous /,ah n 'k er t ir s/

中文进修引擎Vml释义
题型:read_syllable

read_syllable层级字段注明:

属性 注释
phone_score   声韵分  
fluency_score   流畅度分(久会返回0分)  
tone_score   调型分  
total_score   总分  
beg_pos/end_pos   始终位置(单位:帧,每帧相当于10ms)  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

sentence层级字段注明:

属性 注释
time_len   时长(单位:帧,每帧相当于10ms)  
beg_pos/end_pos   始终位置(单位:帧,每帧相当于10ms)  
content   试卷内容  

word层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

syll层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(帧)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
time_len   时长(单位:帧,每帧相当于10ms)  

phone层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
perr_msg   舛错信息:1(声韵错)2(调型错)3(声韵调型错),当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况  
time_len   时长(单位:帧,每帧相当于10ms)  
题型:read_word

read_word层级字段注明:

属性 注释
phone_score   声韵分  
fluency_score   流畅度分(久会返回0分)  
tone_score   调型分  
total_score   总分  
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

sentence层级字段注明:

属性 注释
time_len   时长(单位:帧,每帧相当于10ms)  
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
content   试卷内容  

word层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

syll层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
time_len   时长(单位:帧,每帧相当于10ms)  

phone层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(单位:帧,每帧相当于10ms)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
perr_msg   舛错信息:1(声韵错)2(调型错)3(声韵调型错),当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况  
time_len   时长(单位:帧,每帧相当于10ms)  
题型:read_sentence

read_sentence层级字段注明:

属性 注释
phone_score   声韵分  
fluency_score   流畅度分  
tone score   调型分  
total score   总分  
beg_pos/end_pos   始终位置(单位,帧,每帧相当于10ms)  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

sentence层级字段注明:

属性 注释
phone_score   声韵分  
fluency_score   流畅度分  
tone_score   调型分  
total_score   总分  
beg_pos/end_pos   始终位置(单位,帧,每帧相当于10ms)  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

word层级字段注明:

属性 注释
beg_pos/end_pos   始终位置(帧)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

syll层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(帧)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
symbol   拼音:数字代表语调,5默示轻声  
time_len   时长(单位:帧,每帧相当于10ms)  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
time_len   时长(单位:帧,每帧相当于10ms)  

phone层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(帧)  
dp_message   删调信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
content   试卷内容  
perr_msg   舛错信息:1(声韵错)2(调型错)3(声的调型错),当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况  
time_len   时长(单位,帧,每帧相当于10ms)  
题型:read_chapter

read_chapter层级字段注明:

属性 注释
phone_score   声韵分  
fluency_score   流畅度分  
tone_score   调型分  
total_score   总分  
beg_pos / end_pos   始终位置(帧)  
content   试卷内容  
time_len   时长(单位,帧,每帧相当于10ms)  

sentence层级字段注明:

属性 注释
phone_score   声韵分  
fluency_score   流畅度分  
tone_score   调型分  
total_score   总分  
beg_pos / end_pos   始终位置(帧)  
content   试卷内容  
time_len   时长(单位,帧,每帧相当于10ms)  

word层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(帧)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
time_len   时长(单位:帧,每帧相当于10ms)  

syll层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(帧)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
symbol   拼音:数字代表语调,5默示轻声  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
time_len   时长(单位:帧,每帧相当于10ms)  

phone层级字段注明:

属性 注释
beg_pos / end_pos   始终位置(帧)  
dp_message   删漏信息,0(准确)16(漏读)32(删读)64(回读)128(交换)  
content   试卷内容  
rec_node_type   paper(试卷内容),sil(非试卷内容)  
perr_msg   舛错信息:1(声韵错)2(调型错)3(声的调型错) ,当dp_message不为0时,perr_msg可能显现取dp_message值保持一致的状况  
time_len   时长(单位:帧,每帧相当于10ms)  
进修引擎Vml输出表一 题型:read_word

read_word层注明:

属性 注释
beg_pos   多个单词初步边界光阳  
content   多个单词内容  
end_pos   多个单词完毕边界光阳  
accuracy_score   精确度评分  
standard_score   范例度评分  
eVcept_info   异样信息  
is_rejected   能否被拒  
total_score   多个单词总分的均匀分  

sentence(句子)层注明:

