You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: articles/ai-services/document-intelligence/concept-accuracy-confidence.md
-14Lines changed: 0 additions & 14 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -71,44 +71,30 @@ With the addition of table, row and cell confidence with the ```2024-02-29-previ
71
71
72
72
**A:** Yes. The different levels of table confidence (cell, row, and table) are meant to capture the correctness of a prediction at that specific level. A correctly predicted cell that belongs to a row with other possible misses would have high cell confidence, but the row's confidence should be low. Similarly, a correct row in a table with challenges with other rows would have high row confidence whereas the table's overall confidence would be low.
73
73
74
-
---
75
-
76
74
**Q:** What is the expected confidence score when cells are merged? Since a merge results in the number of columns identified to change, how are scores affected?<br>
77
75
78
76
**A:** Regardless of the type of table, the expectation for merged cells is that they should have lower confidence values. Furthermore, the cell that is missing (because it was merged with an adjacent cell) should have `NULL` value with lower confidence as well. How much lower these values might be depends on the training dataset, the general trend of both merged and missing cell having lower scores should hold.
79
77
80
-
---
81
-
82
78
**Q:** What is the confidence score when a value is optional? Should you expect a cell with a ``NULL`` value and high confidence score if the value is missing?<br>
83
79
84
80
**A:** If your training dataset is representative of the optionality of cells, it helps the model know how often a value tends to appear in the training set, and thus what to expect during inference. This feature is used when computing the confidence of either a prediction or of making no prediction at all (`NULL`). You should expect an empty field with high confidence for missing values that are mostly empty in the training set too.
85
81
86
-
---
87
-
88
82
**Q:** How are confidence scores affected if a field is optional and not present or missed? Is the expectation that the confidence score answers that question?<br>
89
83
90
84
**A:** When a value is missing from a row, the cell has a `NULL` value and confidence assigned. A high confidence score here should mean that the model prediction (of there not being a value) is more likely to be correct. In contrast, a low score should signal more uncertainty from the model (and thus the possibility of an error, like the value being missed).
91
85
92
-
---
93
-
94
86
**Q:** What should be the expectation for cell confidence and row confidence when extracting a multi-page table with a row split across pages?<br>
95
87
96
88
**A:** Expect the cell confidence to be high and row confidence to be potentially lower than rows that aren't split. The proportion of split rows in the training data set can affect the confidence score. In general, a split row looks different than the other rows in the table (thus, the model is less certain that it's correct).
97
89
98
-
---
99
-
100
90
**Q:** For cross-page tables with rows that cleanly end and start at the page boundaries, is it correct to assume that confidence scores are consistent across pages?
101
91
102
92
**A:** Yes. Since rows look similar in shape and contents, regardless of where they are in the document (or in which page), their respective confidence scores should be consistent.
103
93
104
-
---
105
-
106
94
**Q:** What is the best way to utilize the new confidence scores?<br>
107
95
108
96
**A:** Look at all levels of table confidence starting in a top-to-bottom approach: begin by checking a table's confidence as a whole, then drill down to the row level and look at individual rows, finally look at cell-level confidences. Depending on the type of table, there are a couple of things of note:
109
97
110
-
---
111
-
112
98
For **fixed tables**, cell-level confidence already captures quite a bit of information on the correctness of things. This means that simply going over each cell and looking at its confidence can be enough to help determine the quality of the prediction.
113
99
For **dynamic tables**, the levels are meant to build on top of each other, so the top-to-bottom approach is more important.
Copy file name to clipboardExpand all lines: articles/ai-services/document-intelligence/concept-add-on-capabilities.md
+50-31Lines changed: 50 additions & 31 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -54,7 +54,7 @@ Document Intelligence supports more sophisticated and modular analysis capabilit
54
54
>
55
55
> Not all add-on capabilities are supported by all models. For more information, *see*[model data extraction](concept-model-overview.md#analysis-features).
56
56
57
-
The following add-on capability isavailable for`2024-02-29-preview`, `2024-02-29-preview`, and later releases:
57
+
The following add-on capabilities are available for`2024-02-29-preview`, `2024-02-29-preview`, and later releases:
58
58
59
59
*[`keyValuePairs`](#key-value-pairs)
60
60
@@ -81,20 +81,24 @@ Add-On* - Query fields are priced differently than the other add-on features. Se
81
81
82
82
## High resolution extraction
83
83
84
-
The task of recognizing small text from large-size documents, like engineering drawings, is a challenge. Often the text is mixed with other graphical elements and has varying fonts, sizes and orientations. Moreover, the text can be broken into separate parts or connected with other symbols. Document Intelligence now supports extracting content from these types of documents with the `ocr.highResolution` capability. You get improved quality of content extraction from A1/A2/A3 documents by enabling this add-on capability.
84
+
The task of recognizing small text from large-size documents, like engineering drawings, is a challenge. Often the text is mixed with other graphical elements and has varying fonts, sizes, and orientations. Moreover, the text can be broken into separate parts or connected with other symbols. Document Intelligence now supports extracting content from these types of documents with the `ocr.highResolution` capability. You get improved quality of content extraction from A1/A2/A3 documents by enabling this add-on capability.
@@ -298,7 +318,7 @@ Query fields are an add-on capability to extend the schema extracted from any pr
298
318
299
319
> [!NOTE]
300
320
>
301
-
> Document Intelligence Studio query field extraction is currently available with the Layout and Prebuilt models `2024-02-29-preview``2023-10-31-preview` API and later releases except for the ```us.tax.*``` models (W2, 1098s and 1099s models).
