This shows you the differences between two versions of the page.
Both sides previous revision Previous revision | |||
wiki:autolit:tagging:smart_tags_guidelines [2024/01/27 19:47] jthurnham |
wiki:autolit:tagging:smart_tags_guidelines [2024/02/10 02:09] (current) kevinkallmes |
||
---|---|---|---|
Line 2: | Line 2: | ||
Smart Tag Recommendations is most effective when the tag hierarchy is kept as simple and concrete as possible. See below for tips on how to optimise smart tag recommendations: | Smart Tag Recommendations is most effective when the tag hierarchy is kept as simple and concrete as possible. See below for tips on how to optimise smart tag recommendations: | ||
+ | |||
+ | {{youtube> | ||
===== Configuring the Tag Hierarchy ===== | ===== Configuring the Tag Hierarchy ===== | ||
Line 7: | Line 9: | ||
**Mode** | **Mode** | ||
- | | + | * Form-based Tagging mode (default) works best |
- | | + | * Smart Tags work in Standard mode too, but does not have explicit Questions to feed to the LLM |
**Tags** | **Tags** | ||
Ensure you provide: | Ensure you provide: | ||
- | | + | |
+ | | ||
* Question Type (where relevant), | * Question Type (where relevant), | ||
* Question and Description text to give details beyond the | * Question and Description text to give details beyond the | ||
- | * Tag Name, | + | * Tag Name, |
* Aliases, if there are common synonyms or acronyms | * Aliases, if there are common synonyms or acronyms | ||
* All of the above, as well as the general hierarchy structure, is fed to the LLM | * All of the above, as well as the general hierarchy structure, is fed to the LLM | ||
Line 22: | Line 25: | ||
**Tag Name** | **Tag Name** | ||
- | * Limit to 4-5 words max, spell out acronyms, be careful with special characters ($, :, -, % etc.) only use if the term is commonly reported | + | * Limit to 4-5 words max, spell out acronyms, be careful with special characters ($, :, -, % etc.) only use if the term is commonly reported |
* Ensure terms are spelled correctly but don’t worry about variations (e.g. UK vs US English) | * Ensure terms are spelled correctly but don’t worry about variations (e.g. UK vs US English) | ||
- | |||
**Question type** | **Question type** | ||
+ | |||
* Use Single Apply where possible | * Use Single Apply where possible | ||
- | * Use Select types if total answers/ | + | * Use Select types if total answers/ |
* Currently only one recommendation is generated per question, which can be less helpful for Multiple Select. Single Apply is more likely to identify area of paper containing various answers | * Currently only one recommendation is generated per question, which can be less helpful for Multiple Select. Single Apply is more likely to identify area of paper containing various answers | ||
- | |||
**Questions** | **Questions** | ||
+ | |||
* Keep it simple and direct, don’t add too much detail | * Keep it simple and direct, don’t add too much detail | ||
* Present or past tense questions starting with “What” or “Where” tend to work best | * Present or past tense questions starting with “What” or “Where” tend to work best | ||
**Aliases (Optional)** | **Aliases (Optional)** | ||
+ | |||
* When the tag name cannot be identified (this includes both question and child tags for select question types), the LLM searches for aliases | * When the tag name cannot be identified (this includes both question and child tags for select question types), the LLM searches for aliases | ||
* This is most helpful to provide for field-specific terms | * This is most helpful to provide for field-specific terms | ||
**Contents (Tag Tables and Options)** | **Contents (Tag Tables and Options)** | ||
+ | |||
* Where possible, use free-text Tags; if you use a Tag Table, the LLM can scroll you to relevant text, but cannot assist with extraction | * Where possible, use free-text Tags; if you use a Tag Table, the LLM can scroll you to relevant text, but cannot assist with extraction | ||
* Similarly, if employing Options, the LLM can scroll you to relevant text, but will not select Options on your behalf | * Similarly, if employing Options, the LLM can scroll you to relevant text, but will not select Options on your behalf | ||
Line 58: | Line 63: | ||
Only after your tag hierarchy is configured should you toggle on Smart Tags: | Only after your tag hierarchy is configured should you toggle on Smart Tags: | ||
- | In Settings → Tag Recommendations → Choose Type: Smart. | + | In Settings → Tag Recommendations → Choose Type: Smart. By default, Smart Tags searches Full Texts but you may also wish to turn on Abstract Tag Recommendations, |
- | By default, Smart Tags searches Full Texts but you may also wish to turn on Abstract Tag Recommendations, | + | |
