Artificial intelligence is a strategic technology leading the future，The world's main developed countries have taken the development of artificial intelligence as the competitiveness of national competitiveness、Major strategy to maintain national security。The in -depth use of artificial intelligence technology in the field of language processing has achieved a higher level of cloudization、Automation、Technicalization、Professional、Platformization。Where，Innovative combination of artificial intelligence technology and Oracle research，with its forward -looking、Strategic、Leading and other characteristics，attracted the attention of many experts and scholars。After continuous attempts and exploration，Artificial intelligence technology in Oracle Intelligence、Intelligent school weights and other fields have obtained phased results。

But，With the continuous deepening of the discipline，The research paradigm and thinking mode of a single discipline is difficult to achieve scientific and technological innovation、The disadvantages such as solving complex problems gradually appear。These disadvantages are particularly obvious in the two huge interdisciplinary research of artificial intelligence and Oracle research，This is the selection of research goals、The design of the research method and the scientific literacy of the researcher put forward higher requirements。As an Oracle researcher in the new era，Should work hard to overcome traditional research ideas and new technologies、Questions that do not match the new goals，Breakthrough discipline barriers。At the same time，Taking Oracle research talents with artificial intelligence technology ability should be cultivated，Focus on the cultivation of the comprehensive quality of cross -talents，Promoting Oracle Study into the new era。

　　bet365 live casino games

In recent years，Thanks to the rapid development of machine learning technology and the continuous improvement of the country's unprecedented "unique learning" research on oracle bones，Use artificial intelligence technology to process Oracle information to obtain extensive attention from artificial intelligence experts and achieve a series of significant results。Wish to know，The basis for improving the accuracy of machine learning model is to provide a sufficient number of "textbooks" to the machine。In the field of artificial intelligence，These "textbooks" are called data sets，It usually requires a large number of manual labeling。Use the marked data to guide the computer to achieve a specific task target，It is a basic research paradigm for artificial intelligence technology。Specifically to the field of artificial intelligence technology to help Oracle research，The task target that experts have tried involved in the form recognition of Bet365 lotto review Oracle characters、Testing of the Oracle of the Piece、Oracle generation in cross -period、Intelligent consecration of oracle slices、School Weight and other areas。

Oracle character recognition is one of the most applied directions of artificial intelligence experts at present。Generally speaking，Glyph recognition is a category logo that is transformed into the input single image into glyphs through the computer classification model，such as "People", "Bi", "North", etc.。But，In most cases，Oracle Form recognition is limited in the selection of data sets，Because it is difficult to achieve the category of all oracle characters。In the study of Oracle Bone Formation Mission，The training data used by the researcher usually uses the data marked by yourself，The scale of its data is different。For example, the Oracle-20K of the data set mentioned by the Microsoft Asian Research Institute in the identification research contains 261 characters a total of 20039 pictures，The data set used by the Oracle recognition system of Tsinghua University contains a total of 20,420 pictures for 1621 characters。At present, the largest Oracle dataset is the OBC306 by the Key Laboratory of Oracle Information Treatment of Oracle Information Treatment of Anyang Teachers College with the OBC306，This dataset contains a total of 309551 Oracle characters composed of 306 Oracle heads。

But due to the difficulties of the tidal sample of the data set，In most cases，Oracle recognition is considered to be a small sample task。For example, Shanghai Jiaotong University tries to use 1,000 pictures to achieve 50 oracle characters for recognition，East China Normal University collected a total of 4748 portraits as a training data，Japanese Liming Museum University explores the 4257 -shaped images recorded from the "Tibetan Orthodile Bone in the Shanghai Museum" to achieve 125 oracle characters recognition。Wish to know，About 4500 Oracle I have found currently discovered，Only realize the identification of hundreds of oracle characters，This promotion effect on Oracle research is very limited。

　　The current dilemma of the research of intelligent Oracle 

The main reason for the number of oracle bone characters that can be recognized by artificial intelligence models is that it is difficult to establish a data set with sufficient data and label the correct data，Because of the labeling of the formal identification of the Oracle word and the labeling of the universal artificial intelligence model, it has essential differences in terms of difficulty。Universal artificial intelligence model requires a lot of data label，But usually do not need to mark the knowledge of professional fields，It has greater mobility in the choice of marking personnel、Flexibility。And Oracle marking requires the ability to mark the ability to recognize a certain amount of oracle characters，There are fewer people who meet this condition。The knowledge threshold of the Oracle Data Mark makes most of the Oracle Form Form recognition Studies only use part Bet365 lotto review of the fonts that are easy to obtain as a data set，This limits the scope of the application of artificial intelligence technology in Oracle research。

