"Journal of Chinese Phonetics" recently published an article by Zu Yiqing et al. "Recording Endangered Languages ​​Using bet365 casino bonus Replication Method—The Case of Xibe bet365 casino bonus"。This article proposes a new method of recording endangered languages ​​using artificial intelligence technology，From this we see the in-depth combination of large-scale corpora and AI technology for linguistic research、bet365 casino bonus resource protection and AI technology development will play an important role in promoting。

　Preserving bet365 casino bonus ​​through mainstream languages

　2017，American scientists proposed the Speaking Rosetta project，Aims to document unknown languages ​​without writing (usually bet365 casino bonus) through phonetic parallelism of unknown languages ​​and textual parallels of known languages。

Zu Yiqing and others in the "Intelligent Speech Technology and Its Industrialization - Intelligent Speech Technology and Systems for Ethnic Minorities" project of the Ministry of Industry and Information Technology in 2015，Carry out speech synthesis work in Xibe bet365 casino bonus。This work led them to the idea of ​​linguistic replication of endangered languages: using speech synthesis、Technologies such as speech recognition and machine translation for linguistic replication of endangered languages。Multi-bet365 casino bonus、Experience accumulated in speech recognition systems for multiple speakers in bet365 casino bonus classification and bet365 casino bonus commonality research，Can be used for reference in the field of endangered bet365 casino bonus processing，And help developers quickly implement bet365 casino bonus replication of an endangered bet365 casino bonus。

　The bet365 casino bonus replication proposed by this idea，It is a record of the basic appearance of a bet365 casino bonus。Conventional recording data can at most improve the speech sound effect，And once there is a bet365 casino bonus replication system for this endangered bet365 casino bonus，You can enter any text or voice in the mainstream bet365 casino bonus or reference bet365 casino bonus，Output the corresponding voice content of the endangered bet365 casino bonus through conversion。System framework based on mainstream languages ​​or reference languages，Complete bet365 casino bonus copying requires the establishment of a speech synthesis system for the target bet365 casino bonus、Speech recognition system and translation system between this bet365 casino bonus and mainstream languages。The speech output by the bet365 casino bonus copy system is not natural speech，It is a voice synthesized through speech synthesis technology。When an endangered bet365 casino bonus is really lost，People can still interact with the bet365 casino bonus through the voice copy system。

　　Speech technology expands the space for endangered bet365 casino bonus research

The framework of the bet365 casino bonus replication system with written bet365 casino bonus is roughly as follows: assuming that the mainstream bet365 casino bonus or reference bet365 casino bonus is Mandarin Chinese，The target bet365 casino bonus is an endangered bet365 casino bonus，Enter any Chinese text，The system can output the speech of this endangered bet365 casino bonus。Also enter text in any endangered bet365 casino bonus，The system can also output Chinese pronunciation。

The target bet365 casino bonus speech synthesis system is the basis of the bet365 casino bonus replication system。Speech synthesis system in bet365 casino bonus protection、The significance in bet365 casino bonus research is far greater than the practical significance。Traditional speech synthesis method requires text analysis，Convert text into speech units，Then the sequence of speech units is converted into sounds through a synthesizer。In the process of acoustic modeling，Need to define basic speech units (e.g. phonemes、Initial consonants and finals)，At the same time, it is also necessary to clarify the prosodic characteristics of these phonetic units in continuous discourse，Whether to reread、The position of the rhyme structure, etc.。In addition，Basic phonetic units also carry syntactic functions、Pragmatic function、Higher-level linguistic features such as emotional expression。If these linguistic features are marked correctly，The trained acoustic model carries richer linguistic content。When generating synthetic speech，It can produce richer expression。

At the same time，The output effect of the speech synthesis system can also test whether the input of linguistic knowledge is correct。For mainstream languages，For example, Chinese Mandarin，The basic phonetic structure has been clearly defined，Can achieve a level of automation in segment annotation，Including chapters、Dialogue、There is still room for research on other linguistic features such as emotion。For endangered languages，The basic phonetic structure has not yet been revealed，Complete analysis can be obtained using speech synthesis technology。For example, the basic phoneme definition of Sibe bet365 casino bonus has undergone a large number of phonetic changes in continuous utterances，The law of sound changes can be discovered relatively completely during the data annotation process，Isolated words cannot show these segmental changes。If you only rely on manual analysis，In order to exclude the influence of other phonemes，Usually only isolated words can be used for analysis。In the data processing process of speech synthesis，Researchers have the opportunity to perform segmental analysis on each fragment of a continuous utterance，At the same time, it also analyzes the prosody and other linguistic levels of continuous discourse as a whole，Convey linguistic knowledge to speech synthesis system through data annotation，And check the correctness of the knowledge through the output of speech synthesis。In this research mode，It will definitely promote linguistic research。

The previous discussion is limited to the processing of endangered languages ​​with written text。The technical issues involved in recording endangered languages ​​without written texts are more complicated，It’s also more difficult，But speech technology can open up more space for bet365 casino bonus research without words。

　　bet365 casino bonus resource protection and AI development complement each other

Scholars of Chinese and minority languages，You can all use this speech synthesis system to carry out your own research。We think，Except Endangered bet365 casino bonus Records，Linguists can collaborate with artificial intelligence engineers，Take the lead in the following two aspects: phoneticians and linguists use existing knowledge to accurately annotate data，Annotated content includes phonetic structure、Syntactic structure，Up to chapter information structure and supersegment content；The annotated corpus，Using intelligent speech synthesis system as a research platform，Carefully examine whether the input linguistic knowledge is correct through synthetic verification。This research method combines large-scale natural bet365 casino bonus data with artificial intelligence AI，The results can be used to study basic phonetic structures，You can also study the focus of the statement more deeply and comprehensively、Linguistic issues of prosody and discourse，Of course，It can also further improve the naturalness of synthesized speech。

When intelligent bet365 casino bonus technology has arrived，Linguists and phoneticians should take active action，Do a good job in the construction of speech and bet365 casino bonus data resources，Provide solid data support for the development of my country’s AI industry。

(Author’s unit: Institute of Linguistics, bet365 casino bonus Academy of Social Sciences)