{"id":1069920,"date":"2022-10-26T22:56:00","date_gmt":"2022-10-27T05:56:00","guid":{"rendered":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/?post_type=msr-blog-post&#038;p=1069920"},"modified":"2024-09-25T01:42:38","modified_gmt":"2024-09-25T08:42:38","slug":"document-ai","status":"publish","type":"msr-blog-post","link":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/articles\/document-ai\/","title":{"rendered":"\u6587\u6863\u57fa\u7840\u6a21\u578b\u5f15\u9886\u6587\u6863\u667a\u80fd\u8d70\u5411\u591a\u6a21\u6001\u5927\u4e00\u7edf"},"content":{"rendered":"\n<p>\u7f16\u8005\u6309\uff1a\u81ea2019\u5e74\u4ee5\u6765\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5728\u6587\u6863\u667a\u80fd\u9886\u57df\u8fdb\u884c\u4e86\u8bf8\u591a\u63a2\u7d22\uff0c\u5f00\u53d1\u51fa\u4e00\u7cfb\u5217\u591a\u6a21\u6001\u4efb\u52a1\u7684\u6587\u6863\u57fa\u7840\u6a21\u578b (Document Foundation Model)\uff0c\u5305\u62ec LayoutLM (v1\u3001v2\u3001v3) \u3001LayoutXLM\u3001MarkupLM \u7b49\u3002\u8fd9\u4e9b\u6a21\u578b\u5728\u8bf8\u5982\u8868\u5355\u3001\u6536\u636e\u3001\u53d1\u7968\u3001\u62a5\u544a\u7b49\u89c6\u89c9\u5bcc\u6587\u672c\u6587\u6863\u6570\u636e\u96c6\u4e0a\u90fd\u53d6\u5f97\u4e86\u4f18\u5f02\u7684\u8868\u73b0\uff0c\u83b7\u5f97\u4e86\u5b66\u672f\u754c\u548c\u4ea7\u4e1a\u754c\u7684\u5e7f\u6cdb\u8ba4\u53ef\uff0c\u5e76\u5df2\u5e94\u7528\u5728\u5305\u62ec Azure Form Recognizer\u3001AI Builder\u3001Microsoft Syntex \u7b49\u5728\u5185\u7684\u5fae\u8f6f\u4ea7\u54c1\u4e2d\uff0c\u8d4b\u80fd\u4f01\u4e1a\u548c\u673a\u6784\u7684\u6570\u5b57\u5316\u8f6c\u578b\u3002<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p>\u4f60\u662f\u5426\u66fe\u7ecf\u5728\u6838\u9500\u53d1\u7968\u4fe1\u606f\u65f6\uff0c\u88ab\u6db5\u76d6\u62ac\u5934\u3001\u5f00\u7968\u65e5\u671f\u3001\u5546\u54c1\u5185\u5bb9\u6570\u91cf\u3001\u5355\u4ef7\u3001\u91d1\u989d\u7b49\u591a\u79cd\u4fe1\u606f\u4e14\u5f62\u5f0f\u4e0d\u4e00\u7684\u53d1\u7968\u641e\u5f97\u65e0\u4ece\u4e0b\u624b\uff1f\u5904\u7406\u91cd\u8981\u7684\u5546\u4e1a\u5408\u540c\uff0c\u5c0f\u5fc3\u7ffc\u7ffc\uff0c\u751f\u6015\u5f04\u9519\u4e00\u4f4d\u5c0f\u6570\u70b9\uff0c\u9020\u6210\u4e0d\u53ef\u4f30\u91cf\u7684\u7ecf\u6d4e\u635f\u5931\uff1f\u9762\u5bf9\u6d77\u91cf\u7684\u7b80\u5386\uff0c\u52b3\u5fc3\u52b3\u529b\u4e00\u4e00\u8fc7\u76ee\uff0c\u4e0d\u60f3\u9519\u8fc7\u6bcf\u4e00\u4f4d\u4eba\u624d\uff1f\u9664\u6b64\u4e4b\u5916\uff0c\u4fdd\u9669\u4fdd\u5355\u3001\u4e1a\u52a1\u62a5\u8868\u3001\u5546\u52a1\u90ae\u4ef6\u3001\u53d1\u8d27\u8ba2\u5355\u2026\u2026 \u5546\u4e1a\u6d3b\u52a8\u4e2d\u8fd8\u6709\u5404\u79cd\u5404\u6837\u7684\u6587\u6863\u9700\u8981\u5904\u7406\u3002<\/p>\n\n\n\n<p>\u968f\u7740\u4f01\u4e1a\u6570\u5b57\u5316\u8f6c\u578b\uff0c\u5404\u79cd\u6587\u6863\u3001\u56fe\u8868\u3001\u56fe\u50cf\u5185\u5bb9\u7684\u6570\u5b57\u5316\u5df2\u7ecf\u6210\u4e3a\u4f01\u4e1a\u4e00\u9879\u91cd\u8981\u7684\u5de5\u4f5c\u3002\u4f46\u662f\u9762\u5bf9\u5927\u91cf\u8d28\u91cf\u53c2\u5dee\u4e0d\u9f50\u7684\u626b\u63cf\u6587\u4ef6\uff0c\u7248\u5f0f\u5404\u5f02\u7684\u7f51\u9875\u3001\u7535\u5b50\u6587\u6863\uff0c\u4eba\u5de5\u64cd\u4f5c\u4e0d\u4ec5\u8d39\u65f6\u8d39\u529b\u3001\u6548\u7387\u4f4e\uff0c\u8fd8\u5bb9\u6613\u51fa\u9519\uff0c\u5982\u4f55\u624d\u80fd\u9ad8\u6548\u5730\u63d0\u53d6\u3001\u6574\u7406\u548c\u5206\u6790\u6587\u6863\u4e2d\u7684\u4fe1\u606f\uff1f\u5e78\u8fd0\u7684\u662f\uff0c<a href=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/project\/document-ai\/\">\u6587\u6863\u667a\u80fd (Document AI)<\/a> \u6280\u672f\u7684\u51fa\u73b0\u5c06\u5458\u5de5\u548c\u4f01\u4e1a\u4ece\u91cd\u590d\u7e41\u9501\u7684\u6587\u6863\u6570\u5b57\u5316\u5de5\u4f5c\u4e2d\u89e3\u653e\u4e86\u51fa\u6765\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"406\" src=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-1-1024x406.png\" alt=\"\u5e38\u89c1\u7684\u5546\u4e1a\u6587\u6863\u793a\u4f8b\uff08\u4ece\u5de6\u81f3\u53f3\uff09\uff1a\u8868\u5355\u3001\u6536\u636e\u3001\u53d1\u7968\u3001\u62a5\u544a\" class=\"wp-image-1070409\" srcset=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-1-1024x406.png 1024w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-1-300x119.png 