短语音识别的标准版相当常见，比如微信上的语音转文字。对方发送的语音可以直接转换成文字。一般来说，60秒内的语音被准确识别为文本，适用于手机语音输入、智能语音交互、语音指令、语音搜索等短语音交互场景。
[En]
The standard version of short speech recognition is quite common, such as voice-to-text conversion on Wechat. The voice sent by the other party can be converted directly into text. Generally speaking, speech within 60 seconds is accurately recognized as text, which is suitable for short voice interaction scenarios such as mobile phone voice input, intelligent voice interaction, voice instructions, voice search and so on.
各大云工厂都在做这项功能，一般功能如下：
[En]
This function is being done by all major cloud factories, and the general features are as follows:
技术领先识别准确
采用国际领先的流媒体端到端语音语言集成建模方法，结合百度自然语言处理技术，近场普通话识别准确率高达98%。
[En]
Using the leading international streaming end-to-end speech language integration modeling method, integrated with Baidu natural language processing technology, the near-field Mandarin recognition accuracy is up to 98%.
多语种和多方言识别
支持普通话和略带口音的中文识别；支持广东话和四川话识别；支持英语识别。
[En]
Support Mandarin and slightly accented Chinese recognition; support Cantonese and Sichuan dialect recognition; support English recognition.
深度语义解析
支持50多个领域的语义理解，如：天气，交通，娱乐等。还可接入智能对话定制与服务平台UNIT自定义语义理解和对话服务，让您更准确地理解用户意图。
中文标点智能断句
使用大型数据集训练语言模型，以智能地匹配适当的标点符号(包括，.！？)使识别结果的表达更加通俗易懂。
[En]
Train the language model using large data sets to intelligently match appropriate punctuation (including,.!? ) to make the expression of the recognition result more understandable.
数字格式智能转换
根据对语音内容的理解，将数字序列、小数、时间、分数和基本运算符正确转换为数字格式，使识别的数字结果更符合使用习惯，直观自然。
[En]
According to the understanding of speech content, the digital sequence, decimal, time, fraction and basic operators can be correctly converted into digital format, which makes the recognized digital results more in line with the habit of use, intuitive and natural.
支持自助训练专属模型
语音自助培训平台支持自助培训模式，零码上传即可完成培训。业务领域的词汇识别率可精准提升5%-25%，可独家使用。
[En]
Self-help training model is supported on the voice self-training platform, and zero code can be uploaded to complete the training. The vocabulary recognition rate in the business domain can be accurately improved by 5-25%, and can be used exclusively.
不仅是开头的语音，还有其他应用场景：
[En]
Not only the voice at the beginning, but also other application scenarios:
语音输入
破除生词和拼音障碍，使用语音即时输入。可以有效识别带有轻微口音的普通话、广东话、四川话和英语，并可以根据句子的意思添加自动纠错和自动标点符号，使输入更快，沟通更顺畅。
[En]
Get rid of obscure words and pinyin barriers and use voice instant input. Mandarin, Cantonese, Sichuan dialect and English with slight accent can be effectively identified, and automatic error correction and automatic punctuation can be added according to the meaning of the sentence, making the input faster and the communication more smooth.
语音搜索
搜索内容直接语音输入，应用于网页搜索、汽车搜索、手机搜索等多种搜索场景，解放双手让搜索更高效，适用于视频网站、智能硬件、手机厂商等行业。
[En]
The search content is input directly by voice, which is applied to web search, car search, mobile search and other search scenarios, liberating hands to make search more efficient, suitable for video websites, smart hardware, mobile phone manufacturers and other industries.
语音指令
无需手动操作，可以通过语音直接对设备或者软件发布指令，控制操作，适用于智能硬件、车载系统、机器人、手机APP、游戏等多个领域。
社交聊天
在社交聊天中，语音输入直接转换为文本，更加方便，或者当您收到不适合播放的语音消息时，可以将其转换为文本进行查看，以满足更多聊天场景。
[En]
In social chat, the voice input is directly converted to text, which makes it more convenient, or when you receive a voice message that is not suitable for playback, it can be converted to text for viewing to meet more chat scenarios.
游戏娱乐
在游戏中聊天是必不可少的，双手都不能打字，语音输入可以把语音变成文字，让用户在操作的同时也能直观的看到聊天内容，多元化的满足用户的聊天需求。
[En]
Chat in the game is essential, both hands can not type, voice input can turn voice chat into text, so that users can also intuitively see the chat content while operating, diversified to meet the chat needs of users.
Original: https://blog.csdn.net/m0_66194642/article/details/123735742
Author: 打工人何苦为难打工人
Title: 云-短语音识别
原创文章受到原创版权保护。转载请注明出处：https://www.johngo689.com/514973/
转载文章受原作者版权保护。转载请注明原作者出处！