In view of the high threshold of the knowledge of Oracle data，The current available solution is to use a tool book for auxiliary labeling，If the prefix of the "Oracle Word" is used to divide the auxiliary marker, find more glyphs。But this method also has problems，Specific performance as: Experts when compiling a tool book，The basis for dividing the prefix is ​​not completely through the glyph，It also needs to be combined with the word of the word on the pioneer、Text and other information to judge。But，The results finally presented in the reference book，Data concentration with the use of tool books，Oracle recognition model can only be relied on only a single image，This causes the information available in the model reasoning. Compared with the information that can be used in the data set, the information available is attenuation.，This phenomenon is called information attenuation。Requirement of the model to use attenuation information to obtain the same reasoning results as professional labels violated the basic principles of information theory，It is also impossible for artificial intelligence models to complete。

Difficulty in the marking of manual data、Model can use information attenuation，Oracle -based type recognition data also exists in the problem of the data of the general recognition model very rarely involved，That is extreme unevenness of data distribution。Sample cases of universal models can be seen everywhere in life，Each category can easily find enough pictures as samples。But for Oracle，Each word appears frequently，For the low -frequencular bone characters in the original data，Artificial intelligence model is difficult to achieve the ideal recognition effect。To solve this problem，The Institute of Automation of the Chinese Academy of Sciences proposed a classification method for integrating measurement learning，Western Profe Liverpool University tried the mixed boundary decision -making method，Anyang Teachers College used a sufficient amount of handwritten oracle shape to make data amplification and other methods，Improve model recognition ability。Even so，Due to many glyphs too small，When these glyphs or other are other aliens in new research materials，Whether the smart model can correctly identify them still needs to be verified。

Of course，Data labeling problem is not only existed in the field of Oracle character recognition，This problem also exists in other areas of intelligent Oracle research。The data set used for the detection of the Oracle of the pioneer usually has only thousands of oracle slices.，Oracle generation of inter -periodic cross -period、Data available for identifying tasks，Usually there are only hundreds of pairs of glyph images，Oracle Intelligent Smart Symbol Research even has only dozens of correct coincidence results as training data。On the bet365 Play online games surface，The main obstacle to the current artificial intelligence -assisted Oracle research is the problem of data marking，But the essence of it，Its root is the discomfort of the research method of artificial intelligence and the status quo of Oracle research。Also based on Oracle characters as an example，Oracle Form recognition in the field of artificial intelligence is a classification problem of a closed data set，To solve this problem，The number of categories of the data concentration of Oracle font must be clear。For Oracle research，There are still many oracle characters that have not been completely decipher，This has led to academic differences on the issue of individual oracle types，So that the basic problems can be divided into a total of how many categories can be divided into the argument。Visible，Paradigm of artificial intelligence research paradigms in the field of Oracle research is not compatible is the root cause of the data dilemma faced by cross disciplines。

　Disciplinary cross -new exploration in data dilemma 

Oracle Form recognition as one of the representative research directions in the field of intelligent Oracle research，The data dilemma encountered in the research process is the epitome of the current interdisciplinary cross -facing problem。On the one hand，At present, the universal model trained by artificial intelligence is widely used in various living scenes，More accurate、It has huge potential for processing text and voice data more meticulous。But on the other hand，For scientific research scenes such as Oracle Research，The discomfort of the universal model also gradually appears，The discomfort of the universal model also gradually appears，But many times，Scientific research scenarios cannot provide ideal data support for smart algorithms。To release the huge potential of artificial intelligence in Oracle research，Find a breakthrough that is easy to penetrate on the two clear research paths，Researchers in two disciplines need to make greater changes on the inherent research route。

For artificial intelligence researchers，Researchers need to have solid artificial intelligence professional knowledge，Academic insights with Oracle research，Discover algorithm research and development paradigm that really helps Oracle research。With the continuous deepening of the discipline，More and more scholars start to try more suitable for artificial intelligence tasks for Oracle research，​​Among them, the more representative work includes the weight of smart oracle bone colonies and pioneering schools。Taking oracle bonus as an example，Under an artificial intelligence perspective，Oracle condensation problem is transformed into a fitting problem for image contour。The specific research path is: After using a computer vision algorithm to extract the contour curve of the oracle tablets，Select the edge computing curve similarity of part of the part of the two oracle fragments to make the oracle combination judgment。Stitching of the top curve of the fragments and the bottom curve，Zhang Chongsheng team of Henan University proposed a network model Bet365 lotto review that transformed the bone boundary line into a time series and conducts self -supervision and stitching- "Duoduo"。Chen Shanxiong's research group of Chen Shanxiong of Southwest University uses the idea of ​​measuring learning.。These two methods show good model performance，In the first 15 retrieval results, the probability of the correct combination of fragments was higher than 90%。Obviously，If these methods are put into use on a large scale，It is expected to greatly improve the research progress of the oracle condensation，Promoting the explosive growth of the results of the oracle embellishment。But as of now，Only "Duoduo" released 30 new sets of new synthesis results in 2020 to verify that Oracle experts were verified。Visible，Although artificial intelligence experts have made breakthroughs in the interdisciplinary cross，But the help of Oracle research is still very limited。

