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/cognitive-services/QnAMaker/Concepts/knowledge-base.md
+27-27Lines changed: 27 additions & 27 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,7 +1,7 @@
1
1
---
2
2
title: Knowledge base - QnA Maker
3
3
titleSuffix: Azure Cognitive Services
4
-
description: A QnA Maker knowledge base consists of a set of question/answer (QnA) pairs and optional metadata associated with each QnA pair.
4
+
description: A QnA Maker knowledge base consists of a set of question-and-answer (QnA) pairs and optional metadata associated with each QnA pair.
5
5
services: cognitive-services
6
6
author: diberry
7
7
manager: nitinme
@@ -13,25 +13,25 @@ ms.author: diberry
13
13
ms.custom: seodec18
14
14
---
15
15
16
-
# What is a QnA Maker Knowledge base?
16
+
# What is a QnA Maker knowledge base?
17
17
18
-
A QnA Maker knowledge base consists of a set of question/answer (QnA) pairs and optional metadata associated with each QnA pair.
18
+
A QnA Maker knowledge base consists of a set of question-and-answer (QnA) pairs and optional metadata associated with each QnA pair.
19
19
20
20
## Key knowledge base concepts
21
21
22
-
***Questions** - A question contains text that best represents a user query.
23
-
***Answers** - An answer is the response that is returned when a user query is matched with the associated question.
24
-
***Metadata** - Metadata are tags associated with a QnA pair and are represented as key-value pairs. Metadata tags are used to filter QnA pairs and limit the set over which query matching is performed.
22
+
***Questions**: A question contains text that best represents a user query.
23
+
***Answers**: An answer is the response that's returned when a user query is matched with the associated question.
24
+
***Metadata**: Metadata are tags associated with a QnA pair and are represented as key-value pairs. Metadata tags are used to filter QnA pairs and limit the set over which query matching is performed.
25
25
26
-
A single QnA, represented by a numeric QnA ID, has multiple variants of a question (alternate questions) that all map to a single answer. Additionally, each such pair can have multiple metadata fields associated with it: one key, and one value.
26
+
A single QnA, represented by a numeric QnA ID, has multiple variants of a question (alternate questions) that all map to a single answer. Additionally, each such pair can have multiple metadata fields associated with it: one key and one value.
When you ingest rich content into a knowledge base, QnA Maker attempts to convert the content to markdown. Read [this](https://aka.ms/qnamaker-docs-markdown-support)blog to understand the markdown formats understandable by most chat clients.
32
+
When you ingest rich content into a knowledge base, QnA Maker attempts to convert the content to markdown. Read [this blog](https://aka.ms/qnamaker-docs-markdown-support) to learn about the markdown formats that are understandable by most chat clients.
33
33
34
-
Metadata fields consist of key-value pairs separated by a colon**(Product:Shredder)**. Both key and value must be text-only. The metadata key must not contain any spaces. Metadata supports only one value per key.
34
+
Metadata fields consist of key-value pairs separated by a colon, such as Product:Shredder. Both the key and the value must be text-only. The metadata key must not contain any spaces. Metadata supports only one value per key.
35
35
36
36
## How QnA Maker processes a user query to select the best answer
37
37
@@ -41,27 +41,27 @@ The trained and [published](/azure/cognitive-services/qnamaker/quickstarts/creat
41
41
42
42
### Ranker process
43
43
44
-
The process is explained in the following table:
44
+
The process is explained in the following table.
45
45
46
46
|Step|Purpose|
47
47
|--|--|
48
48
|1|The client application sends the user query to the [GenerateAnswer API](/azure/cognitive-services/qnamaker/how-to/metadata-generateanswer-usage).|
49
-
|2|Qna Maker preprocessing the user query with language detection, spellers, and word breakers.|
50
-
|3|This preprocessing is taken to alter user query for best search results.|
51
-
|4|This altered query is sent to Azure Search Index, receiving the `top` number of results. If the correct answer isn't in these results, increase the value of `top` slightly. Generally a value of 10 for `top` works in 90% of queries.|
52
-
|5|QnA Maker applies advanced featurization to determine the correctness of the fetched Azure Search results for user query. |
49
+
|2|QnA Maker preprocesses the user query with language detection, spellers, and word breakers.|
50
+
|3|This preprocessing is taken to alter the user query for the best search results.|
51
+
|4|This altered query is sent to Azure Search Index, which receives the `top` number of results. If the correct answer isn't in these results, increase the value of `top` slightly. Generally, a value of 10 for `top` works in 90% of queries.|
52
+
|5|QnA Maker applies advanced featurization to determine the correctness of the fetched Azure Search results for the user query. |
53
53
|6|The trained ranker model uses the feature score, from step 5, to rank the Azure Search results.|
54
54
|7|The new results are returned to the client application in ranked order.|
55
55
|||
56
56
57
-
Features used include but are not limited to word-level semantics, term-level importance in a corpus, and deep learned semantic models to determine similarity and relevance between two text strings.