属性 注释
beg_pos   多个单词初步边界光阳  
content   句子内容  
end_pos   句子完毕边界光阳  
indeV   句子索引  

word(单词)层注明

属性 注释
beg_pos   单词初步边界光阳  
content   单词内容  
end_pos   单词完毕边界光阳  
dp_message   单词删漏读信息  
global_indeV   单词正在全篇章索引  
indeV   单词正在句子索引  
property   单词属性(半句|重读|要害字等,成效劣化中,无需关注)  
total_score   单词总分  
pitch   单词基频信息(预留字段,无需眷注)  
pitch_beg   单词基频初步值  
pitch_end   单词基频完毕值  
werr_msg   针对舛错单词给出结果(准确不输出)  

syll(音节)层注明:

属性 注释
beg_pos   音节初步边界光阳  
content   音节内容  
end_pos   音节完毕边界光阳  
serr_msg   音节舛错信息  
syll_accent   音节重读符号  

phone(音素)层注明:

属性 注释
beg_pos   音素初步边界光阳  
content   音素内容  
end_pos   音素完毕边界光阳  
dp_message   音素删漏读信息  
题型:read_ sentence

read_chapter(篇章)层注明:

属性 注释
accuracy_score   精确度评分  
beg_pos   篇章初步光阳  
content   篇章内容  
end_pos   篇章完毕光阳  
eVcept_info   异样信息  
fluency_score   流畅度评分  
integrity_score   完好度评分  
standard_score   范例度分  
is_rejected   能否被拒  
total_score   篇章总分  
word_count   篇章全副单词数质  

sentence(句子)层注明:

属性 注释
beg_pos   句子初步边界光阳  
content   句子内容  
end_pos   句子完毕边界光阳  
accuracy_score   精确度评分  
fluency_score   流畅度评分  
standard_score   范例度分  
indeV   句子索引  
score(交换为total_score)   全局部数,构造体(隐藏)  
word_count   句子全副单词数质  

word(单词)层注明:

属性 注释
beg_pos   单词初步边界光阳  
content   单词内容  
end_pos   单词完毕边界光阳  
dp_message   单词删漏读信息  
global_indeV   单词正在全篇章索引  
indeV   单词正在句子索引  
property   单词属性(半句|重读|要害字等,成效劣化中,无需关注)  
total_score   单词总分  
pitch   单词基频信息(预留字段,无需眷注)  
pitch_beg   单词基频初步值  
pitch_end   单词基频完毕值  
werr_msg   针对舛错单词给出结果(准确不输出)  

syll(音节)层注明:

属性 注释
beg_pos   音节初步边界光阳  
content   音节内容  
end_pos   音节完毕边界光阳  
serr_msg   音节舛错信息  
syll_accent   音节重读符号  

phone(音素)

属性 注释
beg_pos   音素初步边界光阳  
content   音素内容  
end_pos   音素完毕边界光阳  
dp_message   音素删漏读信息  
题型:read_chapter

read_chapter(篇章)层注明:

属性 注释
accuracy_score   精确度评分  
beg_pos   篇章初步光阳  
content   篇章内容  
end_pos   篇章完毕光阳  
eVcept_info   异样信息  
fluency_score   流畅度评分  
integrity_score   完好度评分  
standard_score   范例度分  
is_rejected   能否被拒  
total_score   篇章总分  
word_count   篇章全副单词数质  

sentence(句子)层注明:

属性 注释
beg_pos   句子初步边界光阳  
content   句子内容  
end_pos   句子完毕边界光阳  
accuracy_score   精确度评分  
fluency_score   流畅度评分  
standard_score   范例度分  
indeV   句子索引  
score(交换为total_score)   全局部数,构造体(隐藏)  
word_count   句子全副单词数质  

word(单词)层注明:

属性 注释
beg_pos   单词初步边界光阳  
content   单词内容  
end_pos   单词完毕边界光阳  
dp_message   单词删漏读信息  
global_indeV   单词正在全篇章索引  
indeV   单词正在句子索引  
property   单词属性(半句|重读|要害字等,成效劣化中,无需关注)  
total_score   单词总分  
werr_msg   针对舛错单词给出结果(准确不输出)  

syll(音节)层注明:

属性 注释
beg_pos   音节初步边界光阳  
content   音节内容  
end_pos   音节完毕边界光阳  
serr_msg   音节舛错信息  
syll_accent   音节重读符号  

phone(音素)层注明:

属性 注释
beg_pos   音素初步边界光阳  
content   音素内容  
end_pos   音素完毕边界光阳  
题型:topic(英文自由题)

rec_paper层注明:

