|
10 | 10 | </prolog> |
11 | 11 | <body> |
12 | 12 | <p>It's been more than one year since we started working on the <xref |
13 | | - href="ai_positron.dita"/>, back when the <uicontrol>OpenAI GPT 3.5</uicontrol> AI engine was |
14 | | - released and we are currently on version 4.0.0 of the add-on. Our main direction (which we |
15 | | - still keep) was to provide powerful tools based on AI engines for technical documentation |
16 | | - writers (targeted mostly towards DITA XML users) and for our XSLT/Schematron developers. I've |
17 | | - always envisioned the AI set of tools as an exoskeleton, as a way to add more power to a |
18 | | - technical documentation writer's punch throughout the day. The table below contains a list of |
19 | | - concepts (mostly add-on specific features or feature categories), each with a short |
20 | | - description based on personal experience.</p> |
| 13 | + href="ai_positron.dita"/>, back when the <uicontrol>OpenAI GPT 3.5</uicontrol> AI |
| 14 | + engine was released, and we are currently on version 4.0.0 of the add-on. Our main |
| 15 | + direction (which we still keep) was to provide powerful tools based on AI engines for |
| 16 | + technical documentation writers (targeted mostly towards DITA XML users) and for our |
| 17 | + XSLT/Schematron developers. I've always envisioned the AI set of tools as an |
| 18 | + exoskeleton, as a way to add more power to a technical documentation writer's punch |
| 19 | + throughout the day. The table below contains a list of concepts (mostly add-on specific |
| 20 | + features or feature categories), each with a short description based on personal |
| 21 | + experience.</p> |
21 | 22 | <section><title>Eating our own dog food?</title>Are we using AI tools ourselves? Yes, I have |
22 | 23 | started using AI tools more and more to improve readability and fix grammar or logical |
23 | 24 | inconsistencies in articles like this one. I don't ask the AI to create new article |
|
47 | 48 | <p>So are all AI engines equal? Our predefined actions have and are actively tested with |
48 | 49 | <b>OpenAI</b> engines and with engines provided by <b>Anthropic Claude</b>. We |
49 | 50 | also tried to use engines like <b>Llama 3.1</b> or <b>Mistral</b> with our current |
50 | | - predefined actions but the results are not fantastic, the possibility to obtain |
| 51 | + predefined actions but the results are not fantastic, as the possibility to obtain |
51 | 52 | broken <b>DITA XML</b> content with these open source engines is much higher.</p> |
52 | 53 | </section> |
53 | 54 | <section><title>Predefined actions</title>Over time, we improved how we write prompts for |
|
57 | 58 | For each action, we try to give a clear prompt with lots of details so the AI engine can |
58 | 59 | produce the best results. We test our predefined actions using both manual and automatic |
59 | 60 | tests. We got the best results with Open AI GPT-4 engines and the engines provided by |
60 | | - <b>Anthropic Claude</b> like <b>Sonnet</b>.</section> |
| 61 | + <b>Anthropic Claude</b>, like <b>Sonnet</b>.</section> |
61 | 62 | <section><title>Custom actions</title>We wanted to give companies the flexibility to create |
62 | 63 | their own <xref |
63 | 64 | href="https://www.oxygenxml.com/doc/ug-editor/topics/ai_positron.html#ai_positron__section_rly_wp4_wxb" |
|
66 | 67 | aspects of implementing custom AI actions. The key to a successful action is to make it |
67 | 68 | as precise and detailed as possible.</section> |
68 | 69 | <section><title>Content generation</title> |
69 | | - <p>There are two main ways AI based content generation can be useful:</p> |
| 70 | + <p>There are two main ways AI-based content generation can be useful:</p> |
70 | 71 | <ul id="ul_njw_gbs_fdc"> |
71 | | - <li>Summarizing a large amount of text. For example predefined actions like |
| 72 | + <li>Summarizing a large amount of text. For example, predefined actions like |
72 | 73 | <uicontrol>Short Description</uicontrol> try to summarize a big piece of |
73 | 74 | content. These actions usually work very well.</li> |
74 | 75 | <li>Expanding a small amount of text. Content generation actions like <uicontrol>Add |
|
94 | 95 | screenshots.</section> |
95 | 96 | <section><title>Rewrite actions</title>We created useful actions to <b>Correct Grammar</b>, |
96 | 97 | <b>Improve Readability</b>, <b>Use Active Voice</b>, and <b>Improve Structure</b>. |
97 | | - These actions are applied to existing content, they have precise prompts which instruct |
| 98 | + These actions are applied to existing content, they have precise prompts that instruct |
98 | 99 | the AI to preserve the existing content as much as possible, and they are solid examples |
99 | 100 | of how the AI can help a writer during their daily work.</section> |
100 | 101 | <section><title>Review actions</title>At some point, we had the idea that the AI, instead of |
|
108 | 109 | or report on the <b>Readability</b> of a DITA XML topic. We also created actions to use |
109 | 110 | AI to better understand what questions a certain topic answers. In the future, using an |
110 | 111 | AI engine might also include asking the AI to find <b>logical inconsistencies</b> in |
111 | | - already written content, which may be too large for a person to fully digest |
| 112 | + already written content that may be too large for a person to fully digest |
112 | 113 | easily.</section> |
113 | 114 | <section><title>Translate actions</title>There are studies that say the translation |
114 | 115 | abilities in AI engines are better than <b>Google Translate</b> or <b>DeepL</b>. The |
|
152 | 153 | not use all the specialized DITA XML elements.</p></section> |
153 | 154 | <section><title>Third Party CMS support</title>Both Oxygen desktop and in-browser editing |
154 | 155 | tools can be integrated with various content management systems. Most AI predefined |
155 | | - actions should work when Oxygen is integrated with third-party CMSs. However, actions |
| 156 | + actions should work when Oxygen is integrated with a third-party CMS. However, actions |
156 | 157 | that write content on disk (like actions from the <b>Intelligent agents</b> category) |
157 | | - might not be able to create new topics on the CMS. Actions improved by using <b>RAG</b> |
158 | | - (retrieval augmented generation) might also not work well since the project content is |
159 | | - stored on the CMS. Oxygen desktop has an <uicontrol>Enable indexing for remote |
160 | | - resources</uicontrol> checkbox on its <uicontrol>Open/Find Resource</uicontrol> page |
161 | | - in <b>Preferences</b>. This allows our indexing engine to index content from the CMS. |
162 | | - But the in-browser WebAuthor editing tool does not yet support indexing CMS-specific |
163 | | - content.</section> |
| 158 | + might not be able to create new topics on the CMS. Actions that are improved by using |
| 159 | + <b>RAG</b> (retrieval augmented generation) might also not work well since the |
| 160 | + project content is stored on the CMS. Oxygen desktop has an <uicontrol>Enable indexing |
| 161 | + for remote resources</uicontrol> checkbox in the <uicontrol>Open/Find |
| 162 | + Resource</uicontrol> preferences page. This allows our indexing engine to index |
| 163 | + content from the CMS. But the in-browser Web Author editing tool does not yet support |
| 164 | + indexing CMS-specific content.</section> |
164 | 165 | <p>I hope this article, based on my experience working on AI editing tools and with AI |
165 | 166 | editing tools, will be useful to you. As always, you can leave feedback using the Oxygen |
166 | 167 | Feedback form, which appears below this published article.</p> |
|
0 commit comments