300w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-1-768x304.png 768w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-1-240x95.png 240w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-1.png 1186w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>\u5e38\u89c1\u7684\u5546\u4e1a\u6587\u6863\u793a\u4f8b\uff08\u4ece\u5de6\u81f3\u53f3\uff09\uff1a\u8868\u5355\u3001\u6536\u636e\u3001\u53d1\u7968\u3001\u62a5\u544a<\/em><\/figcaption><\/figure>\n\n\n\n<p><\/p>\n\n\n\n<p>\u6587\u6863\u667a\u80fd\u662f\u901a\u8fc7\u8ba1\u7b97\u673a\u8fdb\u884c\u81ea\u52a8\u9605\u8bfb\u3001\u7406\u89e3\u4ee5\u53ca\u5206\u6790\u5546\u4e1a\u6587\u6863\u7684\u8fc7\u7a0b\uff0c\u662f\u81ea\u7136\u8bed\u8a00\u5904\u7406 (NLP) \u548c\u8ba1\u7b97\u673a\u89c6\u89c9 (CV) \u4ea4\u53c9\u9886\u57df\u7684\u91cd\u8981\u7814\u7a76\u65b9\u5411\u3002\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u666e\u53ca\u6781\u5927\u5730\u63a8\u52a8\u4e86\u6587\u6863\u667a\u80fd\u7684\u53d1\u5c55\uff0c\u4ee5\u6587\u6863\u7248\u9762\u5206\u6790\u3001\u6587\u6863\u4fe1\u606f\u62bd\u53d6\u3001\u6587\u6863\u89c6\u89c9\u95ee\u7b54\u4ee5\u53ca\u6587\u6863\u56fe\u50cf\u5206\u7c7b\u7b49\u4e3a\u4ee3\u8868\u7684\u6587\u6863\u667a\u80fd\u4efb\u52a1\u5747\u6709\u663e\u8457\u7684\u6027\u80fd\u63d0\u5347\uff0c\u8be5\u6280\u672f\u4e5f\u5df2\u7ecf\u5728\u5e2e\u52a9\u4f01\u4e1a\u8282\u7ea6\u8fd0\u8425\u6210\u672c\u3001\u63d0\u9ad8\u5458\u5de5\u6548\u7387\u3001\u964d\u4f4e\u4eba\u4e3a\u9519\u8bef\u7b49\u65b9\u9762\u53d1\u6325\u4e86\u91cd\u8981\u4f5c\u7528\u3002<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u4ece\u6587\u672c\u5230\u591a\u6a21\u6001-\u6587\u6863\u667a\u80fd\u6a21\u578b\u9010\u6b65\u8fdb\u5316\u89e3\u9501\u65b0\u6280\u80fd\">\u4ece\u6587\u672c\u5230\u591a\u6a21\u6001\uff0c\u6587\u6863\u667a\u80fd\u6a21\u578b\u9010\u6b65\u8fdb\u5316\u89e3\u9501\u65b0\u6280\u80fd<\/h2>\n\n\n\n<p>\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5bf9\u6587\u6863\u667a\u80fd\u7684\u7cfb\u5217\u7814\u7a76\u59cb\u4e8e2019\u5e74\u3002\u5728\u5bf9\u6df1\u5ea6\u5b66\u4e60\u8fdb\u884c\u6df1\u5165\u7814\u7a76\u65f6\uff0c\u7814\u7a76\u5458\u4eec\u5e0c\u671b\u53ef\u4ee5\u4ece\u516c\u5f00\u7684\u6587\u6863\u4e2d\u62bd\u53d6\u6709\u7528\u7684\u4fe1\u606f\uff0c\u5efa\u7acb\u77e5\u8bc6\u5e93\uff0c\u4ee5\u652f\u6301\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u9884\u8bad\u7ec3\u4efb\u52a1\u3002\u7136\u800c\u6765\u81ea\u771f\u5b9e\u4e16\u754c\u7684\u6587\u6863\u5e76\u4e0d\u662f\u7ed3\u6784\u5316\u7684\u6570\u636e\uff0c\u5982\u4f55\u4ece\u6742\u4e71\u7684\u6587\u6863\u4e2d\u63d0\u53d6\u51fa\u7ed3\u6784\u5316\u7684\u6587\u672c\u4fe1\u606f\u5c31\u6210\u4e86\u7814\u7a76\u5458\u4eec\u8981\u89e3\u51b3\u7684\u7b2c\u4e00\u4e2a\u95ee\u9898\u3002<\/p>\n\n\n\n<p>\u5bf9\u6b64\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u63d0\u51fa\u4e86\u7edf\u4e00\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b UniLM\uff0c\u5b83\u65e2\u80fd\u9605\u8bfb\u6587\u6863\u53c8\u80fd\u81ea\u52a8\u751f\u6210\u5185\u5bb9\u3002UniLM \u6a21\u578b\u5728\u62bd\u8c61\u6458\u8981\u3001\u751f\u6210\u5f0f\u95ee\u7b54\u548c\u8bed\u8a00\u751f\u6210\u6570\u636e\u96c6\u7684\u62bd\u6837\u9886\u57df\u5747\u53d6\u5f97\u4e86\u4f18\u5f02\u7684\u6210\u7ee9\u3002\u540c\u65f6\uff0c\u7814\u7a76\u5458\u4eec\u8fd8\u5c06\u6a21\u578b\u4ece\u82f1\u6587\u6269\u5c55\u5230\u4e86\u66f4\u591a\u8bed\u8a00\uff0c\u63a8\u51fa\u4e86 InfoXLM \u6a21\u578b\u3002\u8fd9\u4e9b\u53ea\u9488\u5bf9\u6587\u672c\u4fe1\u606f\u5904\u7406\u7684\u6a21\u578b\u65b9\u6cd5\uff0c\u6ee1\u8db3\u4e86\u5f53\u65f6\u7814\u7a76\u5de5\u4f5c\u7684\u9700\u6c42\uff0c\u7136\u800c\u5728\u73b0\u5b9e\u573a\u666f\u4e2d\uff0c\u6587\u6863\u5185\u5bb9\u5e76\u4e0d\u662f\u53ea\u6709\u6587\u5b57\uff0c\u8fd8\u5305\u542b\u5404\u79cd\u5404\u6837\u7684\u5b57\u4f53\u3001\u989c\u8272\u3001\u4e0b\u5212\u7ebf\u7b49\u5e03\u5c40\u548c\u98ce\u683c\u4fe1\u606f\u3002<\/p>\n\n\n\n<p>2019\u5e74\u5e95\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7ed3\u5408 NLP \u548c CV \u6280\u672f\uff0c\u63a8\u51fa\u4e86\u901a\u7528\u6587\u6863\u7406\u89e3\u9884\u8bad\u7ec3\u6a21\u578b