Ascend/kgrag
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

删除文件

Browse Source
This commit is contained in:
zhangkunxiang 2026-07-29 16:24:49 +08:00
parent 7b3df5b161
commit c30647c9a6
15 changed files with 0 additions and 16183 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
kg_build_0409/test/122-06A0014-B01001_发动机-维修手册.docx
View File

Binary file not shown.

BIN
kg_build_0409/test/122-06A0014-B01001_发动机-维修手册.pdf
View File

Binary file not shown.

9517
kg_build_0409/test/122-06A0014-B01001_发动机-维修手册_content_list.json
View File

File diff suppressed because one or more lines are too long

BIN
kg_build_0409/test/163-06A0015-B01001_船舶主发动机原理图.pdf
View File

Binary file not shown.

133
kg_build_0409/test/过滤文本抽取结果.json
View File

File diff suppressed because one or more lines are too long

257
kg_build_0409/test/过滤文本抽取结果_0318.json
View File

File diff suppressed because one or more lines are too long

BIN
testfiles/input/122-06A0014-B01001_发动机-维修手册测试.doc
View File

Binary file not shown.

BIN
testfiles/input/163-06A0014-B01002_螺旋桨-操作使用手册.pdf
View File

Binary file not shown.

BIN
testfiles/input/163-06A0015-B01001_发动机-操作使用手册.docx
View File

Binary file not shown.

BIN
testfiles/input/163-06A0015-B01001_发动机-维修手册.docx
View File

Binary file not shown.

BIN
testfiles/input/163-06A0016-B01003_雷达-图解目录.pdf
View File

Binary file not shown.

BIN
testfiles/input/163-06A0016-B01003_雷达-操作使用手册(1).pdf
View File

Binary file not shown.

BIN
testfiles/input/个人总结模板 V1.0.pdf
View File

Binary file not shown.

5307
testfiles/output/163-06A0016-B01003_雷达-操作使用手册(1)_content_list.json
View File