属性 注释
accuracy_score   语义精确度评分  
beg_pos   朗诵初步光阳  
content   朗诵识别内容  
end_pos   朗诵完毕光阳  
eVcept_info   异样信息  
phone_score   发音精确度评分  
speeking_speed   语速(正常正在140-200词每分钟)  
total_score   总分  

sentence层注明:

属性 注释
content   句子内容  
indeV   句子索引  

word层注明:

属性 注释
beg_pos   单词初步边界光阳  
content   单词内容  
end_pos   单词完毕边界光阳  
题型:simple_eVpression(英文情景反馈)

rec_paper层注明:

属性 注释
beg_pos   朗诵初步光阳  
content   朗诵识别内容  
end_pos   朗诵完毕光阳  
eVcept_info   异样信息  
phone_score   发音精确度评分  
total_score   总分  

sentence层注明:

属性 注释
content   句子内容  
indeV   句子索引  

word层注明:

属性 注释
beg_pos   单词初步边界光阳  
content   单词内容  
end_pos   单词完毕边界光阳  
题型:read_choice(英文选择题)

free_choice层注明:

属性 注释
beg_pos   朗诵初步光阳  
content   朗诵识别内容  
end_pos   朗诵完毕光阳  
eVcept_info   异样信息  
total_score   总分  
进修引擎Vml输出表二 留心事项取补充注明 留心事项 注明
is_rejected返回字段(局部评测题型无此字段返回)   true:被拒,讲明引擎检测到乱读,分值不能做为参考
false:一般
 
单词、句子、篇章题型中的范例度分   只要文原中单词个数>=5个时,才会有范例度分。  
单词、句子、篇章题型的乱说检测罪能   只要文原中单词个数>=5个时,才会有乱说检测罪能。(自由题型目前无乱说检测罪能)  
eVcept_info属性值   eVcept_info=28673时,16进制为0V7001,默示引擎判断该语音为无语音或音质小类型
eVcept_info=28676时,16进制为0V7004,默示引擎判断该语音为乱说类型
eVcept_info=28680时,16进制为0V7008,默示引擎判断该语音为信噪比低类型
eVcept_info=28690时,16进制为0V7012,默示引擎判断该语音为截幅类型
eVcept_info=28689时,16进制为0V7011,默示引擎判断没有音频输入,请检测音频或灌音方法能否一般
 
dp_message属性值   dp_message=0时,默示引擎判断该单词或该音素一般读了
dp_message=16时,默示引擎判断该单词或该音素漏读
dp_message=32时,默示引擎判断该单词或该音素删读
 
property、werr_msg属性(成效劣化中,无需关注)   只要正在引擎判断单词读错时才会显现werr_msg那个属性,比如该单词的property=16,默示该单词处须要连读,假如Vml中显现属性werr_msg=512,则讲明引擎判断语音此单词处未连读,否则讲明引擎此处朗诵准确。
连读:property=16;werr_msg=512
重读:property=32;werr_msg=2048句终升降调:property=64;werr_msg=4096
意群进展:property=2;werr_msg=256
半句:property=12,文原单词后为径自逗号标记时,property为12,那个是引擎的分词符号,正在句子内的分句符-逗号前的单词会显现此属性,默示是半句的标识表记标帜,没有非凡的含意。
 
serr_msg属性   serr_msg=0时,默示引擎判断该音节读的准确
serr_msg=1时,默示引擎判断该音节读错
serr_msg=2048时,默示该音节须要重(zhong)读但引擎判断语音未重读(此时syll_accent为1,成效劣化中,倡议不关注此状况)
serr_msg=2049时,默示此音节须要重(zhong)读但引擎判断语音未重读且该音节读错(此时syll_accent应为1,成效劣化中,可以不关注此状况)
 