LayoutLM\uff0c\u5e76\u7b2c\u4e00\u6b21\u5728\u6587\u6863\u7ea7\u9884\u8bad\u7ec3\u4e2d\u5c06\u6587\u672c\u4e0e\u5e03\u5c40\u4fe1\u606f\u8054\u5408\u8bad\u7ec3\uff0c\u5176\u5728 IIT-CDIP Test Collection 1.0 \u6570\u636e\u96c6\u7ea6\u4e00\u5343\u4e00\u767e\u4e07\u5f20\u7684\u626b\u63cf\u6587\u6863\u56fe\u50cf\u4e0a\u8fdb\u884c\u4e86\u9884\u8bad\u7ec3\uff0c\u8be5\u6570\u636e\u96c6\u5305\u542b\u4fe1\u4ef6\u3001\u5907\u5fd8\u5f55\u3001\u7535\u5b50\u90ae\u4ef6\u3001\u8868\u683c\u3001\u7968\u636e\u7b49\u5404\u5f0f\u5404\u6837\u7684\u6587\u6863\u7c7b\u578b\u3002LayoutLM \u5728\u8868\u683c\u7406\u89e3\u3001\u7968\u636e\u7406\u89e3\u3001\u6587\u6863\u56fe\u50cf\u5206\u7c7b\u7b49\u4efb\u52a1\u7684\u5b9e\u9a8c\u4e0a\u83b7\u5f97\u4e86\u4f18\u4e8e\u5176\u5b83\u6a21\u578b\u7684\u7ed3\u679c\uff0c\u5e76\u6709\u6548\u6539\u5584\u4e86\u4ee5\u5f80\u6a21\u578b\u5728\u5177\u4f53\u573a\u666f\u4e2d\u6ca1\u6709\u5229\u7528\u5927\u89c4\u6a21\u65e0\u6807\u6ce8\u6570\u636e\uff0c\u4e14\u6a21\u578b\u96be\u4ee5\u6cdb\u5316\u7684\u95ee\u9898\u3002\u968f\u540e\u5728 LayoutLMv2 \u7248\u672c\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u5c06\u89c6\u89c9\u7279\u5f81\u4fe1\u606f\u878d\u5165\u5230\u9884\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u63d0\u5347\u4e86\u6a21\u578b\u7684\u56fe\u50cf\u7406\u89e3\u80fd\u529b\uff0c\u4ece\u800c\u5c06\u6587\u672c\u3001\u5e03\u5c40\u548c\u56fe\u50cf\u4fe1\u606f\u7edf\u4e00\u5728\u4e00\u4e2a\u6846\u67b6\u4e2d\u5171\u540c\u5efa\u6a21\u8bad\u7ec3\uff0c\u4f7f\u7528\u4e00\u4e2a\u6a21\u578b\u5c31\u80fd\u66f4\u597d\u5730\u5b66\u4e60\u5230\u4e0d\u540c\u6a21\u6001\u4e4b\u95f4\u7684\u5173\u8054\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"934\" height=\"423\" src=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-2.png\" alt=\"\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u6587\u6863\u667a\u80fd\u7cfb\u5217\u7814\u7a76\u53d1\u5c55\u5386\u7a0b\" class=\"wp-image-1070412\" srcset=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-2.png 934w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-2-300x136.png 300w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-2-768x348.png 768w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-2-240x109.png 240w\" sizes=\"auto, (max-width: 934px) 100vw, 934px\" \/><figcaption class=\"wp-element-caption\"><em>\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u6587\u6863\u667a\u80fd\u7cfb\u5217\u7814\u7a76\u53d1\u5c55\u5386\u7a0b<\/em><\/figcaption><\/figure>\n\n\n\n<p><\/p>\n\n\n\n<p>\u800c\u4e3a\u4e86\u6ee1\u8db3\u4e0d\u540c\u7528\u6237\u5bf9\u591a\u8bed\u8a00\u7684\u9700\u6c42\uff0c\u7814\u7a76\u5458\u4eec\u5728 LayoutLMv2 \u7684\u57fa\u7840\u4e0a\u8fdb\u4e00\u6b65\u63d0\u51fa\u4e86\u591a\u8bed\u8a00\u6587\u6863\u7406\u89e3\u4efb\u52a1\u7684\u591a\u6a21\u6001\u9884\u8bad\u7ec3\u6a21\u578b LayoutXLM\u3002LayoutXLM \u6a21\u578b\u4e0d\u4ec5\u4ece\u5404\u79cd\u4e0d\u540c\u8bed\u8a00\u7684\u6587\u6863\u6a21\u677f\u3001\u5e03\u5c40\u3001\u683c\u5f0f\u4e2d\u83b7\u5f97\u4e86\u6587\u672c\u548c\u89c6\u89c9\u4fe1\u53f7\uff0c\u540c\u65f6\u8fd8\u4ece\u6587\u672c\u3001\u89c6\u89c9\u548c\u8bed\u8a00\u5b66\u7684\u89d2\u5ea6\u5229\u7528\u4e86\u5c40\u90e8\u4e0d\u53d8\u7684\u7279\u6027\u3002\u9664\u4e86\u5728\u5c06\u8fd1200\u79cd\u8bed\u8a00\u4e0a\u8fdb\u884c\u4e86\u9884\u8bad\u7ec3\u5916\uff0c\u4e3a\u4e86\u66f4\u7cbe\u51c6\u5730\u8bc4\u4f30\u591a\u8bed\u8a00\u6587\u6863\u7406\u89e3\u9884\u8bad\u7ec3\u6a21\u578b\u7684\u6027\u80fd\uff0c\u7814\u7a76\u5458\u4eec\u8fd8\u521b\u5efa\u4e86\u591a\u8bed\u8a00\u6587\u6863\u7406\u89e3\u6570\u636e\u96c6 XFUND\uff0c\u5176\u6db5\u76d67\u79cd\u8bed\u8a00\uff1a\u4e2d\u6587\u3001\u65e5\u6587\u3001\u897f\u73ed\u7259\u6587\u3001\u6cd5\u6587\u3001\u610f\u5927\u5229\u6587\u3001\u5fb7\u6587\u3001\u8461\u8404\u7259\u6587\u3002<\/p>\n\n\n\n<p>\u53e6\u5916\uff0c\u5728\u4f17\u591a\u89c6\u89c9\u6548\u679c\u56fa\u5b9a\u4e0d\u53d8\u7684\u6587\u6863\u4e4b\u5916\uff0c\u73b0\u5b9e\u573a\u666f\u4e2d\u8fd8\u5b58\u5728\u5927\u91cf\u5b9e\u65f6\u6e32\u67d3\u7684\u52a8\u6001\u89c6\u89c9\u5bcc\u6587\u672c\u6587\u6863\uff0c\u5982\u57fa\u4e8e HTML \u7684\u7f51\u9875\uff0c\u6216\u57fa\u4e8e XML \u7684 Office \u6587\u4ef6\u7b49\u3002\u4e3a\u6b64\uff0c\u7814\u7a76\u5458\u4eec\u53c8\u5f00\u53d1\u4e86 MarkupLM \u6a21\u578b\uff0c\u53ef\u76f4\u63a5\u5bf9\u52a8\u6001\u6587\u6863\u7684\u6807\u8bb0\u8bed\u8a00\u6e90\u4ee3\u7801\u8fdb\u884c\u5904\u7406\uff0c\u4e0d\u9700\u8981\u4efb\u4f55\u989d\u5916\u7684\u8ba1\u7b97\u8d44\u6e90\u5373\u53ef\u6e32\u67d3\u751f\u6210\u52a8\u6001\u6587\u6863\u7684\u5b9e\u9645\u89c6\u89c9\u6548\u679c\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e MarkupLM \u663e\u8457\u4f18\u4e8e\u8fc7\u53bb\u57fa\u4e8e\u7f51\u9875\u5e03\u5c40\u7684\u65b9\u6cd5\uff0c\u5177\u6709\u5f88\u9ad8\u7684\u5b9e\u7528\u6027\u3002<\/p>\n\n\n\n<p>\u4ece\u6700\u521d\u7684\u6587\u672c\u4fe1\u606f\u5230\u5e03\u5c40\u4fe1\u606f\uff0c\u518d\u5230\u56fe\u50cf\u4fe1\u606f\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u6301\u7eed\u8fed\u4ee3\u6587\u6863\u667a\u80fd\u6280\u672f\u548c\u6a21\u578b\uff0c\u5e76\u4e8e\u4eca\u5e74\u53d1\u5e03\u4e86\u6700\u65b0\u6210\u679c LayoutLMv3\uff0c\u4ee5\u7edf\u4e00\u7684\u6587\u672c\u548c\u56fe\u50cf\u63a9\u7801\u5efa\u6a21\u76ee\u6807\u6765\u9884\u8bad\u7ec3\u591a\u6a21\u6001\u6a21\u578b\u3002LayoutLMv3 \u7684\u521b\u65b0\u4e4b\u5904\u5728\u4e8e\u63d0\u51fa\u4e86\u4e00\u4e2a\u8bcd\u5757\u5bf9\u9f50\u9884\u8bad\u7ec3\u76ee\u6807\uff0c\u901a\u8fc7\u9884\u6d4b\u4e00\u4e2a\u6587\u672c\u8bcd\u7684\u5bf9\u5e94\u56fe\u50cf\u5757\u662f\u5426\u88ab\u906e\u76d6\uff0c\u5e76\u628a\u56fe\u50cf\u7ec6\u7c92\u5ea6\u5bf9\u9f50\u5173\u7cfb\u770b\u4f5c\u4e00\u79cd\u8bed\u8a00\uff0c\u6765\u5b66\u4e60\u8de8\u6a21\u6001\u7684\u5bf9\u9f50\u5173\u7cfb\u3002\u4e0e\u6b64\u540c\u65f6\uff0cLayoutLMv3 \u9996\u6b21\u5c06\u6587\u672c\u548c\u56fe\u50cf\u540c\u65f6\u8fdb\u884c\u63a9\u7801\u9884\u6d4b\uff0c\u8fdb\u4e00\u6b65\u589e\u5f3a\u4e86\u8de8\u6a21\u6001\u5b66\u4e60\u7684\u6709\u6548\u6027\u3002\u800c\u5728\u6a21\u578b\u67b6\u6784\u4e0a\uff0cLayoutLMv3 \u4e0d\u4f9d\u8d56\u590d\u6742\u7684 CNN \u6216 Faster R-CNN \u7f51\u7edc\u6765\u8868\u5f81\u56fe\u50cf\uff0c\u76f4\u63a5\u5229\u7528\u6587\u6863\u56fe\u50cf\u7684\u56fe\u50cf\u5757\uff0c\u5927\u5927\u8282\u7701\u4e86\u53c2\u6570\u5e76\u907f\u514d\u4e86\u590d\u6742\u7684\u6587\u6863\u9884\u5904\u7406\uff0c\u8fdb\u800c\u8ba9 LayoutLMv3 \u53ef\u9002\u7528\u4e8e\u4ee5\u6587\u672c\u4e3a\u4e2d\u5fc3\u548c\u4ee5\u56fe\u50cf\u4e3a\u4e2d\u5fc3\u7684\u6587\u6863\u667a\u80fd\u4efb\u52a1\u3002<\/p>\n\n\n\n<p>\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u9996\u5e2d\u7814\u7a76\u5458\u97e6\u798f\u5982\u8868\u793a\uff1a\u201cLayout(X)LM \u7cfb\u5217\u6a21\u578b\u662f\u5927\u89c4\u6a21\u9884\u8bad\u7ec3\u57fa\u7840\u7814\u7a76\uff0c\u63a8\u8fdb\u4e0d\u540c\u4efb\u52a1\u3001\u8bed\u8a00\u548c\u6a21\u6001\u57fa\u7840\u6a21\u578b\u2018\u5927\u4e00\u7edf\u2019 (Big Convergence)\uff0c\u4ee5\u53ca\u6784\u5efa\u901a\u7528\u57fa\u7840\u6a21\u578b\u7b49\u7814\u7a76\u7684\u91cd\u8981\u7ec4\u6210\u90e8\u5206\u3002\u201d<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1020\" height=\"687\" src=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-3.png\" alt=\"LayoutLMv3 \u7684\u67b6\u6784\u548c\u9884\u8bad\u7ec3\u76ee\u6807\" class=\"wp-image-1070415\" srcset=\"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-3.png 1020w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-3-300x202.png 300w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-3-768x517.png 768w, https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-content\/uploads\/2024\/08\/document-ai-3-240x162.png 240w\" sizes=\"auto, (max-width: 1020px) 100vw, 1020px\" \/><figcaption class=\"wp-element-caption\"><em>LayoutLMv3 \u7684\u67b6\u6784\u548c\u9884\u8bad\u7ec3\u76ee\u6807<\/em><\/figcaption><\/figure>\n\n\n\n<p><\/p>\n\n\n\n<p>\u201c\u6211\u4eec\u770b\u5230\uff0c\u5728\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u7814\u7a76\u4e2d\uff0c\u5305\u62ec NLP\u3001CV \u7b49\u4e0d\u540c\u6a21\u6001\u7684\u7814\u7a76\u90fd\u5728\u5448\u73b0\u5927\u4e00\u7edf (Big Convergence) \u7684\u8d8b\u52bf\uff0c\u4e0d\u540c\u9886\u57df\u90fd\u5728\u8fdb\u884c\u7edf\u4e00\u6a21\u578b\u7684\u7814\u7a76\u3002LayoutLM \u7684\u524d\u4e24\u4e2a\u7248\u672c\u7740\u91cd\u89e3\u51b3\u7684\u662f\u8bed\u8a00\u5904\u7406\u95ee\u9898\uff0c\u800c LayoutLMv3 \u6700\u5927\u7684\u7279\u70b9\u662f\u53ef\u4ee5\u540c\u65f6\u5e94\u5bf9 