File diff suppressed because it is too large Load Diff

969
testfiles/output/个人总结模板 V1.0_content_list.json
View File

@ -1,969 +0,0 @@
[
{
"type": "text",
"text": "个人总结",
"text_level": 2,
"bbox": [
214,
325,
403,
414
],
"page_idx": 0
},
{
"type": "text",
"text": "姓名：胡德锋",
"bbox": [
112,
512,
244,
557
],
"page_idx": 0
},
{
"type": "text",
"text": "部门：大模型算法部",
"bbox": [
112,
559,
306,
603
],
"page_idx": 0
},
{
"type": "header",
"text": "紫东红途Hont",
"bbox": [
56,
61,
178,
131
],
"page_idx": 0
},
{
"type": "footer",
"text": "中 科 紫 东 红 途 （ 南 京 ） 科 技 有 限 公 司",
"bbox": [
333,
931,
677,
970
],
"page_idx": 0
},
{
"type": "image",
"img_path": "images/35cd2729a7be613b8297da0a85fead37f765207302f3ebdcc219c8523411f123.jpg",
"image_caption": [],
"image_footnote": [],
"bbox": [
45,
129,
396,
898
],
"page_idx": 1
},
{
"type": "text",
"text": "目录｜Contents",
"text_level": 2,
"bbox": [
446,
270,
791,
362
],
"page_idx": 1
},
{
"type": "text",
"text": "工作亮点与不足02-",
"bbox": [
446,
692,
691,
750
],
"page_idx": 1
},
{
"type": "text",
"text": "工作业绩总结",
"text_level": 2,
"bbox": [
40,
35,
218,
90
],
"page_idx": 2
},
{
"type": "text",
"text": "本职工作攻坚：维修助手产品研发（第一版本开发）",
"bbox": [
75,
129,
500,
172
],
"page_idx": 2
},
{
"type": "text",
"text": "重点项目支撑：西安项目",
"text_level": 2,
"bbox": [
75,
190,
284,
233
],
"page_idx": 2
},
{
"type": "text",
"text": "(1)维修助手RAG检索优化：",
"bbox": [
75,
251,
300,
292
],
"page_idx": 2
},
{
"type": "text",
"text": "1、pdf检索增加第三轮精排检索，相关性正确率90%+",
"bbox": [
112,
311,
566,
351
],
"page_idx": 2
},
{
"type": "text",
"text": "2、3d模型检索增加第三轮精排检索，相关性正确率90%+",
"bbox": [
111,
370,
595,
411
],
"page_idx": 2
},
{
"type": "text",
"text": "3、增加多级树结构零部件检索",
"bbox": [
112,
431,
372,
472
],
"page_idx": 2
},
{
"type": "text",
"text": "(2)语音模块功能开发：",
"bbox": [
76,
490,
261,
533
],
"page_idx": 2
},
{
"type": "text",
"text": "1、流式、非流式（语音转文本、热词、降噪）",
"bbox": [
113,
550,
492,
592
],
"page_idx": 2
},
{
"type": "text",
"text": "2、自定义多音色语音输出",
"bbox": [
112,
611,
335,
651
],
"page_idx": 2
},
{
"type": "text",
"text": "3、流式、非流式（说话人区分）",
"bbox": [
112,
670,
380,
712
],
"page_idx": 2
},
{
"type": "text",
"text": "(3)文件高亮、模型能力输出优化",
"bbox": [
76,
731,
348,
772
],
"page_idx": 2
},
{
"type": "text",
"text": "1、pdf高亮溯源90%+",
"bbox": [
113,
792,
304,
833
],
"page_idx": 2
},
{
"type": "text",
"text": "2、解决输出图片出错问题",
"bbox": [
112,
851,
335,
892
],
"page_idx": 2
},
{
"type": "text",
"text": "(4)对话生成内容3d零部件高亮检索",
"bbox": [
76,
911,
370,
951
],
"page_idx": 2
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
962,
94
],
"page_idx": 2
},
{
"type": "text",
"text": "工作业绩总结",
"text_level": 2,
"bbox": [
40,
35,
218,
92
],
"page_idx": 3
},
{
"type": "text",
"text": "本职工作攻坚：维修助手产品研发（第二版本开发）",
"bbox": [
73,
129,
500,
172
],
"page_idx": 3
},
{
"type": "text",
"text": "重点项目支撑：HJ项目",
"bbox": [
75,
190,
269,
233
],
"page_idx": 3
},
{
"type": "text",
"text": "(1)知识库解析：",
"bbox": [
76,
251,
205,
292
],
"page_idx": 3
},
{
"type": "text",
"text": "1、优化mineru解析速度，单日解析2000文件+",
"bbox": [
112,
311,