For Oracle researchers，On the basis of understanding the limitations of artificial intelligence research paradigm，Disable brought by the differences in the perspective of the subject，and actively use intelligent tools for Oracle research。The same use of oracle condensation as an example，On the one hand，Artificial Intelligence Research Statement is limited by algorithm design。Under normal circumstances，The evaluation method of the success of the algorithm design is: verify the algorithm in the labeled test data and statistical correct rate。But due to the difficulty of labeling the Oracle data，Label data that can be used for testing limited。For example，The test range of "Duoduo" is 1,000 oracle fragments images，SSF-SIAM's test scope is 480 patchable pads and 1083 unqualified pads。Whether the accuracy of stitching query in tens of thousands of oracle bone pads can meet expectations，It is still a problem to be verified。On the other hand，Under the perspective of artificial intelligence disciplines，Its research tasks are mainly designed with high accuracy with high accuracy.，Once the algorithm design is completed，can be regarded as a research goal。Under the perspective of Oracle research，Its research goals are to find more unsatisfactory results that have not been discovered before，Smart conjunction model can only be regarded as a optional Oracle research tool。Not to mention that Oracle Study as an unpopular "Study"，Experts in this field are mainly engaged in archeology、Text、Studies in the fields of history，Generally does not have artificial intelligence discipline background，How to use intelligent algorithms to engage in Oracle research，It is also a challenge for them。This leads to the potential and role of experts and scholars in the field of Oracle research，But the case that really uses this tool to solve the actual problems of Oracle research is not rich。Therefore，Cultivation with algorithm programming ability，and really use artificial intelligence as a tool for cross -disciplinary research for oracle research，It is the top priority bet365 Play online games of current talent training。

Current，Some experts and scholars have realized，In the interdisciplinary cross -background，Only realized the adaptation of researchers in research ideas，Artificial intelligence technology can provide continuous results output for Oracle research。Intelligent Synthetic Studies in the Key Laboratory of Oracle Information Treatment of Anyang Teachers College are a typical case。They have developed a smart algorithm -based computer tool while completing the development of the algorithm，and cultivated researchers who can use tools for oracle conjugate。Since 2019，This laboratory began to continue publishing conjunction results，As of now, 44 groups have been released，The correctness of the results has been verified by multiple experts。

other，Smart school heavy work has also achieved rich results output on the basis of cross -fusion of research ideas。Cooperation with the Microsoft Asian Research Institute of Oracle Research Center of the Capital Normal University，jointly developed a model for Oracle School -"Oracle Literary School Heavy Assistant Piner"，Application of deep neural network extraction global features，Determine whether it is a heavy film。The new school "Oracle Collection" (hereinafter referred to as "He") 50 cases of weight，"He" and "Oracle Collection Supplementary Compilation" (hereinafter referred to as "Complement") each weight 140 cases，"Completion" 10 cases，and supplement "Combine" and "Complement" material source Tables of 100 cases。Based on this，Oracle Information Treatment of Oracle Information Treatment of Anyang Teachers College Union and Tencent Company combined with Top Oracle Test、Piece Oracle Non -noise reduction、Oracle font matching and other artificial intelligence models，School weight through the similarity of the content of the film，New school "Complement" 25 cases of weight。Different research ideas from traditional artificial intelligence help Oracle research，Intelligent school weight research no longer use algorithm performance research as the main goal，Instead, the results of the school weight directly produced。This research method makes artificial intelligence technology truly play a role in assisting Oracle research，It plays a good demonstration and leading role for the integration and development of cross -bone research and artificial intelligence technology cross disciplines。

State's Soul，Wen Hua，Text to cast it。General Secretary Xi Jinping pointed out in the 120th anniversary of the discovery and research of Oracle，Oracle is the source of Chinese characters and the root veins of the excellent Chinese traditional culture，It is worthy of treasure、Better inheritance Development。Comprehensive use of artificial intelligence identification and other technical means，Promoting multidisciplinary cross -integration，Promoting research and achieving new breakthroughs is the meaning of further deepening the research and application of Oracle。To further promote the cross -disciplinary cross，enable artificial intelligence technology to play a direct role bet365 Play online games for the output of Oracle research results，It is required to have the basic literacy of two disciplines at the same time，When designing intelligent algorithms，Can find reasonable task goals，In Oracle Study，The ability to use intelligent algorithms for assisted research。Only this，can really make the two disciplines have a deeper collision，Push the creative conversion of ancient text research such as Oracle，Innovative development，Make new and greater contributions to promoting the development of modern Chinese civilization and the progress of human society。

　(This article is the ancient text and the development of the Chinese civilization inheritance development project "Yin Qi Wenyuan -Oracle Data Data Platform" (G2812)、"Identification and Extraction Technology of heterogeneous data in Oracle Literature" (G1806) phased results) 

(The author is the Director of the Big Data Analysis Research Office of Oracle Information Treatment of Oracle Information Treatment of the Oracle Information Treatment of the Ministry of Education、Lecturer; Member of the Chinese Academy of Social Sciences、Chief Professor of Zhengzhou University Chinese Character Civilization Research Center)