NLP \u548c CV \u4e24\u79cd\u6a21\u6001\u7684\u4efb\u52a1\uff0c\u5728\u8ba1\u7b97\u89c6\u89c9\u9886\u57df\u53d6\u5f97\u4e86\u8f83\u5927\u7684\u7a81\u7834\uff0c\u201d\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u9ad8\u7ea7\u7814\u7a76\u5458\u5d14\u78ca\u8868\u793a\u3002<\/p>\n\n\n\n<p>\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7cfb\u5217\u6587\u6863\u667a\u80fd\u6a21\u578b GitHub \u94fe\u63a5\uff1a<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/github.com\/microsoft\/unilm\">https:\/\/github.com\/microsoft\/unilm<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u5f15\u9886\u4e1a\u754c\u7684\u57fa\u51c6\u6a21\u578b\">\u5f15\u9886\u4e1a\u754c\u7684\u57fa\u51c6\u6a21\u578b<\/h2>\n\n\n\n<p>\u65e0\u8bba\u5728\u5927\u89c4\u6a21\u65e0\u6807\u6ce8\u6570\u636e\u7684\u4f7f\u7528\u4e0a\uff0c\u8fd8\u662f\u5bf9\u6587\u672c\u3001\u56fe\u7247\u3001\u591a\u6a21\u6001\u3001\u591a\u7248\u5f0f\u3001\u591a\u8bed\u8a00\u7684\u5bcc\u6587\u672c\u5185\u5bb9\u7684\u7406\u89e3\u4e0a\uff0cLayoutLM \u90fd\u6781\u5177\u9886\u5148\u6027\uff0c\u5c24\u5176\u662f LayoutLMv3 \u66f4\u9ad8\u7684\u901a\u7528\u6027\u548c\u4f18\u8d8a\u6027\uff0c\u4f7f\u4e4b\u6210\u4e3a\u4e1a\u754c\u7814\u7a76\u7684\u57fa\u51c6\u6a21\u578b\uff0c\u4f17\u591a\u5934\u90e8\u4f01\u4e1a\u548c\u673a\u5668\u4eba\u81ea\u52a8\u5316 (RPA) \u9886\u57df\u4f01\u4e1a\u7684\u6587\u6863\u667a\u80fd\u4ea7\u54c1\u4e2d\u90fd\u6709 LayoutLM \u7684\u8eab\u5f71\u3002<\/p>\n\n\n\n<p>\u201c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u4e0d\u4ec5\u5728\u57fa\u7840\u6a21\u578b\u548c\u57fa\u51c6\u6570\u636e\u96c6\u7684\u521b\u65b0\u4e0a\u53d6\u5f97\u4e86\u8bf8\u591a\u6210\u679c\uff0c\u6211\u4eec\u7684\u6a21\u578b\u8fd8\u652f\u6491\u4e86\u5f88\u591a\u4e0a\u5c42\u5e94\u7528\uff0c\u8ba9\u7528\u6237\u53ea\u7528\u4e00\u4e2a\u57fa\u7840\u6a21\u578b\u5c31\u80fd\u5b8c\u6210\u591a\u9879\u4efb\u52a1\u7684\u8bad\u7ec3\u3002\u5f88\u591a\u5b66\u672f\u754c\u548c\u4ea7\u4e1a\u754c\u7684\u540c\u4ec1\u90fd\u5728\u7528 LayoutLM \u6216 LayoutXLM \u8fdb\u884c\u66f4\u591a\u6709\u610f\u4e49\u7684\u63a2\u7d22\uff0c\u4fc3\u8fdb\u6587\u6863\u667a\u80fd\u9886\u57df\u5411\u524d\u53d1\u5c55\u3002\u201d\u5d14\u78ca\u8bf4\u3002<\/p>\n\n\n\n<p>\u5fae\u8f6f\u81ea\u8eab\u7684\u4ea7\u54c1\u66f4\u662f\u4e00\u9a6c\u5f53\u5148\uff0c\u76ee\u524d\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5728\u6587\u6863\u667a\u80fd\u9886\u57df\u7684\u4e00\u7cfb\u5217\u6a21\u578b\u5df2\u5e94\u7528\u5230\u8bf8\u591a\u5fae\u8f6f\u7684\u76f8\u5173\u4ea7\u54c1\u4e2d\uff0c\u5305\u62ec Azure Form Recognizer\u3001AI Builder\u3001Microsoft Syntex \u7b49\u3002\u5fae\u8f6f Azure AI \u5408\u4f19\u4eba\u7814\u53d1\u7ecf\u7406\u5f20\u5bdf\u8868\u793a\uff0c\u201c\u6211\u4eec\u5f88\u9ad8\u5174\u80fd\u548c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u8fd9\u4e9b\u9876\u5c16\u7684\u7814\u7a76\u5458\u4eec\u5408\u4f5c\u3002\u6587\u6863\u667a\u80fd\u7684\u57fa\u7840\u6a21\u578b\u6781\u5927\u5730\u63d0\u9ad8\u4e86\u6211\u4eec\u5728\u8be5\u9886\u57df\u5e94\u7528\u3001\u5f00\u53d1\u7684\u6548\u7387\uff0c\u540c\u65f6\uff0c\u4e5f\u5bf9\u6587\u6863\u667a\u80fd\u7684\u666e\u53ca\u6709\u7740\u79ef\u6781\u7684\u63a8\u52a8\u4f5c\u7528\u3002\u6211\u4eec\u671f\u5f85\u672a\u6765\u5728\u8fd9\u4e00\u9886\u57df\u6709\u66f4\u591a\u6fc0\u52a8\u4eba\u5fc3\u7684\u8fdb\u5c55\u3002\u201d<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"\u6587\u6863\u667a\u80fd\u7684\u4e0b\u4e00\u6b65-\u5927\u89c4\u6a21\u7684\u7edf\u4e00\u6846\u67b6\">\u6587\u6863\u667a\u80fd\u7684\u4e0b\u4e00\u6b65\uff1a\u5927\u89c4\u6a21\u7684\u7edf\u4e00\u6846\u67b6<\/h2>\n\n\n\n<p>\u968f\u7740\u6280\u672f\u9010\u6b65\u8d70\u5411\u6210\u719f\uff0c\u6587\u6863\u667a\u80fd\u5df2\u5728\u91d1\u878d\u3001\u533b\u7597\u3001\u80fd\u6e90\u3001\u653f\u52a1\u3001\u7269\u6d41\u7b49\u4e0d\u540c\u884c\u4e1a\u5b9e\u73b0\u4e86\u4e0d\u540c\u7c7b\u578b\u7684\u5e94\u7528\u3002\u4f8b\u5982\uff0c\u5728\u91d1\u878d\u9886\u57df\u53ef\u5b9e\u73b0\u8d22\u62a5\u5206\u6790\u548c\u667a\u80fd\u51b3\u7b56\u5206\u6790\uff1b\u5728\u533b\u7597\u9886\u57df\u63a8\u52a8\u75c5\u4f8b\u6570\u5b57\u5316\uff0c\u5206\u6790\u533b\u5b66\u6587\u732e\u548c\u75c5\u4