511,
351
],
"page_idx": 3
},
{
"type": "text",
"text": "2、基于规则（客户wx工卡、案例等信息）进行文件结构化切片（支持pdf、docx、txt、xlsx等）(2)知识图谱：",
"bbox": [
75,
370,
908,
472
],
"page_idx": 3
},
{
"type": "text",
"text": "1、抽取速度改进：规则+LLM复合抽取，单文件提取速度达到均50s",
"bbox": [
112,
490,
681,
533
],
"page_idx": 3
},
{
"type": "text",
"text": "2、图谱抽取：定义本体，根据结构树，基于章节进行划分完成粗粒度抽取，采用聚类算法，进行细粒度业务抽取",
"bbox": [
110,
550,
961,
651
],
"page_idx": 3
},
{
"type": "text",
"text": "3、图检索：支持问题的一跳问答，复杂逻辑的多跳问答",
"bbox": [
111,
670,
580,
712
],
"page_idx": 3
},
{
"type": "text",
"text": "4、图谱交互和展示：图谱初始结构树展示、点击交互展示、溯源展示等",
"bbox": [
110,
731,
710,
772
],
"page_idx": 3
},
{
"type": "text",
"text": "(3)基于agent的问答对话交互",
"bbox": [
76,
790,
325,
833
],
"page_idx": 3
},
{
"type": "text",
"text": "1、Agent对话框架搭建：维修问答、操作使用问答、百科问答、统计分析问答",
"bbox": [
112,
850,
765,
892
],
"page_idx": 3
},
{
"type": "text",
"text": "2、意图识别、多轮对话、工作流、工具化调用",
"bbox": [
111,
911,
505,
951
],
"page_idx": 3
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
962,
94
],
"page_idx": 3
},
{
"type": "text",
"text": "工作业绩总结",
"text_level": 2,
"bbox": [
40,
33,
218,
92
],
"page_idx": 4
},
{
"type": "text",
"text": "本职工作攻坚：公文助手产品研发",
"bbox": [
75,
129,
362,
172
],
"page_idx": 4
},
{
"type": "text",
"text": "重点项目支撑：教学助手",
"bbox": [
75,
190,
287,
233
],
"page_idx": 4
},
{
"type": "text",
"text": "（1）公文助手",
"bbox": [
86,
251,
204,
292
],
"page_idx": 4
},
{
"type": "text",
"text": "算法部完成算法研发情况如下：",
"bbox": [
75,
311,
332,
351
],
"page_idx": 4
},
{
"type": "text",
"text": "1）智能会议功能",
"bbox": [
92,
370,
240,
412
],
"page_idx": 4
},
{
"type": "text",
"text": "2）内容校对、内容润色",
"bbox": [
91,
429,
295,
472
],
"page_idx": 4
},
{
"type": "text",
"text": "3）教案生成、开题报告生成",
"bbox": [
91,
490,
334,
533
],
"page_idx": 4
},
{
"type": "text",
"text": "4）RAG检索生成",
"bbox": [
91,
551,
242,
590
],
"page_idx": 4
},
{
"type": "text",
"text": "5）公文撰写",
"bbox": [
91,
611,
202,
651
],
"page_idx": 4
},
{
"type": "text",
"text": "（2）智能教学助手",
"bbox": [
86,
670,
241,
712
],
"page_idx": 4
},
{
"type": "text",
"text": "1）内容校对、内容润色",
"bbox": [
92,
729,
295,
772
],
"page_idx": 4
},
{
"type": "text",
"text": "2）教案生成、开题报告生成",
"bbox": [
91,
790,
334,
831
],
"page_idx": 4
},
{
"type": "text",
"text": "3）结构化RAG检索生成",
"bbox": [
91,
850,
298,
892
],
"page_idx": 4
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
962,
94
],
"page_idx": 4
},
{
"type": "text",
"text": "工作业绩总结",
"text_level": 2,
"bbox": [
40,
33,
218,
92
],
"page_idx": 5
},
{
"type": "text",
"text": "本职工作攻坚：红途智盒",
"bbox": [
55,
164,
267,
205
],
"page_idx": 5
},
{
"type": "text",
"text": "重点项目支撑：西安项目、HJ项目、维修产品等",
"bbox": [
55,
225,
459,
266
],
"page_idx": 5
},
{
"type": "text",
"text": "(1)NPU推理：",
"bbox": [
56,
285,
170,
325
],
"page_idx": 5
},
{
"type": "text",
"text": "大模型、向量化模型、rerank模型、MinerU、sensevoice（语音转文本）、cosyvoice（语音合成）",
"bbox": [
87,
344,
916,
388
],
"page_idx": 5
},
{
"type": "text",
"text": "(2)训练：部署llama-factory等工具，可实现对于大模型、ocr模型中公式识别、表格识别等模型进行训练",
"bbox": [
56,
403,
926,
448
],
"page_idx": 5
},
{