===== Recommendation Limits ===== | ===== Recommendation Limits ===== | ||
- | **Abstracts: | + | **Abstracts: |
- | + | ||
- | **Full Texts:** Smart Tags is currently able to run on 250 full texts or less, and runs only on included reports. | + | |
+ | **Full Texts: | ||
===== Using Smart Tags ===== | ===== Using Smart Tags ===== | ||
- | Smart Tags can take some time to generate. Feel free to click out of the modal and let it run in the background, you can check back in on its progress anytime using the icon. | + | Smart Tags can take some time to generate. Feel free to click out of the modal and let it run in the background, you can check back in on its progress anytime using the icon. |
As applicable, Abstract or Full Tag Recs are displayed as a right hand tab (in red below) wherever you can view a record: Screening & Tagging modules and Study Inspector. Ensure you have Questions open (in blue below), with a question selected for the Smart Tag Recommendation to be displayed (in green). Clicking the rec will auto-scroll to its suggested text excerpt within the Abstract/ | As applicable, Abstract or Full Tag Recs are displayed as a right hand tab (in red below) wherever you can view a record: Screening & Tagging modules and Study Inspector. Ensure you have Questions open (in blue below), with a question selected for the Smart Tag Recommendation to be displayed (in green). Clicking the rec will auto-scroll to its suggested text excerpt within the Abstract/ | ||
- | {{ : | + | {{ : |
===== Advanced: Bulk Apply Smart Tags ===== | ===== Advanced: Bulk Apply Smart Tags ===== | ||
- | While we encourage users to oversee the LLM by reviewing and either accepting or rejecting each recommendation, | + | While we encourage users to oversee the LLM by reviewing and either accepting or rejecting each recommendation, |
- | To do so, head to [[wiki: | + | To do so, head to [[:wiki: |
- | + | ||
- | <WRAP center round important 60%> | + | |
- | Bulk Actions is an advanced tool, make sure to familiarize yourself with the use of Bulk Actions and Study Inspector first before going ahead with this action! | + | |
- | </ | + | |
+ | <WRAP center round important 60%> Bulk Actions is an advanced tool, make sure to familiarize yourself with the use of Bulk Actions and Study Inspector first before going ahead with this action! </ | ||
===== FAQs ===== | ===== FAQs ===== | ||
- | **What data does OpenAI GPT 4 use? Are there restrictions in place on the use of my data? | + | **What data does OpenAI GPT 4 use? Are there restrictions in place on the use of my data? ** |
- | ** | + | |
When using Smart Tag Recommendations, | When using Smart Tag Recommendations, | ||
Line 98: | Line 97: | ||
First, make sure recommendations were successfully generated in Background Jobs. Then make sure you are on either the Abstract or Full Text page, you have the right hand Questions and Tag Recs tabs open and a Question is selected. Try going through a few questions in the form, it may be that fewer recommendations were generated than expected. Still nothing? You may need to simplify your hierarchy and regenerate recs. If that doesn’t work, contact support@nested-knowledge.com. | First, make sure recommendations were successfully generated in Background Jobs. Then make sure you are on either the Abstract or Full Text page, you have the right hand Questions and Tag Recs tabs open and a Question is selected. Try going through a few questions in the form, it may be that fewer recommendations were generated than expected. Still nothing? You may need to simplify your hierarchy and regenerate recs. If that doesn’t work, contact support@nested-knowledge.com. | ||
- | ** | + | **The recommendation is highlighting the same excerpt for multiple questions or isn’t applicable to my question, can I retrain the AI?** |
- | The recommendation is highlighting the same excerpt for multiple questions or isn’t applicable to my question, can I retrain the AI?** | + | |
OpenAI GPT 4 works to find the most succinct answer to the question, which means it often highlights the abstract/ | OpenAI GPT 4 works to find the most succinct answer to the question, which means it often highlights the abstract/ | ||
Line 106: | Line 104: | ||
No, after configuring new questions, Smart Tags must be regenerated in Settings. Note: each nest is limited to one regeneration due to LLM costs. If you need to regenerate Smart Tags after using your one regeneration, | No, after configuring new questions, Smart Tags must be regenerated in Settings. Note: each nest is limited to one regeneration due to LLM costs. If you need to regenerate Smart Tags after using your one regeneration, | ||
- | |||