f8b\u5173\u8054\u6027\uff0c\u53d1\u73b0\u6f5c\u5728\u6cbb\u7597\u65b9\u6848\uff1b\u5728\u8d22\u52a1\u9886\u57df\u5b9e\u73b0\u53d1\u7968\u548c\u8ba2\u5355\u7684\u81ea\u52a8\u5316\u4fe1\u606f\u63d0\u53d6\uff0c\u8282\u7701\u5927\u91cf\u4eba\u5de5\u5904\u7406\u7684\u65f6\u95f4\u6210\u672c\u3002<\/p>\n\n\n\n<p>\u4f46\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5e76\u4e0d\u4f1a\u6b62\u6b65\u4e8e\u6b64\uff0c\u5d14\u78ca\u8868\u793a\uff0c\u4e0b\u4e00\u6b65\u7814\u7a76\u5458\u4eec\u5c06\u4ece\u63d0\u5347\u6a21\u578b\u89c4\u6a21\u3001\u6269\u5927\u8bad\u7ec3\u6570\u636e\u89c4\u6a21\u548c\u7edf\u4e00\u6846\u67b6\u4e09\u4e2a\u65b9\u9762\u7740\u624b\uff0c\u8fdb\u4e00\u6b65\u63a8\u8fdb\u6587\u6863\u667a\u80fd\u7684\u57fa\u7840\u7814\u7a76\u3002\u201cNLP \u9886\u57df\u7684 GPT-3 \u8bc1\u660e\u4e86\u8d85\u5927\u6a21\u578b\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u6a21\u578b\u7684\u6027\u80fd\uff0c\u4e0e\u6b64\u540c\u65f6\u5f53\u524d\u6587\u6863\u667a\u80fd\u6a21\u578b\u8bad\u7ec3\u4f7f\u7528\u7684\u6570\u636e\u8fd8\u4e0d\u53ca\u4e92\u8054\u7f51\u6570\u636e\u7684\u5341\u5206\u4e4b\u4e00\uff0c\u8fd8\u6709\u5f88\u5927\u7684\u63d0\u5347\u7a7a\u95f4\u3002\u6211\u4eec\u5e0c\u671b\u4e0d\u65ad\u6269\u5c55\u6570\u636e\u548c\u6a21\u578b\u89c4\u6a21\uff0c\u5e76\u5b9e\u73b0\u4e00\u4e2a\u7edf\u4e00\u7684\u6846\u67b6\uff0c\u628a\u6574\u4e2a\u6587\u6863\u667a\u80fd\u5305\u542b\u7684\u8981\u7d20\u90fd\u7edf\u4e00\u8d77\u6765\uff0c\u8fd9\u6b63\u662f\u6211\u4eec\u5f53\u524d\u548c\u672a\u6765\u7814\u7a76\u5de5\u4f5c\u7684\u91cd\u70b9\u3002\u201d<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u7f16\u8005\u6309\uff1a\u81ea2019\u5e74\u4ee5\u6765\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5728\u6587\u6863\u667a\u80fd\u9886\u57df\u8fdb\u884c\u4e86\u8bf8\u591a\u63a2\u7d22\uff0c\u5f00\u53d1\u51fa\u4e00\u7cfb\u5217\u591a\u6a21\u6001\u4efb\u52a1\u7684\u6587\u6863\u57fa\u7840\u6a21\u578b (Document Foundation Model)\uff0c\u5305\u62ec LayoutLM (v1\u3001v2\u3001v3) \u3001LayoutXLM\u3001MarkupLM \u7b49\u3002\u8fd9\u4e9b\u6a21\u578b\u5728\u8bf8\u5982\u8868\u5355\u3001\u6536\u636e\u3001\u53d1\u7968\u3001\u62a5\u544a\u7b49\u89c6\u89c9\u5bcc\u6587\u672c\u6587\u6863\u6570\u636e\u96c6\u4e0a\u90fd\u53d6\u5f97\u4e86\u4f18\u5f02\u7684\u8868\u73b0\uff0c\u83b7\u5f97\u4e86\u5b66\u672f\u754c\u548c\u4ea7\u4e1a\u754c\u7684\u5e7f\u6cdb\u8ba4\u53ef\uff0c\u5e76\u5df2\u5e94\u7528\u5728\u5305\u62ec Azure Form Recognizer\u3001AI Builder\u3001Microsoft Syntex \u7b49\u5728\u5185\u7684\u5fae\u8f6f\u4ea7\u54c1\u4e2d\uff0c\u8d4b\u80fd\u4f01\u4e1a\u548c\u673a\u6784\u7684\u6570\u5b57\u5316\u8f6c\u578b\u3002 \u4f60\u662f\u5426\u66fe\u7ecf\u5728\u6838\u9500\u53d1\u7968\u4fe1\u606f\u65f6\uff0c\u88ab\u6db5\u76d6\u62ac\u5934\u3001\u5f00\u7968\u65e5\u671f\u3001\u5546\u54c1\u5185\u5bb9\u6570\u91cf\u3001\u5355\u4ef7\u3001\u91d1\u989d\u7b49\u591a\u79cd\u4fe1\u606f\u4e14\u5f62\u5f0f\u4e0d\u4e00\u7684\u53d1\u7968\u641e\u5f97\u65e0\u4ece\u4e0b\u624b\uff1f\u5904\u7406\u91cd\u8981\u7684\u5546\u4e1a\u5408\u540c\uff0c\u5c0f\u5fc3\u7ffc\u7ffc\uff0c\u751f\u6015\u5f04\u9519\u4e00\u4f4d\u5c0f\u6570\u70b9\uff0c\u9020\u6210\u4e0d\u53ef\u4f30\u91cf\u7684\u7ecf\u6d4e\u635f\u5931\uff1f\u9762\u5bf9\u6d77\u91cf\u7684\u7b80\u5386\uff0c\u52b3\u5fc3\u52b3\u529b\u4e00\u4e00\u8fc7\u76ee\uff0c\u4e0d\u60f3\u9519\u8fc7\u6bcf\u4e00\u4f4d\u4eba\u624d\uff1f\u9664\u6b64\u4e4b\u5916\uff0c\u4fdd\u9669\u4fdd\u5355\u3001\u4e1a\u52a1\u62a5\u8868\u3001\u5546\u52a1\u90ae\u4ef6\u3001\u53d1\u8d27\u8ba2\u5355\u2026\u2026 \u5546\u4e1a\u6d3b\u52a8\u4e2d\u8fd8\u6709\u5404\u79cd\u5404\u6837\u7684\u6587\u6863\u9700\u8981\u5904\u7406\u3002 \u968f\u7740\u4f01\u4e1a\u6570\u5b57\u5316\u8f6c\u578b\uff0c\u5404\u79cd\u6587\u6863\u3001\u56fe\u8868\u3001\u56fe\u50cf\u5185\u5bb9\u7684\u6570\u5b57\u5316\u5df2\u7ecf\u6210\u4e3a\u4f01\u4e1a\u4e00\u9879\u91cd\u8981\u7684\u5de5\u4f5c\u3002\u4f46\u662f\u9762\u5bf9\u5927\u91cf\u8d28\u91cf\u53c2\u5dee\u4e0d\u9f50\u7684\u626b\u63cf\u6587\u4ef6\uff0c\u7248\u5f0f\u5404\u5f02\u7684\u7f51\u9875\u3001\u7535\u5b50\u6587\u6863\uff0c\u4eba\u5de5\u64cd\u4f5c\u4e0d\u4ec5\u8d39\u65f6\u8d39\u529b\u3001\u6548\u7387\u4f4e\uff0c\u8fd8\u5bb9\u6613\u51fa\u9519\uff0c\u5982\u4f55\u624d\u80fd\u9ad8\u6548\u5730\u63d0\u53d6\u3001\u6574\u7406\u548c\u5206\u6790\u6587\u6863\u4e2d\u7684\u4fe1\u606f\uff1f\u5e78\u8fd0\u7684\u662f\uff0c\u6587\u6863\u667a\u80fd (Document AI) \u6280\u672f\u7684\u51fa\u73b0\u5c06\u5458\u5de5\u548c\u4f01\u4e1a\u4ece\u91cd\u590d\u7e41\u9501\u7684\u6587\u6863\u6570\u5b57\u5316\u5de5\u4f5c\u4e2d\u89e3\u653e\u4e86\u51fa\u6765\u3002 \u6587\u6863\u667a\u80fd\u662f\u901a\u8fc7\u8ba1\u7b97\u673a\u8fdb\u884c\u81ea\u52a8\u9605\u8bfb\u3001\u7406\u89e3\u4ee5\u53ca\u5206\u6790\u5546\u4e1a\u6587\u6863\u7684\u8fc7\u7a0b\uff0c\u662f\u81ea\u7136\u8bed\u8a00\u5904\u7406 (NLP) \u548c\u8ba1\u7b97\u673a\u89c6\u89c9 (CV) \u4ea4\u53c9\u9886\u57df\u7684\u91cd\u8981\u7814\u7a76\u65b9\u5411\u3002\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u7684\u666e\u53ca\u6781\u5927\u5730\u63a8\u52a8\u4e86\u6587\u6863\u667a\u80fd\u7684\u53d1\u5c55\uff0c\u4ee5\u6587\u6863\u7248\u9762\u5206\u6790\u3001\u6587\u6863\u4fe1\u606f\u62bd\u53d6\u3001\u6587\u6863\u89c6\u89c9\u95ee\u7b54\u4ee5\u53ca\u6587\u6863\u56fe\u50cf\u5206\u7c7b\u7b49\u4e3a\u4ee3\u8868\u7684\u6587\u6863\u667a\u80fd\u4efb\u52a1\u5747\u6709\u663e\u8457\u7684\u6027\u80fd\u63d0\u5347\uff0c\u8be5\u6280\u672f\u4e5f\u5df2\u7ecf\u5728\u5e2e\u52a9\u4f01\u4e1a\u8282\u7ea6\u8fd0\u8425\u6210\u672c\u3001\u63d0\u9ad8\u5458\u5de5\u6548\u7387\u3001\u964d\u4f4e\u4eba\u4e3a\u9519\u8bef\u7b49\u65b9\u9762\u53d1\u6325\u4e86\u91cd\u8981\u4f5c\u7528\u3002 \u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u5bf9\u6587\u6863\u667a\u80fd\u7684\u7cfb\u5217\u7814\u7a76\u59cb\u4e8e2019\u5e74\u3002\u5728\u5bf9\u6df1\u5ea6\u5b66\u4e60\u8fdb\u884c\u6df1\u5165\u7814\u7a76\u65f6\uff0c\u7814\u7a76\u5458\u4eec\u5e0c\u671b\u53ef\u4ee5\u4ece\u516c\u5f00\u7684\u6587\u6863\u4e2d\u62bd\u53d6\u6709\u7528\u7684\u4fe1\u606f\uff0c\u5efa\u7acb\u77e5\u8bc6\u5e93\uff0c\u4ee5\u652f\u6301\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u7684\u9884\u8bad\u7ec3\u4efb\u52a1\u3002\u7136\u800c\u6765\u81ea\u771f\u5b9e\u4e16\u754c\u7684\u6587\u6863\u5e76\u4e0d\u662f\u7ed3\u6784\u5316\u7684\u6570\u636e\uff0c\u5982\u4f55\u4ece\u6742\u4e71\u7684\u6587\u6863\u4e2d\u63d0\u53d6\u51fa\u7ed3\u6784\u5316\u7684\u6587\u672c\u4fe1\u606f\u5c31\u6210\u4e86\u7814\u7a76\u5458\u4eec\u8981\u89e3\u51b3\u7684\u7b2c\u4e00\u4e2a\u95ee\u9898\u3002 \u5bf9\u6b64\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u63d0\u51fa\u4e86\u7edf\u4e00\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b UniLM\uff0c\u5b83\u65e2\u80fd\u9605\u8bfb\u6587\u6863\u53c8\u80fd\u81ea\u52a8\u751f\u6210\u5185\u5bb9\u3002UniLM \u6a21\u578b\u5728\u62bd\u8c61\u6458\u8981\u3001\u751f\u6210\u5f0f\u95ee\u7b54\u548c\u8bed\u8a00\u751f\u6210\u6570\u636e\u96c6\u7684\u62bd\u6837\u9886\u57df\u5747\u53d6\u5f97\u4e86\u4f18\u5f02\u7684\u6210\u7ee9\u3002\u540c\u65f6\uff0c\u7814\u7a76\u5458\u4eec\u8fd8\u5c06\u6a21\u578b\u4ece\u82f1\u6587\u6269\u5c55\u5230\u4e86\u66f4\u591a\u8bed\u8a00\uff0c\u63a8\u51fa\u4e86 InfoXLM \u6a21\u578b\u3002\u8fd9\u4e9b\u53ea\u9488\u5bf9\u6587\u672c\u4fe1\u606f\u5904\u7406\u7684\u6a21\u578b\u65b9\u6cd5\uff0c\u6ee1\u8db3\u4e86\u5f53\u65f6\u7814\u7a76\u5de5\u4f5c\u7684\u9700\u6c42\uff0c\u7136\u800c\u5728\u73b0\u5b9e\u573a\u666f\u4e2d\uff0c\u6587\u6863\u5185\u5bb9\u5e76\u4e0d\u662f\u53ea\u6709\u6587\u5b57\uff0c\u8fd8\u5305\u542b\u5404\u79cd\u5404\u6837\u7684\u5b57\u4f53\u3001\u989c\u8272\u3001\u4e0b\u5212\u7ebf\u7b49\u5e03\u5c40\u548c\u98ce\u683c\u4fe1\u606f\u3002 2019\u5e74\u5e95\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7ed3\u5408 NLP \u548c CV \u6280\u672f\uff0c\u63a8\u51fa\u4e86\u901a\u7528\u6587\u6863\u7406\u89e3\u9884\u8bad\u7ec3\u6a21\u578b LayoutLM\uff0c\u5e76\u7b2c\u4e00\u6b21\u5728\u6587\u6863\u7ea7\u9884\u8bad\u7ec3\u4e2d\u5c06\u6587\u672c\u4e0e\u5e03\u5c40\u4fe1\u606f\u8054\u5408\u8bad\u7ec3\uff0c\u5176\u5728 IIT-CDIP Test Collection 1.0 \u6570\u636e\u96c6\u7ea6\u4e00\u5343\u4e00\u767e\u4e07\u5f20\u7684\u626b\u63cf\u6587\u6863\u56fe\u50cf\u4e0a\u8fdb\u884c\u4e86\u9884\u8bad\u7ec3\uff0c\u8be5\u6570\u636e\u96c6\u5305\u542b\u4fe1\u4ef6\u3001\u5907\u5fd8\u5f55\u3001\u7535\u5b50\u90ae\u4ef6\u3001\u8868\u683c\u3001\u7968\u636e\u7b49\u5404\u5f0f\u5404\u6837\u7684\u6587\u6863\u7c7b\u578b\u3002LayoutLM \u5728\u8868\u683c\u7406\u89e3\u3001\u7968\u636e\u7406\u89e3\u3001\u6587\u6863\u56fe\u50cf\u5206\u7c7b\u7b49\u4efb\u52a1\u7684\u5b9e\u9a8c\u4e0a\u83b7\u5f97\u4e86\u4f18\u4e8e\u5176\u5b83\u6a21\u578b\u7684\u7ed3\u679c\uff0c\u5e76\u6709\u6548\u6539\u5584\u4e86\u4ee5\u5f80\u6a21\u578b\u5728\u5177\u4f53\u573a\u666f\u4e2d\u6ca1\u6709\u5229\u7528\u5927\u89c4\u6a21\u65e0\u6807\u6ce8\u6570\u636e\uff0c\u4e14\u6a21\u578b\u96be\u4ee5\u6cdb\u5316\u7684\u95ee\u9898\u3002\u968f\u540e\u5728 LayoutLMv2 \u7248\u672c\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u5c06\u89c6\u89c9\u7279\u5f81\u4fe1\u606f\u878d\u5165\u5230\u9884\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u63d0\u5347\u4e86\u6a21\u578b\u7684\u56fe\u50cf\u7406\u89e3\u80fd\u529b\uff0c\u4ece\u800c\u5c06\u6587\u672c\u3001\u5e03\u5c40\u548c\u56fe\u50cf\u4fe1\u606f\u7edf\u4e00\u5728\u4e00\u4e2a\u6846\u67b6\u4e2d\u5171\u540c\u5efa\u6a21\u8bad\u7ec3\uff0c\u4f7f\u7528\u4e00\u4e2a\u6a21\u578b\u5c31\u80fd\u66f4\u597d\u5730\u5b66\u4e60\u5230\u4e0d\u540c\u6a21\u6001\u4e4b\u95f4\u7684\u5173\u8054\u3002 \u800c\u4e3a\u4e86\u6ee1\u8db3\u4e0d\u540c\u7528\u6237\u5bf9\u591a\u8bed\u8a00\u7684\u9700\u6c42\uff0c\u7814\u7a76\u5458\u4eec\u5728 LayoutLMv2 \u7684\u57fa\u7840\u4e0a\u8fdb\u4e00\u6b65\u63d0\u51fa\u4e86\u591a\u8bed\u8a00\u6587\u6863\u7406\u89e3\u4efb\u52a1\u7684\u591a\u6a21\u6001\u9884\u8bad\u7ec3\u6a21\u578b LayoutXLM\u3002LayoutXLM \u6a21\u578b\u4e0d\u4ec5\u4ece\u5404\u79cd\u4e0d\u540c\u8bed\u8a00\u7684\u6587\u6863\u6a21\u677f\u3001\u5e03\u5c40\u3001\u683c\u5f0f\u4e2d\u83b7\u5f97\u4e86\u6587\u672c\u548c\u89c6\u89c9\u4fe1\u53f7\uff0c\u540c\u65f6\u8fd8\u4ece\u6587\u672c\u3001\u89c6\u89c9\u548c\u8bed\u8a00\u5b66\u7684\u89d2\u5ea6\u5229\u7528\u4e86\u5c40\u90e8\u4e0d\u53d8\u7684\u7279\u6027\u3002\u9664\u4e86\u5728\u5c06\u8fd1200\u79cd\u8bed\u8a00\u4e0a\u8fdb\u884c\u4e86\u9884\u8bad\u7ec3\u5916\uff0c\u4e3a\u4e86\u66f4\u7cbe\u51c6\u5730\u8bc4\u4f30\u591a\u8bed\u8a00\u6587\u6863\u7406\u89e3\u9884\u8bad\u7ec3\u6a21\u578b\u7684\u6027\u80fd\uff0c\u7814\u7a76\u5458\u4eec\u8fd8\u521b\u5efa\u4e86\u591a\u8bed\u8a00\u6587\u6863\u7406\u89e3\u6570\u636e\u96c6 XFUND\uff0c\u5176\u6db5\u76d67\u79cd\u8bed\u8a00\uff1a\u4e2d\u6587\u3001\u65e5\u6587\u3001\u897f\u73ed\u7259\u6587\u3001\u6cd5\u6587\u3001\u610f\u5927\u5229\u6587\u3001\u5fb7\u6587\u3001\u8461\u8404\u7259\u6587\u3002 \u53e6\u5916\uff0c\u5728\u4f17\u591a\u89c6\u89c9\u6548\u679c\u56fa\u5b9a\u4e0d\u53d8\u7684\u6587\u6863\u4e4b\u5916\uff0c\u73b0\u5b9e\u573a\u666f\u4e2d\u8fd8\u5b58\u5728\u5927\u91cf\u5b9e\u65f6\u6e32\u67d3\u7684\u52a8\u6001\u89c6\u89c9\u5bcc\u6587\u672c\u6587\u6863\uff0c\u5982\u57fa\u4e8e HTML \u7684\u7f51\u9875\uff0c\u6216\u57fa\u4e8e XML [&hellip;]<\/p>\n","protected":false},"author":34512,"featured_media":1070406,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-content-parent":1012650,"msr_hide_image_in_river":0,"footnotes":""},"research-area":[],"msr-locale":[268881],"msr-post-option":[],"class_list":["post-1069920","msr-blog-post","type-msr-blog-post","status-publish","has-post-thumbnail","hentry","msr-locale-zh_cn"],"msr_assoc_parent":{"id":1012650,"type":"lab"},"_links":{"self":[{"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1069920","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post"}],"about":[{"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-blog-post"}],"author":[{"embeddable":true,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/users\/34512"}],"version-history":[{"count":3,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1069920\/revisions"}],"predecessor-version":[{"id":1070421,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1069920\/revisions\/1070421"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/media\/1070406"}],"wp:attachment":[{"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1069920"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1069920"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1069920"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/newed.any0.dpdns.org\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1069920"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}