321
+
> Document Intelligence Studio query field extraction is currently available with the Layout and Prebuilt models `2024-02-29-preview``2023-10-31-preview` API and later releases except for the `US tax` models (W2, 1098s, and 1099s models).
302
322
303
323
### Query field extraction
304
324
@@ -308,7 +328,7 @@ For query field extraction, specify the fields you want to extract and Document
308
328
309
329
:::image type="content" source="media/studio/query-fields.png" alt-text="Screenshot of the query fields button in Document Intelligence Studio.":::
310
330
311
-
* You can pass a list of field labels like `Party1`, `Party2`, `TermsOfUse`, `PaymentTerms`, `PaymentDate`, and `TermEndDate`" as part of the `analyze document` request.
331
+
* You can pass a list of field labels like `Party1`, `Party2`, `TermsOfUse`, `PaymentTerms`, `PaymentDate`, and `TermEndDate` as part of the `analyze document` request.
312
332
313
333
:::image type="content" source="media/studio/query-field-select.png" alt-text="Screenshot of query fields selection window in Document Intelligence Studio.":::
314
334
@@ -318,8 +338,8 @@ For query field extraction, specify the fields you want to extract and Document
The Document Intelligence credit/debit card model uses powerful Optical Character Recognition (OCR) capabilities to analyze and extract key fields from credit and debit cards. Credit cards and debit cards can be of various formats and quality including phone-captured images, scanned documents, and digital PDFs. The API analyzes document text; extracts key information such as Card Number, Issuing Bank, and Expiration Date; and returns a structured JSON data representation. The model currently supports English-language document formats.
25
25
26
-
## Automated Credit/Debit Card processing
26
+
## Automated card processing
27
27
28
28
Automated Credit/Debit card processing is the process of extracting key fields from bank cards. Historically, bank card analysis process is achieved manually and, hence, very time consuming. Accurate extraction of key data from bank cards s is typically the first and one of the most critical steps in the contract automation process.
29
29
@@ -42,7 +42,7 @@ Document Intelligence v4.0 (2024-02-29-preview) supports the following tools, ap
Copy file name to clipboardExpand all lines: articles/ai-services/document-intelligence/concept-custom-lifecycle.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -26,7 +26,7 @@ With the v3.1 (GA) and later APIs, custom models introduce a expirationDateTime
26
26
27
27
With the v3.1 API, custom models introduce a new model expiration property. The model expiration is set to two years from the date the model is built for all requests that use a GA API to build a model. To continue to use the model past the expiration date, you need to train the model with a current GA API version. The API version can be the one that the model was originally trained with or a later API version. The following figure illustrates the options when you need to retrain an expiring or expired model.
28
28
29
-
:::image type="content" source="media/model-lifecycle.png" alt-text="Screenshot showing how to chose an API version to re-train a model.":::
29
+
:::image type="content" source="media/model-lifecycle.png" alt-text="Screenshot showing how to choose an API version and retrain a model.":::
Copy file name to clipboardExpand all lines: articles/ai-services/document-intelligence/concept-general-document.md
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -19,7 +19,7 @@ ms.author: lajanuar
19
19
:::moniker range="doc-intel-4.0.0"
20
20
21
21
> [!IMPORTANT]
22
-
> Starting with `Document Intelligence versions **2024-02-29-preview, 2023-10-31-preview** and going forward, the general document model (prebuilt-document) is deprecated. To extract key-value pairs, selection marks, text, tables, and structure from documents, use the following models:
22
+
> Starting with Document Intelligence versions **2024-02-29-preview, 2023-10-31-preview** and going forward, the general document model (prebuilt-document) is deprecated. To extract key-value pairs, selection marks, text, tables, and structure from documents, use the following models:
Copy file name to clipboardExpand all lines: articles/ai-services/document-intelligence/concept-incremental-classifier.md
+2-2Lines changed: 2 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,5 +1,5 @@
1
1
---
2
-
title: Document Intelligence support for Incremental Classifier Training.
2
+
title: Document Intelligence support for incremental classifier training
3
3
titleSuffix: Azure AI services
4
4
description: Incrementally train custom classifiers by adding new samples to existing classes or adding new classes.
5
5
author: laujan
@@ -74,7 +74,7 @@ The incremental classifier build request is similar to the [classify document bu
74
74
}
75
75
```
76
76
77
-
#### POST Response
77
+
#### POST response
78
78
79
79
All Document Intelligence APIs are asynchronous, polling the returned operation location provides a status on the build operation. Classifiers are fast to train and your classifier can be ready to use in a minute or two.
The Document Intelligence Marriage Certificate model uses powerful Optical Character Recognition (OCR) capabilities to analyze and extract key fields from Marriage Certificates. Marriage certificates can be of various formats and quality including phone-captured images, scanned documents, and digital PDFs. The API analyzes document text; extracts key information such as Spouse names, Issue date, and marriage place; and returns a structured JSON data representation. The model currently supports English-language document formats.
25
25
26
-
## Automated Marriage Certificate processing
26
+
## Automated marriage certificate processing
27
27
28
28
Automated marriage certificate processing is the process of extracting key fields from Marriage certificates. Historically, the marriage certificate analysis process is achieved manually and, hence, very time consuming. Accurate extraction of key data from marriage certificates is typically the first and one of the most critical steps in the marriage certificate automation process.
0 commit comments