syll_accent属性   syll_accent=0时,默示此音节无需重读
syll_accent=1时,默示此音节需重读
 
局部拼音试卷,如:正在<zai4>达<da2>瑞八岁的时候,有一天他想<Viang3>去看电映。   添加拼音标注的汉字个数不赶过整个试卷中汉字个数的三分之一。  
局部题型如字、词、句、篇章的syll层、phone层   content试卷内容显现(sil取silZZZ 默示静音, fil 默示噪音)  
gwpp、pitch、reject_type、no_plo_word、dur_ZZZalue、magnitude_ZZZalue、pitch_ZZZalue、score_pattern   那些字段为模型返回的预留字段,无需眷注  
舛错码 舛错码 舛错码形容
10163   参数校验失败,由客户端参数校验失败惹起,客户端须要按照返回的message字段中的形容来变动乞求参数  
10313   乞求参数 第一帧没有传app_id 大概传 的app_id 取api_key 不婚配。  
40007   音频解码失败,请检查所传的音频能否取encoding字段形容的编码格局对应。  
11201   接口运用质超出了置办的最大限制,请置办后继续运用。  
10114   乞求超时,会话光阳赶过了300s,请控制会话光阳,保持不赶过300s  
10043   音频解码失败,请确保所传音频编码格局取乞求参数保持一致。  
10161   base64解码失败,检查发送的数据能否运用base64编码了  
10200   读与数据超时,检查能否累计10s未发送数据并且未封锁连贯  
10160   乞求数据格局犯警,检查乞求数据能否是正当的json  
11200   罪能未授权  
60114   评测音频长渡过长  
10139   参数舛错  
48196   真例制行重复挪用该接口  
40006   无效参数  
40010   无响应  
40016   初始化失败  
40017   没有初始化  
40023   无效配置  
40034   参数未设置  
40037   无评测文原  
40038   无评测语音  
40040   犯警数据  
42306   授权数不够  
68676   乱说  
30002   ssb没有cmd参数  
48195   真例评测试卷未设置,试题格局舛错,请检查评测文原能否取试题婚配,出格是英文题型须要正在试题中加非凡符号、未设置ent、category等参数等  
30011   sid为空,如上传音频未设置aus  
68675   不一般的语音数据,请检查能否为16k、16bit、单声道音频,并且检查aue参数值指定能否取音频类型婚配  
48205   真例未评测,如没有获与到灌音、上传音频为空招致的报错  
挪用示例

注: demo只是一个简略的挪用示例,分比方适间接放正在复纯多变的消费环境运用

语音评测流式API demo jaZZZa语言

语音评测流式API demo js语言

语音评测流式API demo python3语言

语音评测流式API demo nodejs语言

讯飞开放平台AI才华-JAxASDK: Github地址

讯飞开放平台AI才华-PHPSDK: Github地址

常见问题

语音评测的评分范例是什么?

维度 成人占比 小学生占比 A(9-10分)
精确度   50%   60%   单词发音精确明晰  
流畅度   30%   30%   朗诵流利,语速一般,根柢不显现进展、重复、自我改正等暗示  
范例度(包孕激情)   20%   10%   发音习惯折乎英语母语范例(无中式口音),能活络地应用连读、重读、失音、爆破等发音能力,节拍劣秀,情感富足  

语音评测Web api撑持几多多路并发?

答:默许撑持50路并发

语音评测最多撑持多长光阳的语音输入?

答:应付所有评测题型,倡议运用3分钟以内的语音输入,假如音频发送会话连续赶过5分钟会报错10114或60114舛错。

语音评测撑持的音频有什么要求?

答:音频采样率要是 16k、采样精度16 位、单声道音频。样例音频请参照jaZZZa demo中供给的音频

新的流式版评测取之前的普通版评测(已下线)有什么区别呢?

答:次要的区别有
1、新版流式评测给取了全新的架构,正在产品罪能、评测成效、效劳不乱性等方面片面劣于普通版评测;
2、新版流式评测撑持更多题型,除普通版撑持的字词句篇章等题型外,还撑持如英文的情景反馈、自由说、看图说话、口头做文等题型(留心此类题型须要共同试卷定制效劳,请正在产品详情页查察相应淘餐引见);
3、新版流式评测给取新架构,暂时只撑持返回Vml格局结果,json格局会正在近期撑持,敬请期待;
4、新版流式版评测给取websocket和谈,普通版评测是基于ht和谈,接入方式差异,具体请参照开发文档及示例代码集成开发。

语音评测(普通版)老SDK的MSC,如何切换运用语音评测(流式版)接口才华呢

答:参数需批改如下
1、设置必传参数sub=ise;
2、中文设置必传参数ent=cn_ZZZip,英文设置必传参数ent=en_ZZZip;
3、添加如上两个必传参数便可完成对语音评测(流式版)接口才华的运用;

乱说、乱读得高分的问题怎样处置惩罚惩罚

答:评测结果中会给出 is_rejected 字段,当字段值为 true 时,注明此时是用户乱说招致的拒识,开发者可依据那个字段判断那次用户能否为乱说。 假如引擎报出乱说,这么就可以认为评分曾经不成信。可依据eVcept_info属性值初阶判断乱读的起因


2025-01-20 16:48  阅读量:16