57
+
Features used include but aren't limited to word-level semantics, term-level importance in a corpus, and deep learned semantic models to determine similarity and relevance between two text strings.
58
58
59
59
## HTTP request and response with endpoint
60
-
When you publish your knowledge base, the service creates a REST-based HTTP **endpoint** that can be integrated into your application, commonly a chat bot.
60
+
When you publish your knowledge base, the service creates a REST-based HTTP endpoint that can be integrated into your application, commonly a chat bot.
61
61
62
62
### The user query request to generate an answer
63
63
64
-
A **user query** is the question that the end user asks of the knowledge base, such as,`How do I add a collaborator to my app?`. The query is often in a natural language format or a few keywords that represent the question, such as,`help with collaborators`. The query is sent to your knowledge from an HTTP **request** in your client application.
64
+
A user query is the question that the end user asks of the knowledge base, such as `How do I add a collaborator to my app?`. The query is often in a natural language format or a few keywords that represent the question, such as `help with collaborators`. The query is sent to your knowledge base from an HTTP request in your client application.
65
65
66
66
```json
67
67
{
@@ -82,9 +82,9 @@ You control the response by setting properties such as [scoreThreshold](./confid
82
82
83
83
Use [conversation context](../how-to/metadata-generateanswer-usage.md#use-question-and-answer-results-to-keep-conversation-context) with [multi-turn functionality](../how-to/multiturn-conversation.md) to keep the conversation going to refine the questions and answers, to find the correct and final answer.
84
84
85
-
### The response from a call to generate answer
85
+
### The response from a call to generate an answer
86
86
87
-
The HTTP **response** is the answer retrieved from the knowledge base, based on the best match for a given user query. The response includes the answer and the prediction score. If you asked for more than one top answer, with the `top` property, you get more than one top answer, each with a score.
87
+
The HTTP response is the answer retrieved from the knowledge base, based on the best match for a given user query. The response includes the answer and the prediction score. If you asked for more than one top answer with the `top` property, you get more than one top answer, each with a score.
88
88
89
89
```json
90
90
{
@@ -113,13 +113,13 @@ The HTTP **response** is the answer retrieved from the knowledge base, based on
113
113
```
114
114
115
115
### Test and production knowledge base
116
-
A Knowledge base is the repository of questions and answers created, maintained, and used through QnA Maker. Each QnA Maker tier can be used for multiple Knowledge bases.
116
+
A knowledge base is the repository of questions and answers created, maintained, and used through QnA Maker. Each QnA Maker tier can be used for multiple knowledge bases.
117
117
118
-
A knowledge base has two states - Test and published.
118
+
A knowledge base has two states: *test*and *published*.
119
119
120
-
The **test knowledge base** is the version that is being edited, saved, and tested for accuracy and completeness of responses. Changes made to the test knowledge base do not affect the end user of your application/chat bot. The test knowledge base is known as `test` in the HTTP request.
120
+
The *test knowledge base* is the version that's being edited, saved, and tested for accuracy and completeness of responses. Changes made to the test knowledge base don't affect the end user of your application or chat bot. The test knowledge base is known as `test` in the HTTP request.
121
121
122
-
The **published knowledge base** is the version that is used in your chat bot/application. The action of publishing a knowledge base puts the content of the Test knowledge base in the Published version of the knowledge base. Since the published knowledge base is the version that the application uses through the endpoint, care should be taken to ensure that the content is correct and well-tested. The published knowledge base is known as `prod` in the HTTP request.
122
+
The *published knowledge base* is the version that's used in your chat bot or application. The action of publishing a knowledge base puts the content of the test knowledge base in the published version of the knowledge base. Because the published knowledge base is the version that the application uses through the endpoint, make sure that the content is correct and welltested. The published knowledge base is known as `prod` in the HTTP request.
123
123
124
124
## Next steps
125
125
@@ -130,10 +130,10 @@ The **published knowledge base** is the version that is used in your chat bot/ap
0 commit comments