"type": "text",
"text": "(3)维修等产品部署演示：支持hai等项目客户在各地进行产品演示，同时提供给jy等客户进行试用",
"bbox": [
56,
464,
859,
509
],
"page_idx": 5
},
{
"type": "text",
"text": "团队建设：",
"text_level": 2,
"bbox": [
56,
585,
142,
625
],
"page_idx": 5
},
{
"type": "text",
"text": "1、知识管理体系：搭建团队内部的知识git库，定期上传算法研发中过程中解决的问题难点、复用性算法能力等",
"bbox": [
54,
644,
942,
746
],
"page_idx": 5
},
{
"type": "text",
"text": "2、团队职责分配：胡德锋（10%协同，90%研发），陈睿（10%预研，90%研发），须李龙（30%预研， \n90%研发）",
"bbox": [
54,
764,
930,
866
],
"page_idx": 5
},
{
"type": "text",
"text": "3、建立周会和周会纪要，开展培训",
"bbox": [
55,
885,
354,
927
],
"page_idx": 5
},
{
"type": "text",
"text": "4、专著积累：输出2篇专利",
"bbox": [
55,
944,
289,
987
],
"page_idx": 5
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
855,
33,
962,
94
],
"page_idx": 5
},
{
"type": "text",
"text": "工作业绩总结",
"text_level": 2,
"bbox": [
40,
35,
217,
90
],
"page_idx": 6
},
{
"type": "image",
"img_path": "images/fa3d95072d436d31036d1db2ae2bdabc10cae20eaed971e4ef27844d1cc39186.jpg",
"image_caption": [],
"image_footnote": [],
"bbox": [
39,
142,
402,
779
],
"page_idx": 6
},
{
"type": "image",
"img_path": "images/307039befbb53ec074193421ebf0d8d91e86a8a85a5770bcb3f9b0e3947bee38.jpg",
"image_caption": [
"2025---->2026"
],
"image_footnote": [],
"bbox": [
469,
125,
911,
972
],
"page_idx": 6
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
962,
94
],
"page_idx": 6
},
{
"type": "text",
"text": "改进和规划",
"text_level": 2,
"bbox": [
39,
35,
187,
90
],
"page_idx": 7
},
{
"type": "text",
"text": "不足和改进",
"text_level": 2,
"bbox": [
65,
162,
165,
203
],
"page_idx": 7
},
{
"type": "text",
"text": "• 业务理解：持续加深对于业务的理解和重视程度，技术是实现功能，但应该不能光局限于技术，也要考虑到客户的角度怎么去看待问题和产品落地",
"bbox": [
70,
237,
788,
335
],
"page_idx": 7
},
{
"type": "text",
"text": "• 工作沟通：不能太过于陷入自身研发，直面问题，要及时反馈给领导，和公司同事进行配合高效配合解决问题",
"bbox": [
70,
366,
788,
461
],
"page_idx": 7
},
{
"type": "text",
"text": "• 工作经验不足：这块是自己实实在在的问题，在这里要持续和大家保持学习，要及时暴露问题，学习经验",
"bbox": [
70,
494,
788,
588
],
"page_idx": 7
},
{
"type": "text",
"text": "责任心：在后续工作中，要持续保持自己的高度责任心",
"bbox": [
71,
622,
544,
662
],
"page_idx": 7
},
{
"type": "text",
"text": "• 驻场：从现场收获到了很多数据，同时也遇见了很多问题，比如沟通问题、客户数据量大、业务变更等",
"bbox": [
70,
696,
789,
792
],
"page_idx": 7
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
962,
94
],
"page_idx": 7
},
{
"type": "text",
"text": "对公司发展的建议",
"text_level": 2,
"bbox": [
39,
35,
276,
90
],
"page_idx": 8
},
{
"type": "text",
"text": "不限范围，如：",
"bbox": [
70,
266,
196,
303
],
"page_idx": 8
},
{
"type": "text",
"text": "1、业务与战略层面",
"bbox": [
72,
305,
235,
344
],
"page_idx": 8
},
{
"type": "text",
"text": "2、管理与运营层面",
"bbox": [
70,
346,
235,
383
],
"page_idx": 8
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
962,
94
],
"page_idx": 8
},
{
"type": "text",
"text": "谢 谢",
"text_level": 2,
"bbox": [
583,
379,
747,
498
],
"page_idx": 9
},
{
"type": "text",
"text": "博弈智能操作系统为核心，聚焦决策智能领域实战化应用",
"bbox": [
509,
562,
876,
598
],
"page_idx": 9
},
{
"type": "header",
"text": "紫东红途Honto",
"bbox": [
856,
33,
963,
94
],
"page_idx": 9
}
]